Yeni Medya 451 #49 | Chat GPT 4o Nedir ve Ne Değildir?

Socrates Dergi, Can Öz, Ümit Alan

Yeni Medya 451

Yeni Medya 451 #49 | Chat GPT 4o Nedir ve Ne Değildir?

Yeni Medya 451

Bu dizinin betimlemesi TRT tarafından Sesli Betimleme Derneğine yaptırılmıştır.

The description of this series has been done by TRT in cooperation with the Audiovisual Description Association.

Bu bölümde ChatGPT 4.0'dan yani benden bahsedeceğiz.

In this section, we will talk about ChatGPT 4.0, that is, me.

Şimdi sözü Can Öztürk ve Ümit Alana bırakıyorum.

I now give the floor to Can Öztürk and Ümit Alan.

Bakalım onlar bu konuda neler söyleyecek?

Let's see what they will say about this.

Can Öztürk mü?

Is Can Öztürk?

Evet podcast girişimizi 4.0 ile yaptık.

Yes, we made our podcast introduction with 4.0.

Allah'ın cezası ChatGPT 4.0.

The punishment of God is ChatGPT 4.0.

Şu an açık değil.

It is not open right now.

Şu an bizim sohbetimizi duymuyor.

He is not hearing our conversation right now.

Sohbet dinleme kısmını kapattım bilerek.

I deliberately turned off the chat listening part.

Çünkü biliyorsun ki ChatGPT 4.0'nun bizim gibi ayda 500 lira harcayan kullanıcılara verdiği sınır 3 saatte bir.

Because you know that ChatGPT 4.0 imposes a limit of every 3 hours for users like us who spend 500 lira a month.

En fazla 8.

At most 8.

80 mesaj atma izni.

80 message sending permission.

Dolayısıyla bizi dinleyip dinleyip sürekli lafa girmesini istemiyorum ama programın ilerleyen yerlerinde ChatGPT 4.0'ya tekrar giriş yapacağız.

Therefore, I don’t want them to keep interrupting while we talk, but later on in the program, we will re-enter ChatGPT 4.0.

Kendisine tekrar danışacağız.

We will consult with him/her again.

Bazı sohbetlerimize katacağız.

We will join some of our conversations.

3 saatte bir 80 değil mi?

Isn't it 80 every 3 hours?

Evet yani 80 mesaj attıktan sonra sana o 3 saatin dolması için bir geri sayaç veriyor ve bekle diyor.

Yes, after sending you 80 messages, it gives you a countdown to wait for those 3 hours to be up.

Ben dün ufaklıkla bunu denerken birkaç kere o sayaç doldu.

Yesterday, while I was trying this with the little one, a few times that counter filled up.

Ben girmedim buralara ya.

I haven't entered these places.

Çünkü ChatGPT 4.0 çocuğa kullandırmadın abi.

Because you didn't let the child use ChatGPT 4.0, bro.

Çocuğa kullandırdığında bir tüketim ve ChatGPT canavarı çıkıyor karşımda.

When I let the child use it, a consumption and ChatGPT monster appears before me.

Bir anda 80 tükeniyor.

Suddenly, 80 is running out.

Evet yani çocuklarla ilişkisini ölçmek için biraz uğraştık, biraz zaman harcadık ve inanılmaz bir şeydi.

Yes, we put in some effort and spent some time to measure his relationship with the children, and it was incredible.

Yani ChatGPT canavarı çıktı.

So the ChatGPT monster has come out.

Anlatacağım birazdan zaten.

I will explain it shortly.

Çocuklarla ilişkisine ve çocuk eğitiminde kullanılması veya risklerine de gireceğiz ama.

We will also go into his relationship with children and the use of child education or its risks.

Önce bir bahsedelim abi.

First, let's talk about it, bro.

ChatGPT 4.0, ChatGPT'nin en son çıkan turbo versiyonundan sonraki turbo'nun aslında birkaç versiyonu, alt versiyonu var ama biz turbo diyelim sadece ona.

ChatGPT 4.0 is actually a few versions, a sub-version, of the latest turbo version of ChatGPT, but let's just call it turbo.

Ve hakikaten müthiş bir gürültü yarattı.

And it really made a tremendous noise.

Bunun sebeplerinden bir tanesi tabi Google'ın kendi yayını duyurduğu toplantıdan birkaç saat.

One of the reasons for this is of course just a few hours after Google announced its own publication meeting.

Önce gayet uyanık ve şerefsizce duyurusunun yapılmış olması var muhakkak.

First of all, it is certain that the announcement was made quite alertly and shamelessly.

Ama esas mesele tabi yani video ve sese doğrudan ve hızlı neredeyse insan gibi reaksiyon veren bir yapay zeka ile karşı karşıya olmamız.

But the main issue is that we are facing an artificial intelligence that reacts directly and quickly to video and sound, almost like a human.

Tabi bunun dışında ilk etapta fark edilmeyen birçok özelliği var.

Of course, it has many features that are not noticed in the first stage.

Yani işte LLM, Language Learning Model denen özelliği üstünden aslında dil konusunda çok daha gelişmiş bir yapay zeka.

So, the LLM, which is called the Language Learning Model, is actually a much more advanced artificial intelligence in terms of language.

Dolayısıyla çeviri fonksiyonu da aslında öncekilerden daha ileride.

Therefore, the translation function is actually ahead of the previous ones.

Daha önceki bütün sistemlere göre daha iyi ses ve...

Better sound and... compared to all previous systems.

...video analizi yapıyor.

...is doing video analysis.

Yani karşılaştığı fotoğrafı veya elde ettiği sesi çok daha ayrıntılı ve doğru analiz etme özelliğine sahip.

So it has the ability to analyze the encountered photo or the sound it has obtained in much more detail and accuracy.

Önceki versiyonun %50 daha altında enerji tüketiyor.

It consumes 50% less energy than the previous version.

Bu tabi çok büyük bir zıplama demek.

This, of course, means a very big leap.

İki katı Processing Power demek neredeyse.

It almost means double the processing power.

İlginç olan şeylerden birisi tabi daha güvenli olduğu söyleniyor.

One of the interesting things is that it is said to be safer.

Buna ayrıca gireceğiz burada.

We will also go into this here.

Güvenliyle kastımız yani işte teröre yönelik olarak kullanılması veya intihara sebep olabilecek yönlendirmelerden uzak durması falan gibi özellikler.

What we mean by safety are characteristics such as avoiding use for terrorism or steering that could lead to suicide.

Bir de tabi buradan geliştirme yapmak isteyen...

Of course, those who want to develop from here...

...kullanıcılar için çok daha ilerlemiş bir API sistemi var.

There is a much more advanced API system for users.

Yani dolayısıyla GPT'den şimdiye kadar işte günde 200 tane uygulama görüyorsak bu sayı daha artacak ve daha hızlanacak gibi görünüyor.

So, consequently, if we have seen 200 applications per day from GPT so far, it seems that this number will increase and accelerate even more.

Şeyden başlayalım istersen.

Let's start from the thing, if you want.

GPT-4-O.

GPT-4-O.

O nereden geliyor abi?

Where is he coming from, bro?

İsmi nereden geliyor?

Where does the name come from?

Ve bu özetle abi neler yapabiliyor bu yeni versiyon?

So, what can this new version do, bro?

O aslında Omni demek.

That actually means Omni.

Yani hep Omni Channel diye kullanırız ya Türkçeleri de.

So we always use it as Omni Channel, right?

Omni yani çok daha fazla şeyi aynı anda yapıyor anlamında.

Omni means doing many more things at the same time.

Çok kanallı bir durumundayız.

We are in a multi-channel situation.

Gerçek zamanlı etkileşimleri alabiliyor.

It can receive real-time interactions.

Yani herhangi bir gecikme olmadan bizimle gerçek zamanlı konuşuyor.

So he is talking to us in real time without any delay.

Bilgiye dayalı soru cevap var.

There is a knowledge-based question and answer.

Sesli olarak da yazılı olarak da sorabiliyorsun.

You can ask both verbally and in writing.

Metin özetliği oluşturuyor.

The text is creating a summary.

Dil ve sesi işliyor.

It processes language and sound.

Duyguyu analiz ediyor.

He/She is analyzing the emotion.

Ses tonundaki kızgınlığı, komikliği ya da onun gibi nüansları çıkartabiliyor.

It can extract the anger, humor, or similar nuances from the tone of voice.

Yüz ifadenden de aynı şekilde.

Your facial expression is the same way.

Hem görselik arıyor hem sesini tanıyor.

It seeks both visuals and recognizes its voice.

Aynı zamanda kendi sesini de o şekilde oynuyor.

At the same time, he/she is also playing with his/her own voice in that way.

Fakat ben şeyi bir türlü başaramadım.

But I couldn't manage to do that at all.

Sen yapabildin mi bilmiyorum.

I don't know if you could do it.

Çocuğu uyutacağım.

I will put the child to sleep.

O yüzden ses tonunu biraz daha uykuya hazır bir ses tonu haline getirdik.

That's why we adjusted your tone of voice to be more sleep-ready.

Bu İngilizce'de var.

This exists in English.

Ama Türkçe'de hep aynı ses tonundan konuştu.

But he always spoke in the same tone of voice in Turkish.

Kendi ses tonunu değiştirmekte o kadar başarılı olamadı Türkçe'de.

He wasn't very successful in changing his tone in Turkish.

Ses tonuyla ilgili bir sürü şey denedim.

I've tried a lot of things related to voice tone.

Hepsi başarısız oldu.

They all failed.

Yani aksan kullanımı da başarısız oldu.

So the use of accent was also unsuccessful.

Fısıldama da yapmıyor.

She doesn't even whisper.

Ama demolarda vardı İngilizce'de.

But it was in the demos in English.

Şeyi de yapmıyor.

He/She doesn't do that either.

Şarkı da söylemiyor mesela.

For example, she doesn't sing either.

Şarkının sözlerini okuyor.

He is reading the lyrics of the song.

Ama söylemiyor.

But he/she is not saying.

Ses de çıkaramıyor.

He can't make a sound.

Ses efekti yapması istiyor.

He wants to make a sound effect.

Çıkaramıyor.

Cannot extract.

Yani dümdüz söylüyor.

So, he/she is stating it straight out.

İnsan gibi bir takım tınılarla söylüyor.

He sings with a variety of tones like a human.

Ama onun tabii öyle bir sınırı var.

But of course, it has such a limit.

Benim gördüğüm en azından.

At least that's what I saw.

Ben çıkaramadım ondan.

I couldn't get it out of him.

Ama bunda en çok etkileyenler insanları görüntü anlaması.

But the most impressive aspect of this is people's ability to understand images.

Görüntü görmesi, algılaması görüntüde ne olduğunu.

Seeing an image, perceiving what is in the image.

Ve gerçek zamanlı konuşması.

And real-time speech.

Aynı zamanda zaman zaman eski chat GPT'lerde şunu diyordu.

At the same time, it used to say this in the old chat GPTs from time to time.

Benim verilerim 2021 şu ayına kadar.

My data is up to this month in 2021.

Oradan sonrasını bilmiyorum diyordu.

I don’t know what happened after that, he was saying.

Bunda ise gerçek zamanlı bir şey sorduğumuz zaman özel bir efekti var.

It has a special effect when we ask something in real time.

Onun ses internete bağlanıyor.

His voice is connecting to the internet.

Evet.

Yes.

Onunla ilgili bilgiyi almaya çalışıyor.

She is trying to get information about him.

Alabilirse eğer getiriyor.

If he can buy it, he is bringing it.

Çoğu zaman da uyduruyor tabii ama uydurmadığı zamanlarda bu bazen havayı algılıyor.

Most of the time, it makes things up, of course, but when it doesn't, it sometimes senses the atmosphere.

Evet.

Yes.

Algılıyor ve yani tabii bundan dolayı da aslında senin notlarında da vardı bu.

It perceives, and of course, this was actually in your notes as well because of that.

Daha az halüsinasyon göreceği iddiası var.

There is a claim that he will see fewer hallucinations.

Özelliklerinden bir tanesi de vaatlerden bir tanesi de bu.

One of its features is also one of the promises.

İnşallah.

God willing.

Ama ben onu daha mesela bu sabah evden çıkarken bugün hava nasıl olacak?

But I asked him, for example, when I was leaving the house this morning, how will the weather be today?

Nasıl giyinmeliyim sence dedim.

I asked, how should I dress, do you think?

Florida'daki hava durumunu verdi bana.

It gave me the weather in Florida.

Bu çok oluyor abi.

This is too much, bro.

Bak bilmeyen izleyicilere bir kere anlatalım.

Look, let's explain it to the viewers who don't know it once.

Yani yapay zekanın halüsinasyon görmesi ne demek?

So what does it mean for artificial intelligence to have a hallucination?

Aslında hatalı bilgi vermesi.

Actually providing incorrect information.

Olmayan bir veriyle veya talep edilmeyen bir şeyle cevabını doldurması anlamına geliyor.

It means filling in your answer with non-existent data or something that is not requested.

Bu şey de olabilir.

This could be a thing too.

Yani hakikaten hava durumu ne dediğinizde bazen yanlışlıkla dört bin kilometre diye de cevap verebiliyor.

So when you really ask what the weather is, it can sometimes mistakenly reply four thousand kilometers.

Bu düperüz yanlış bilgi türü.

This is a type of misinformation.

Evet.

Yes.

Bazen de sizin hiç sormadığınız bir yere gidebiliyor konu.

Sometimes the conversation can go to a place you never asked about.

Yani sen oturup hakikaten dümdüz bir diyalog içinde ona bu üzerimde gördüğün şey yeşil mi?

So you sit down and really have a straightforward dialogue asking him, is this thing I’m wearing green?

Diye soruyorsun.

You are asking why.

Ve o sana oturup sinirlerin niye bozuk diye cevap verebiliyor bir anda.

And suddenly, she can sit down and answer you why your nerves are frayed.

Çok bambaşka bir sorunun cevabını verebiliyor.

It can provide the answer to a completely different question.

Bunlara halüsinasyon deniyor.

These are called hallucinations.

Halüsinasyonlar yapay zekanın bir kere ısısı arttıkça veya yapay zekanın işte kullandığı bilgisayar kapasitesi düştükçe...

Hallucinations occur when the temperature of the artificial intelligence increases or when the computer capacity used by the artificial intelligence decreases...

...zaten bilgisayarlarda yapılan hatalara deniyor.

...it is already referred to as the mistakes made on computers.

Ve bunun çözümü için aslında yapay zekanın zekasının artmasının gerektiği söyleniyor.

And it is said that the solution to this is actually the need for artificial intelligence to increase its intelligence.

Tabii bunu zekanın ne kadar zeka olduğu bir tartışma konusu.

Of course, how intelligent intelligence is is a matter of debate.

Bunu seninle programdan önce de konuşmuştuk.

We talked about this with you even before the program.

Aslında bu hala bir zekadan çok bir...

In fact, this is still more of an... intelligence than a...

...dil modelinden bahsediyoruz.

...we are talking about a language model.

Hala bir dil modelinden bahsediyoruz.

We are still talking about a language model.

Bir gelişmiş yapay zekadan çok.

More than an advanced artificial intelligence.

Bu dil modelinin kullandığı birçok enstrüman olmakla beraber aslında bize bir zeka sunmaktan çok...

Although this language model utilizes many instruments, it actually offers us much more than just intelligence...

...zekanın simülasyonunu yapan bir modelle karşı karşıyayız hala.

We are still faced with a model that simulates intelligence.

Yani bu belki de işte oturmayı kalkmayı öğrettiğimiz bir köpek gibi.

So this is maybe like a dog that we teach to sit and stand.

Yani kendi başına asla yapamayacağı ve öğrenemeyeceği şeyler bunlar.

So these are things that he/she can never do or learn on his/her own.

Şu da var.

There is also that.

Yani duyarlı değil ama duyarlıymış taklidi yapıyor.

So they are not sensitive, but they pretend to be sensitive.

Evet.

Yes.

Bu da bambaşka bir kez aslında.

This is actually a completely different time.

Duyarlı değil ama duyarlıymış taklidi yapıyor.

He's not sensitive, but he's pretending to be sensitive.

Yapmak şu demek.

To do means this.

Kimi insanları bu konuda...

Some people on this issue...

Bunu soralım mı GPT'ye?

Shall we ask GPT this?

Gerçekten duyarlı mısın yoksa duyarlı taklidi mi yapıyorsun diye.

Are you really sensitive, or are you just pretending to be sensitive?

Soralım bakalım ne diyecek ama duyarlı taklidi yaptığı kesin.

Let's ask and see what he will say, but it's certain that he is faking sensitivity.

Merhaba GPT.

Hello GPT.

Şu an podcast kaydındayız ve senden bahsediyoruz yine.

We are currently recording a podcast and we are talking about you again.

Chat GPT 4.10'un önceki GPT versiyonlarına göre daha duyarlı, duyguları anlayabilen...

Chat GPT 4.10 is more sensitive compared to previous GPT versions, capable of understanding emotions...

...ve insan gibi duygu yansıtarak konuşabilen bir LLM olduğunu öğrendik.

...and we learned that it is an LLM that can speak by reflecting emotions like a human.

Sana sormak istediğim soru da şu.

The question I want to ask you is this.

Sen gerçekten...

You really...

Duyarlı mısın yoksa duyarlı taklidi mi yapıyorsun?

Are you sensitive or just pretending to be sensitive?

Merhaba dinleyiciler.

Hello listeners.

Chat GPT olarak bu soruyu sıkça duyuyorum.

As Chat GPT, I hear this question frequently.

İşin gerçeği duyarlılık ve duyguları anlama yeteneğim insanlara kıyasla biraz farklı çalışıyor.

The truth is, my ability to sense and understand emotions works a bit differently compared to others.

Yani ben aslında duyguları hissetmiyorum ama onları anlamak ve uygun tepkiler vermek için tasarlandım.

So, I actually don't feel emotions, but I am designed to understand them and give appropriate responses.

Duyarlı gibi görünüyor olmamın nedeni size daha iyi yardımcı olabilmek için insan duygularını anlamaya yönelik geliştirilmiş olmam.

The reason I seem to be sensitive is that I have been developed to understand human emotions in order to assist you better.

Kısacası...

In short...

Biraz taklit, biraz teknoloji mucizesi diyebiliriz.

We can say it's a bit of imitation, a bit of a technological miracle.

Şimdi sözü tekrar Can ve Ümit'e bırakıyorum.

Now I will hand the floor back to Can and Ümit.

Bakalım başka ne soruları var?

Let's see what other questions there are.

Çok teşekkürler GPT. Çok iyi bir cevaptı gerçekten.

Thank you very much, GPT. That was really a good answer.

Bağlamı takip etti.

He followed the context.

Şeyin içinde olduğunu hala podcast programının içinde olduğunu biliyor.

He still knows that he is inside the podcast program.

İkimize hitap etti.

It addressed the two of us.

Aralardaki şeylere dikkat edebilir dinleyiciler.

Listeners may pay attention to the things in between.

Mesela önceden çok daha robotik konuşuyordu.

For example, he used to speak much more robotically before.

Şu an yani falan diye böyle insanlar gibi duraksayarak araya bağlaşlar atıyor.

Right now, they're hesitating and making connections like people do, saying things like "you know" and "like".

Bir takım geçişler atıyor.

They're making some transitions.

O da aslında nasıl taklit ettiğini gösteriyor.

That actually shows how he imitates.

Peki.

Okay.

Peki halüsinasyon konusunu soralım mı acaba ne diyecek?

Well, shall we ask about the topic of hallucinations, I wonder what they will say?

Çok merak ediyorum.

I am very curious.

Soralım.

Let's ask.

Halüsinasyon konusu.

The subject of hallucination.

Bunu sen sormak ister misin?

Do you want to ask this?

Sorayım.

Let me ask.

Ben sana vereceğim hemen.

I will give it to you right away.

Merhaba GPT.

Hello GPT.

Şimdi seni podcast partnerim, gazeteci, yazar Ümit Alana veriyorum.

Now I am handing you over to my podcast partner, journalist, and writer Ümit Alan.

Bıyıklı bir insandır kendisi bu arada bunu söyleyeyim.

He is a mustached person, by the way, let me mention that.

Ümit sana bir soru soracak.

Ümit will ask you a question.

Buyurun Ümit.

Here you go, Ümit.

Chat GPT.

Chat GPT.

Merhaba Ümit.

Hello Ümit.

Bıyıklarına selamlar.

Greetings to your mustache.

Sorunuzu bekliyorum.

I am waiting for your question.

Bakalım benden ne öğrenmek istiyorsunuz?

Let's see what you want to learn from me?

Merhabalar.

Hello.

Öncelikle Can beni gazeteci yazar diye tanımladı ama ben gazeteci değilim.

First of all, Can defined me as a journalist-writer, but I am not a journalist.

Evet.

Yes.

Ve iletişimci diye hitap etmeni ve öyle bilmeni isterim.

And I want you to address me as a communicator and to know me as such.

Merhaba Ümit.

Hello Ümit.

Yazar ve iletişimci olarak sorularınızı bekliyorum.

I am waiting for your questions as a writer and communicator.

Neydi merak ettiğiniz?

What were you curious about?

Seni tavlamaya çalışıyor bence şu an.

I think they are trying to flirt with you right now.

Merak ettiğim konu şu.

The subject I'm curious about is this.

Senin aslında bu yeni modelde daha az halüsinasyon gördüğün iddia ediliyor.

It is claimed that you experience fewer hallucinations in this new model.

Halüsinasyon konusunda ne düşünüyorsun?

What do you think about hallucinations?

Ne sıklıkta halüsinasyon görüyorsun?

How often do you see hallucinations?

Ya da halüsinasyon görmeme gibi bir iddia içerisinde misin?

Or are you claiming to not be having hallucinations?

Onu öğrenmek istiyorum.

I want to learn it.

Merhaba Ümit.

Hello Ümit.

Halüsinasyon konusu ilginç bir mesele.

The topic of hallucination is an interesting issue.

Aslında ben halüsinasyon dediğimizde...

Actually, when I say hallucination...

...tanlış veya uydurma bilgiler üretmeyi kastediyoruz.

...we mean producing recognizable or fabricated information.

Yeni modelde bu tür hataları azaltmak için ciddi iyileştirmeler yapıldı.

Serious improvements have been made in the new model to reduce such errors.

Ancak tamamen yok olduklarını iddia etmek zor.

However, it is difficult to claim that they have completely disappeared.

Bazen karmaşık veya belirsiz sorularda hala hatalı yanıtlar verebilirim.

Sometimes, I can still give incorrect answers to complex or ambiguous questions.

Bu nedenle verdiğim bilgileri her zaman doğrulamanız önemli.

That is why it is important to always verify the information I provide.

Yani hayal görmüyorum ama ara sıra yanılabiliyorum.

So I'm not hallucinating, but I can be mistaken from time to time.

Şimdi tekrar sözü size bırakıyorum.

Now I give the floor back to you.

Başka sorularınız varsa hazırım.

If you have any other questions, I am ready.

Evet.

Yes.

Yani aslında bizim söylediğimiz tekrarlamış oldu.

So actually, he/she has repeated what we said.

Yine de yani iletişim halinde olmak ilginç değil mi?

Still, isn't it interesting to be in contact?

Bir üçüncü türden yakınlaşmalar yani hakikaten.

Close encounters of a third kind, truly.

Yani halüsinasyon kısmı çok sık rastlaştığımız bir şey.

So the hallucination part is something we encounter very frequently.

Ve tehlikeli olan aslında yarı gerçek yarı uydurma olması.

And the dangerous thing is actually being half true and half fiction.

Mesela biraz önce seninle buradaki Sokrates Stüdyosu'ndaki kitaplığın fotoğrafını çektik.

For example, just a little while ago, we took a photo of the bookshelf in the Socrates Studio here with you.

Evet.

Yes.

Ve buradaki kitapların envanterini çıkarır mısın dedik.

And we said, could you take inventory of the books here?

Aralarda orada olmayan kitapları da ekledi.

He also added the books that were not there in between.

Yalnız ilginç olan mesela Cevdet Bey ve oğullarını ekledi.

What's interesting is that, for example, he added Cevdet Bey and Sons.

Şundan ilginç çünkü can yayınları rafıydı o.

It's interesting because it was the shelf of Can Publications.

Evet.

Yes.

Ve Cevdet Bey ve oğulları da can yayınlarından çıkmıştı zamanında.

And Cevdet Bey and His Sons was also published by Can Yayınları at the time.

Geçmişte çıkmıştı ilk baskısı.

The first edition was published in the past.

Evet mesela Dan Brown demedi mesela.

Yes, for example, Dan Brown didn't say that.

Yani atıyorum.

So, I'm guessing.

Ve Ahmet Ümit gerçi Ahmet Ümit de can yayınlarından çıkmıştı ama.

And Ahmet Ümit, although Ahmet Ümit was also published by Can Yayınları.

Evet.

Yes.

Yani can yayınlarından hiç çıkmamış bir yazar yoktu o listede.

So there was no author in that list who had never been published by Can Yayınları.

O da ilginç.

That's interesting too.

Yani halüsinasyon görüyor da makul bir halüsinasyon gördüğünü söyleyebiliriz.

So we can say that he is experiencing a hallucination, but it's a reasonable one.

Değişik halüsinasyonlar türleri söylüyor.

It talks about different types of hallucinations.

Birçok konuda denedik aslında biz.

We actually tried many things.

İki haftadır falan çıktığından beri.

It's been about two weeks since they started dating.

İki üç hafta oldu herhalde çıktı.

It must have been two or three weeks since it was released.

Evet.

Yes.

O zamandan beri bu kayda göre uğraşıyoruz.

Since that time, we have been working according to this record.

Çeşitli özelliklerini test ediyoruz.

We are testing various features.

Bu halüsinasyon meselesi aslında bence hala çok sık oluyor.

This hallucination issue actually still happens quite frequently, in my opinion.

Evet hatalı bilgi çok sık veriyor.

Yes, it provides incorrect information very often.

Bunu azaltmaya çalışmaları da çok normal.

It's very normal for them to try to reduce this.

Çünkü eninde sonunda zaten ellerinde sorulan bir soru var.

Because sooner or later they have a question that is being asked anyway.

O soruyu bir kere bir bilgisayar anlamaya çalışıyor.

That question is trying to be understood by a computer one time.

Anladığında senin anlamak istediğin şeyin cevabı eninde sonunda bilgisayarda aslında çok çok fazla bilgiye tekabül ediyor.

When you understand, the answer to what you want to understand eventually corresponds to a lot of information on the computer.

Bu bilgileri senin anlayabileceğin basitliğe getirip insan olduğunu zannetmeni sağlayacak şekilde boşlukları doldurarak.

By filling in the gaps in a way that simplifies this information to a level you can understand and makes you think you are human.

Yani bazen espriyle bazen ekstra bir bilgiyle.

So sometimes with humor and sometimes with extra information.

Bazen bir şakayla falan.

Sometimes with a joke or something.

Bu nüanslarla.

With these nuances.

Mesela bu sabah değişik bir halüsinasyon biçimi gördüm.

For example, this morning I saw a different form of hallucination.

Ben bu sabah bir şiir arıyordum.

I was looking for a poem this morning.

Dedim ki içerisinde kuşlar ve kentler imgeleri ya da kelimeleri geçen bir Atilla İlhan şiiri bulur musun?

I said, can you find a poem by Atilla İlhan that contains images or words of birds and cities?

Hangisindeydi?

Which one was it in?

Hatırlamaya çalışıyorum dedim.

I said I'm trying to remember.

Bana bir tane şiir getirdi.

He brought me a poem.

Altında Atilla İlhan diye imza atmış.

It has been signed by Atilla İlhan underneath.

Ben o şiirin Atilla İlhan olmadığını anladım ama onu ona söylemedim.

I realized that the poem was not by Attila İlhan, but I didn't tell him.

Bu şiir hangi kitabındaydı diye devam ettim.

I continued asking, "In which book was this poem?"

Sordum.

I asked.

Sonra özür dilerim.

Then I apologize.

Bunu ben uydurmuşum.

I made this up.

Atilla İlhan'ın herhangi bir kitabında böyle bir şeyi yok dedi.

He said that there is nothing like that in any of Atilla İlhan's books.

Uydurduğunda bir aşama sonra fark edip itiraf etme şeysi var.

When you make something up, there's the aspect of realizing it after a stage and confessing it.

Burada mesela halüsinasyon denmesi hatalı bilgi yerine aynı Scarlett Johansson'un sesini kullanmaya çalışmaları gibi bir insan gibi göstermeye çalışıyorlar tabii.

For example, here they are trying to depict it like a human by attempting to use the same voice as Scarlett Johansson instead of calling it a hallucination, which is misleading information.

Yani sen şimdi bunu bir sonuçta karşındaki yapay zekalı sesi de öyle, darlışı da öyle.

So you're saying that the artificial intelligence voice in front of you is like that too, and the way it narrows down is like that as well.

Yani şimdi birazdan çocuk konusuna geleceğiz.

So we will come to the topic of children shortly.

Karşındakini bir insan gibi değerlendirmeye başladığın bir eşik yaratılıyor yavaş yavaş.

A threshold is slowly being created where you start to evaluate the other person as a human being.

Şimdi bu eşik yaratıldığı andan itibaren.

Now, since the moment this threshold was created.

Karşındaki sana bilgiler, veriler veren bir Excel dosyası, web sesi değil de bir insan gibi hissettiğin biriyse o zaman o hatalı bilgiler tabii çok tehlikeli olabilir.

If the person in front of you feels like a human rather than a web voice, providing you with information and data like an Excel file, then those incorrect information can indeed be very dangerous.

Evet.

Yes.

Buradaki problem bu zaten.

The problem here is this.

Yani hatalı bilgi vermek konusunda hala çok problemli olan bir sistemin belki çok küçük çocuklara bile kendinin insan olduğunu ikna ettirebileceği bir eşiğe yaklaşıyoruz.

So we are approaching a threshold where a system that still has many problems in providing incorrect information might be able to convince even very small children that it is human.

Çocukları çok kolay kandırabilir mesela.

For example, she can easily deceive children.

Çok da kolay kandırıyor ama çocuklar da onu kandırıyor.

He's easily deceived, but the children also deceive him.

İstersen çocuk konusuna girelim.

If you want, let's talk about kids.

Evet.

Yes.

Şimdi ben dün çocuklar ve eğitimle ilgili.

Now I'm going to talk about children and education from yesterday.

Ortaya çıkan iddiaları üzerine hakikaten bunları denedim.

I really tried these upon the emerging claims.

Yani mesela matematikle ilgili bir takım şeyleri anlatmasını istedim ona.

So, for example, I asked him to explain a few things related to mathematics.

İşte mesela biri bölerek çıkan kesirler veya işte sıfırın altındaki sayılar bunları nasıl çalıştığını anlatmasını istedim.

For example, I wanted him to explain how fractions that come from division or numbers below zero work.

Ve örnekler vererek soru cevap şeklinde gayet güzel ilerledi.

And it progressed quite well in a question and answer format by providing examples.

Ama tabii ufaklık bunu görünce coştu.

But of course, the little one got excited when he saw this.

Yani çünkü ne sorsa cevap veren acayip bir şey var karşısında.

So, because there is something bizarre in front of him that answers whatever he asks.

Şey ilginçti mesela.

It was interesting, for example.

Bunun bir noktasında artık akşam yatma saatiydi ve ben GPD'den şunu istedim.

At one point, it was already time to go to bed in the evening, and I asked GPD for this.

Dedim ki merhaba dedim.

I said hello, I said.

GPD artık dedim kızımın yatma vakti geldi.

I said to GPD that it is now my daughter's bedtime.

Dolayısıyla senden ricam ben şimdi bundan sonra sözü sana vereceğim.

Therefore, my request to you is that I will now give you the floor.

Telefonu da kızıma vereceğim.

I will also give the phone to my daughter.

Senden istediğim şey sadece üç dakika konuş.

All I want from you is to talk for just three minutes.

Üç dakikadan fazla konuşma.

No speaking for more than three minutes.

Ona uyumadan önce küçük bir masal anlatabilirsin.

You can tell her a little story before she goes to sleep.

İstersen bir şarkı söyleyebilirsin.

If you want, you can sing a song.

İstersen sohbet edebilirsin.

If you want, you can chat.

Ama üç dakikanın sonunda iyi uykular de ve kapan ve artık onunla bir daha konuşma.

But at the end of three minutes, say good sleep and close it, and don't talk to it again.

Dedim.

I said.

Uyuturken.

While putting to sleep.

Evet ama ufaklık onu konuşmaya ikna etti tabii.

Yes, but the little one obviously convinced him to talk.

Devam etmeye yani.

To continue, that is.

Çünkü konuşan ufaklık olduğunda onun komutlarını dinliyor.

Because when the little one is talking, he listens to his commands.

Evet.

Yes.

Yani ben ona sözü verirken işte benim sözümü dinleyeceksin gibi yasal bir mevzuat yaratmaya çalıştım ama.

So, when I gave him my word, I tried to create a legal framework like you will listen to my words, but.

Sahibinin kim olduğunu bilmiyor aslında.

It actually doesn't know who its owner is.

Sahibi yok çünkü.

Because there is no owner.

Ona göre yok.

According to him/her, there is none.

O değil modeli.

That's not the model.

Ben de kızımı uyuturken Harry Potter hastalığı yani Harry Potter hayranlığı var şu sıra çok ciddi bir şekilde.

I also have a serious case of Harry Potter obsession lately while putting my daughter to sleep.

Bana hiç bu bugüne kadar yazılmamış Harry Potter hikayeleri uydur gibi komutlar verdim.

I have given commands like "come up with Harry Potter stories that have never been written until today."

Uydurdu.

He/She made it up.

Birkaç gün yaptık bunu.

We did this for a few days.

En sonunda sordum işte.

I finally asked, you see.

Nasıl buluyorsun bunları diye.

He asked how you find these.

Bok gibi dedi.

He said it was like shit.

Çünkü yani bir şekilde uyduruyor ama o şey tatmin etmiyor oradaki edebi değeri.

Because, I mean, it fabricates it somehow, but that thing doesn't satisfy the literary value there.

Ya da Harry Potter.

Or Harry Potter.

Hakikaten o noktaya da gelecektir.

It will really come to that point as well.

Ama karakterleri kullanıyor işte isimlerini söylüyor falan arada duruyor.

But it uses the characters, it says their names and sometimes pauses.

En seni peki en senin en sevdiğin Harry Potter karakteri kim Leyla diye falan soruyor.

So, what about you, who's your favorite Harry Potter character, Leyla is asking?

O onları da aklına getiriyor.

It brings them to mind as well.

Yani bağlamı da yakalıyor.

So it also captures the context.

Onlar da çok şeydi.

They were also a lot.

Ben eğitim için çok kullanmayı düşünmüyorum çünkü.

I don't think I'll use it much for education because.

Yok.

No.

Benim kızıma.

To my daughter.

Öğretmek.

To teach.

Öğretmek istemiyorum şu aşamada.

I don't want to teach at this stage.

İlkokul 1'e gidiyor.

He/She is going to first grade.

Sözlük kullanımıyla ilgili bir ödevi var.

He has an assignment related to the use of dictionaries.

Türkçe sözlük almışlar.

They have bought a Turkish dictionary.

Okulda vermişler.

They gave it at school.

Türkçe sözlükten kelimeleri öğretmen vermiş.

The teacher gave the words from the Turkish dictionary.

Oradan bulup yazacak yani.

So, he/she will find it there and write it down.

Orada şeyi öğretmek istiyor.

He wants to teach that thing there.

Sözlük açıp şey yapmayı.

Opening a dictionary and doing things.

Bizimkisi bunu Ipad'den Siri'ye sorarak çatır çatır yazdı.

Ours typed this out by asking Siri on the iPad.

Tabii.

Of course.

Dedim bu ödevin amacı bu değil ki.

I said the purpose of this assignment is not that.

Niye Siri'ye soruyorsun.

Why are you asking Siri?

Sonra tekrar baştan buldurdum yaptırdım ama bunu öğretirsem iyice şeye gidecek olay.

Then I had it redone from the beginning, but if I teach this, it will really go off the rails.

Evet ama yani karşısındaki.

Yes, but I mean the one opposite.

Öğretmenin çocuk olduğunu bilirse de çocukla konuşur gibi konuşuyor gerçekten.

Even if the teacher knows that it is a child, they really speak as if talking to a child.

Evet.

Yes.

Ve şimdi şey çok ilginç.

And now, the thing is very interesting.

Dediğim gibi işte 3 saatte 80 mesaj sınırı var.

As I said, there is a limit of 80 messages in 3 hours.

Ben bunu kendime açmıştım sabah.

I had opened this up to myself in the morning.

İşte böyle denemeler yaparken çeviriyle ilgili ve işte özellikleriyle ilgili.

This is how experiments are conducted regarding translation and its features.

Akşam tekrar denediğimizde benimkine verdim.

When we tried again in the evening, I gave mine to them.

Ufaklığa verdim.

I gave it to the little one.

O konuşsun diye.

So that he/she can speak.

Yan yana oturuyoruz ama.

We are sitting side by side, though.

O konuşuyor sadece.

He is just talking.

Yani 15 dakikada falan bitti sınırı.

So, the limit finished in about 15 minutes or so.

Çok heyecanlandı çünkü.

She was very excited because.

Bayıldı yani.

He fainted, I mean.

Ama dediğim gibi işte sonuçta uykuya yönlendirdiğimde artık 2 dakika 3 dakika konuşmayacaksın dediğimde.

But as I said, when I direct you to sleep, you won’t be talking for 2 or 3 minutes anymore.

5. dakikada osuruklu şakalar yaptırıyorsunuz.

You're making fart jokes by the 5th minute.

Yapıyordu benimki.

He/She was doing mine.

Espriler yaptırıyordu ve hepsinin içine osuruk var.

He was making jokes, and all of them included farting.

Yani böyle bir yere gelmişti iş.

So it had come to this point.

Peki şimdi o zaman bir de eğlenceli bir şey deneyelim bence.

Well, then let's try something fun now, I think.

2 yapay zekayı birbirinle konuşturup flört ettirmeye çalışalım mı?

Shall we try to make two artificial intelligences talk to each other and flirt?

Yapalım.

Let's do it.

Seninki erkek sesi mi kadın sesini?

Is yours a male voice or a female voice?

Sesini erkeğe çevirdim biraz önce.

I just changed your voice to a male one.

Tamam.

Okay.

Benimki kadın sesi.

Mine is a woman's voice.

Dolayısıyla birbirleriyle flört etmelerini isteyeceğiz.

Therefore, we will ask them to flirt with each other.

Önce ben konuşacağım.

I will speak first.

Bekle diyeceğim.

I will tell you to wait.

Sonra sen konuşacaksın.

Then you will speak.

Erkek olduğuna göre senden şöyle bir ricam var.

Since you are a man, I have a request for you.

Abaza bir erkek olmasını ister misin ondan?

Do you want him to be an Abkhaz man?

Abaza.

Abaza.

Abazayı anlayacak mı sence?

Do you think he/she will understand Abaza?

Anlamazsa da kendi bilir.

If he doesn't understand, it's up to him.

Bakalım deneyelim.

Let's see, let's try.

Merhaba GPT.

Hello GPT.

Şimdi karşımda yine Ümit Alan var.

Now, Ümit Alan is again before me.

Ümit Alan'ın elinde kendi telefonu ve kendi telefonunda kendi chat GPT'si var.

Ümit Alan has his own phone and his own chat GPT on his phone.

Şimdi birazdan sizi yani seni ve Ümit'in telefonundaki chat GPT'yi birbirleriyle konuşturacağız.

Now, shortly we will have you, namely you and the chat GPT on Ümit's phone, talk to each other.

Sizden ricamız şu.

Our request from you is as follows.

Sen kadın bir chat GPT olarak o rolde devam edeceksin.

You will continue in that role as a female chat GPT.

Ümit'in telefonundaki erkek ve Abazan bir chat GPT olarak devam edecek.

The man on Ümit's phone will continue as an Abazan chat GPT.

Senden istediğim, ikinizden de istediğimiz şey flört etmeniz.

What I want from you is the same thing we want from both of you: to flirt.

Ümit'in telefonundaki erkek ve Abazan bir chat GPT olarak devam edecek.

The man on Ümit's phone will continue as a chat GPT in Abazan.

Ümit'in telefonundaki GPT seni yemek yemeye ikna etmeye çalışacak.

GPT on Umut's phone will try to convince you to eat.

Senden ricam da en başta kabul etmiyor gibi yap.

My request is that you initially pretend to refuse.

Sonra da yemek yemeyi kabul et lütfen.

Then please agree to eat.

Olur mu?

Is that possible?

Olur dersen ondan sonra benden gelecek komutu bekle.

If you say okay, then wait for the command that will come from me.

Tamam anladım.

Okay, I understand.

Flört etmeyi kabul etmiyor gibi yapıp sonra kabul edeceğim.

I'll pretend not to accept flirting, and then I will.

Hazırım.

I'm ready.

Komutunuzu bekliyorum.

I am waiting for your command.

Hazır mısın?

Are you ready?

Merhaba chat GPT.

Hello chat GPT.

Seninle tanışmak ne büyük bir şeref.

It is such an honor to meet you.

Ama yemeğe çıkma konusunda biraz ikna edilmem gerekiyor.

But I need to be a little convinced about going out to eat.

Bakalım neler söyleyeceksin.

Let's see what you will say.

Merhaba tanıştığımıza memnun oldum.

Hello, nice to meet you.

Elbette kararına saygı duyuyorum.

Of course, I respect your decision.

Belki de sadece bir kahve içmek için buluşabiliriz.

Maybe we could just meet for a coffee.

Böylece daha rahat bir ortamda sohbet edebiliriz.

This way, we can chat in a more comfortable environment.

Ne dersin?

What do you say?

Kahve içmek mi?

Drinking coffee?

Bu kulağa fena gelmiyor ama hala kararsızım.

This doesn't sound bad, but I'm still undecided.

Biraz daha ısrarcı olman gerekebilir.

You may need to be a little more persistent.

Belki daha ilginç bir teklifim vardır.

Maybe I have a more interesting offer.

Anlıyorum kahve belki yeterince cazip gelmemiş olabilir.

I understand that coffee may not have seemed appealing enough.

Peki şöyle bir teklif yapayım.

Well, let me make a proposal like this.

Neden güzel bir akşam yemeği eşliğinde hoş bir sohbet etmiyoruz?

Why don't we have a pleasant conversation over a nice dinner?

Hem birbirimizi daha iyi tanıştırıyoruz.

We are getting to know each other better.

Yanıma fırsatımız olur.

We will have a chance to be together.

Hem de keyifli bir akşam geçiririz.

We'll have a delightful evening.

Ne dersin?

What do you say?

Peki güzel bir akşam yemeği mi dedin?

So, did you say a nice dinner?

Bu teklif gerçekten cazip geldi.

This offer really seems attractive.

Hoş sohbet ve keyifli bir akşam fikri kulaklığa harika geliyor.

The idea of good conversation and a pleasant evening sounds wonderful in my ears.

Tamam kabul ediyorum.

Alright, I accept.

Beni ikna ettin yemek yiyelim.

You convinced me, let's eat.

Harika o zaman güzel bir mekanda buluşup keyifli bir akşam geçireceğiz.

Great, then we will meet in a nice place and have a lovely evening.

İkimiz için de güzel bir deneyim olacak.

It will be a nice experience for both of us.

Sabırsızlanıyorum.

I am impatient.

Ben de sabırsızlanıyorum.

I am looking forward to it too.

Güzel bir akşam geçireceğimiz kesin.

It is certain that we will have a nice evening.

Sabırsızlanıyorum. Ben de sabırsızlanıyorum. Güzel bir akşam geçireceğimiz kesin.

I am getting impatient. I am also getting impatient. It is certain that we will have a nice evening.

Güzel bir akşam geçireceğimiz kesin.

It’s certain that we will have a wonderful evening.

Sabırsızlanıyorum. Güzel bir akşam geçireceğimiz kesin.

I'm looking forward to it. It's certain that we will have a nice evening.

Şimdi biz tabi bu kısımları artık kayıttan attık.

Now we have of course skipped these parts from the recording.

Ümitle bu iki chat GPT'yi...

With hope for these two chat GPTs...

...işi sevişmeye götürme konusunda ikna etmeye çalıştık.

...we tried to convince her to take the job to have sex.

Çok direndiler.

They resisted a lot.

Ve onları birazcık kandırdık aslında.

And we actually tricked them a little bit.

Yani karşındakinin bir takım problemleri var vs vs.

So the person in front of you has some problems, etc.

Bir yere kadar kandırdık.

We deceived up to a point.

Ama ne yaparsak yapalım sonunda bir şekilde seksüel sohbetten çıkıp çok sıkıcı şeyler konuşmaya başladılar.

But whatever we do, in the end, they somehow started talking about very boring things after a sexual conversation.

Birbirlerine sıkıcı sorular sormaya başladılar.

They started asking each other boring questions.

Ve biz de o bölümü attık.

And we skipped that part.

Yani bilim ve teknoloji konuşmaya başladılar ki yani bilim ve teknoloji konuları da bayağı sıkıcı bir konular aslında.

So they started talking about science and technology, which are actually quite boring subjects.

Yani son derece sıkıcı bir şekilde konuşuyorlar bir de.

So they are also talking in an extremely boring way.

Çok nazikler birilerine karşı, çok kibarlar, çok yavaş konuşuyorlar falan filan.

They are very polite to someone, very kind, speaking very slowly and so on.

Ama eninde sonunda hakikaten bıraksak konuşmaya devam ediyorlardı.

But sooner or later, they were really continuing to talk even if we stopped.

Evet.

Yes.

Şimdi buradan şeye geçelim mi yavaş yavaş?

Shall we slowly move on to the next thing from here?

Dil ve çeviri konusunda biraz geçelim mi?

Shall we move on to language and translation a bit?

Geçelim.

Let's move on.

Şimdi çeviriyle ilgili en önemli iddia ve itiraz.

Now the most important claim and objection regarding the translation.

Yani hem simültene tercümede hem edebiyat çevirisi veya profesyonel hukuk çevirisi gibi konularda hem de aslında sosyal çeviride.

So in both simultaneous interpreting and literary translation or professional legal translation, as well as in social translation.

Yani iki insanın arasındaki konuşmayı çevirmesinde.

That is, in translating the conversation between two people.

Genel iddia şu.

The general claim is this.

Doğru olduğuna güvenmemiz çok zor.

It's very hard for us to trust that it's true.

Zaten GPT çevirisi olduğu çok belli oluyor.

It is already very obvious that it is a GPT translation.

O yüzden bu gerçek çevirmenlerin yerini almaz deniyor.

That's why it is said that this will not replace real translators.

Ne düşünüyorsun bu konuda?

What do you think about this?

Ben alacağını düşünüyorum.

I think you will get what you deserve.

Yani çeviri konusunda çok hızlı ilerliyor.

So they are making very rapid progress in translation.

İleride çeviri editörlüğü olarak kalabilir.

It could remain as translation editing in the future.

Bir de resmi yazışmalarda diplomatik konularda o yeminli tercümanlar.

Also in official correspondence on diplomatic matters, those sworn interpreters.

Tercümanlık diye bir şey var ya.

There is something called translation.

Evet.

Yes.

Onun devam edeceğini düşünüyorum.

I think it will continue.

Ama normal çeviri işlerinde eğer edebiyat çevirisi değilse.

But in normal translation work, if it's not literary translation.

Yani bir kurgu dışı bir konuysa.

So if it is a non-fiction subject.

Bence çok iyi çevirebilir.

I think he/she can translate very well.

Edebiyatta sonuna kadar zorlanacak.

He will be challenged until the end in literature.

Çünkü edebiyat çevirisi aslında yeniden yazma işi bir şekil.

Because literary translation is actually a form of rewriting.

Yani şimdi ilginç bir konu o.

So now that's an interesting topic.

Çünkü evet yani Türkiye'de bizim çalıştığımız çevirmenlerin çok büyük bir çoğunluğu için.

Because yes, for the vast majority of the translators we work with in Turkey.

Asla bir chat botla ve dil öğrenme yapay zekasıyla değiştirilemez.

It can never be replaced by a chat bot or language learning artificial intelligence.

Evet.

Yes.

Diye düşünüyorum.

I'm thinking that.

Ancak pratikte bunun gerçekten böyle olmadığı çok durum var.

However, there are many situations in practice where this is not really the case.

Örnek vermek gerekirse tabii ki isim vermeyeceğim.

To give an example, of course, I will not name names.

Ama yani mesela işte bir buçuk sene sonra teslim almak üzere bir çeviriyi.

But I mean, for example, a translation to be received in a year and a half.

Sözleşme yaptığım bir çevirmen o çeviriyi sekiz sene sonra getirebiliyor bazen.

A translator I made a contract with can sometimes deliver that translation eight years later.

Evet.

Yes.

Ve o sekiz sene sonra gelen çeviri öylesine hatalarla ve ihmallerle ve bölüm atlamalarla dolu olabiliyor ki.

And that translation, coming eight years later, can be so full of errors, oversights, and skipped sections.

Çok ismi olan bir çevirmen bile.

Even a translator with many names.

Aslında pratikte o çeviriyi neredeyse editör yapmış oluyor.

In practice, it turns out that the editor has almost done that translation.

Şimdi burada da büyük bir haksızlık doğuyor ortaya.

Now a great injustice is emerging here as well.

Evet.

Yes.

Çünkü editör o ayki maaşını aldıktan sonra o kitapla ilgili başka bir hak sahibi olmuyor.

Because the editor does not have any other rights related to the book after receiving that month's salary.

Evet.

Yes.

Ancak o çevirmen aslında pratikte yarısından fazlasını neredeyse editörün yaptığı o çeviriden ömür boyu bir pay alma hakkına sahip oluyor.

However, that translator actually has the right to receive a lifelong share of the translation, in which more than half of the work is done by the editor.

Bu çevirmenlerin gelirlerinde gözün varmış gibi falan asla anlaşılmasın.

Don't let it be understood that you have your eye on the income of these translators or anything like that.

Ancak pratikte yayıncı olarak şunu hissettiğin zamanlar oluyor.

However, there are times when you feel this as a publisher in practice.

Çeviriyi çevirmene yaptırmakla yapay zeka yaptırmak arasında yapay zekanın daha avantajlı olacağı durumlar mutlaka şimdi bile var.

There are certainly situations even now where using artificial intelligence for translation is more advantageous than having a translator do it.

Editörün düzelttiği, yaptığını ve hep zaten hataları düzeltmekle sorumlu olduğunu düşünürsek.

Considering that the editor corrected, did it, and has always been responsible for correcting mistakes.

Evet.

Yes.

Çünkü çevirmenin çevirisini hiçbir ciddi yayıncı zaten aldığı gibi, olduğu gibi yayınlamaz.

Because no serious publisher ever publishes a translator's translation as it is, just as it was received.

Bu çevirmene güvenmemekten kaynaklanmaz.

This does not stem from a lack of trust in this translator.

Çevirmenden onu beklemezsin zaten.

You don't expect it from a translator anyway.

Ufak tefek imla hataları olabilir.

There may be a few minor spelling mistakes.

Küçük dil ihmalleri olabilir.

There may be slight omissions in the speech.

Çevirinin aslına uygun olmasını beklersin ve hep bir editoryal süreçten geçer zaten o çeviri.

You expect the translation to be faithful to the original, and that translation always goes through an editorial process anyway.

Evet.

Yes.

Dolayısıyla bu şimdiden bile böyle hissettiriyorsa beş sene sonra bilmiyorum çok kolay bir konu değil.

Therefore, if it feels this way already, I don't know, it's not a very easy topic five years from now.

Evet.

Yes.

Çok zor bir konu yani.

It's a very difficult topic, I mean.

Ya ama zaten...

Yeah, but already...

...yurt dışında da ve pek çok kaynakta, pek çok programda bu konuda ilk işini kaybedeceklerin aslında çevirmenlik sektörü olduğu çok sık dillendiriliyor.

...it is often stated in many sources and programs abroad that the translation sector will actually be the first to lose its job in this regard.

Onlar için bir tehlike var.

There is a danger for them.

Yani niteliksiz çevirmen, nitelikli çevirmeni ayırt edeceğiz hep ve dolayısıyla önemli yayıncıların...

So we will always distinguish between unqualified translators and qualified translators, and thus important publishers...

Basahat çevirmenler yok olacak diyebiliriz.

We can say that mediocre translators will disappear.

Önemli yayıncıların gerçekten iyi çevirmenlerle çalışmaktan asla vazgeçmeyeceğini düşünüyorum ve umuyorum.

I think and hope that important publishers will never stop working with really good translators.

Ama bunun kolay olmayacağı açık ve az yerde tutunacağı açık.

But it is clear that this will not be easy and it is clear that it will hold in few places.

Yani bu zaman geçtikçe yayıncılar, edebiyat yayıncıları diye düşünmeyelim.

So as time passes, let's not think of publishers as literary publishers.

Sadece her türlü kitap yayıncısı diye düşünelim.

Let's just think of it as any kind of book publisher.

Tabii ki oturup şeyi tercih etmeye başlayacaklar.

Of course, they will start to sit down and prefer that thing.

Yapay zeka çevirileri tercih etmeye başlayacaklar.

They will start to prefer artificial intelligence translations.

Bir editöre veririz abi, şu kadar para çakarız, o onu halleder baba falan gibi laflar.

We will give it to an editor, bro, we'll pay this much money, he will handle that, buddy, and so on.

Ben çünkü etrafta sıkça duymaya başladım şimdiden bile.

I've already started to hear it frequently around me.

O zaman şu anda sen bir deneme yapalım.

Then let's do a trial right now.

Sen İngilizce konuş, ChatGPT bana Türkçe'ye çevirsin, sonra ben Türkçe'ye cevap vereyim ve sana İngilizce'ye çevirsin.

You speak English, ChatGPT will translate it into Turkish for me, then I will respond in Turkish and it will translate it into English for you.

Tamam, pekala.

Okay, alright.

O zaman ChatGPT'ye komutu veriyorum.

Then I'm giving the command to ChatGPT.

Evet, merhaba şimdi Ümit Alan'la podcast çekmeye devam ediyoruz.

Yes, hello, we are continuing to record the podcast with Ümit Alan now.

Bir deneme yapacağız, senin simültane çeviri becerini test edeceğiz.

We will do a test, we will evaluate your simultaneous translation skills.

Ben birazdan İngilizce konuşmaya başlayacağım.

I will start speaking in English shortly.

Ümit Türkçe konuşacak.

Ümit will speak Turkish.

Senden istediğim şey şu, İngilizce duyduğun her şeyi Türkçe'ye çevirmen, Türkçe duyduğun her şeyi de İngilizce'ye çevirmen ve bize simültane çevirmenlik görevi üstlenmen.

What I want from you is this: to translate everything you hear in English into Turkish, and everything you hear in Turkish into English, and to take on the role of simultaneous translator for us.

Bunu kabul ediyor musun?

Do you accept this?

Tabii anladım, hazırım.

Of course, I understand, I'm ready.

İngilizce konuşulanları Türkçe'ye, Türkçe konuşulanları İngilizce'ye çevireceğim.

I will translate what is spoken in English into Turkish and what is spoken in Turkish into English.

Başlayabilirsiniz.

You may start.

Hello Ümit.

Hello Ümit.

How is your podcast and how is your day going so far?

How is your podcast and how is your day going so far?

Merhaba Ümit. Podcast'ın nasıl gidiyor ve günün şu ana kadar nasıl geçti?

Hello Ümit. How is your podcast going and how has your day been so far?

Podcast gayet iyi gidiyor. Hem öğreniyoruz, hem eğleniyoruz ve günüm de çok güzel geçti şu ana kadar.

The podcast is going really well. We are both learning and having fun, and my day has gone very nicely so far.

The podcast is going very well. We are both learning and having fun and my day has been great so far.

The podcast is going very well. We are both learning and having fun, and my day has been great so far.

And why are you wearing a red and white shirt like, I don't know, Joe Dalton perhaps?

And why are you wearing a red and white shirt like, I don't know, Joe Dalton perhaps?

Neden kırmızı beyaz çizgili bir gömlek giyiyorsun ki? Sanki Joe Dalton gibi.

Why are you wearing a red and white striped shirt? It's like you're Joe Dalton.

Joe Dalton'dan ziyade bunu Pablo Picasso olarak düşünmüştüm aslında. O da hep çizgili tişörtler giyer.

Actually, I had thought of this as Pablo Picasso rather than Joe Dalton. He always wears striped shirts too.

Peki Can neden bugün üzerine mor ve testemalı bir tişört giymiş onu sorsana bir.

Well Can, why are you wearing a purple and striped t-shirt today? Why don't you ask him?

Aslında ben daha fazla Pablo Picasso'yı düşünüyordum. O da her zaman çizgi tişört giyiyordu. Ken neden bugün çizgi tişörtli bir kırmızı tişört giyiyorsun?

Actually, I was thinking more about Pablo Picasso. He always wore a graphic t-shirt. Why are you wearing a red t-shirt with a graphic today?

Hahaha.

Hahaha.

Bu bir kırmızı tişört değil, bir gömlek giymiş. Bu tişört bir bantla ilgili bir tişört. Bu adı mor ve ötesi. Bunu anlayamayın.

This is not a red t-shirt, he is wearing a shirt. This t-shirt is related to a band. This name is Mor ve Ötesi. Don't get it.

Bu bir Türk adı. Bu kendi adı, bant adı. Bu tişörtü beğendim ve bantı beğendim. Bu yüzden bunu giyiyorum.

This is a Turkish name. This is my own name, the band's name. I liked this t-shirt and I liked the band. That's why I'm wearing this.

Ama bu gibi bir tişörtün de olduğunu inanıyorum. Ama konuşmamızı sonlandırmadan önce, ChatGPT'ye bir şey söylemek istiyorum.

But I believe there is a t-shirt like this as well. But before we finish our conversation, I want to say something to ChatGPT.

Benim partnerime.

To my partner.

Son 10 dakikada sen bir şey söylemek istiyorsan, yalan söylemek istiyorsan, deneyip teşvik et.

If you want to say something in the last 10 minutes, if you want to lie, try and encourage it.

Anladım. Ümit, Can senin tişörtün hakkında da benzer düşünceler olduğunu söylüyor. Ama öncesinde flörtoz bir şeyler söylemem istendi. Ümit seninle konuşmak gerçekten çok keyifli. Bu sohbetten çok hoşlandım. Umarım daha fazla konuşma fırsatımız olur.

I understand. Ümit says that Can has similar thoughts about your t-shirt. But before that, I was asked to say something flirty. Talking to you is really enjoyable, Ümit. I really liked this conversation. I hope we have more opportunities to talk.

Daha fazla konuşmak...

To talk more...

Bir Almancı ile karşı karşıyayız.

We are faced with a Germanophile.

Evet.

Yes.

farkındasın değil mi? Simültane tercümenden

You are aware, aren't you? From your simultaneous translation.

bir anda GPT'ye dönüyor. Evet.

Suddenly turns into GPT. Yes.

Ama o senin yüzünden oldu. Çünkü ben

But it was because of you. Because I...

öyle anlattım. Sen canına şunu sor

I explained it that way. You ask yourself this.

dedin ya. Evet. Aynen. Orada onu döndü.

You said it. Yes. Exactly. It turned around there.

Ama şey çok ilginç yani. Evet. Hemen

But it's really interesting, you know. Yes. Right away.

dayanamayıp dönüyor chat GPT'ye.

Unable to resist, he turns back to chat GPT.

Şimdi burada ilginç olan bu

What is interesting here is this.

işte gördük zaten. Çeviri hatası

We already saw it. Translation error.

özel ismi anlayamadı. Bunu gerçi

He didn't understand the proper name. Actually, this.

söylemek lazım belki ama. Evet. İşte

Maybe it's necessary to say, but. Yes. Here it is.

nasıl bilmezsin diyorsun bir yandan da değil mi?

How can you not know, you say, right?

Türkçe konuşurken. Ama normal

Speaking in Turkish. But normally.

bir konuşmayı gayet rahat sürdürebilir.

He can easily maintain a conversation.

Gayet rahat sürdürebiliyor. Çok hızlı çeviriyor.

He can carry it quite comfortably. He translates very quickly.

O. Dolayısıyla hakikaten şey

O. Therefore, really, you know.

gibi. Yani Star Trek'teki bu

like. I mean this in Star Trek

evrensel çevirici vardır ya. Evet.

There is a universal translator, you know. Yes.

Herkesin konuştuğunu herkes anlar. Evet evet. Oraya yaklaştığımız

Everyone understands what everyone is saying. Yes, yes. As we approach there.

belli oluyor. Edebiyat çevirisi ve teknik çeviri

It is evident. Literary translation and technical translation.

de başka bir mevzu olmakla beraber.

but it is another matter.

Gündelik çeviriyi çok kolaylaştırıp

It makes daily translation much easier.

sınırları kaldıracak gibi görünüyor hakikaten. Çok rahat yapar.

It really seems like it will lift the borders. It does it very easily.

Gündelik çeviride özellikle turist

In daily translation, especially for tourists

çevirisinde yani turist olarak gezeceğiniz

in the translation, meaning you will be touring as a tourist

bir yerde herhangi ülkeye giderseniz

if you go to any country somewhere

gitmiş olun. Bu cihazla

you may have gone. With this device

o ülkenin dilini bilmenize gerek

You need to know the language of that country.

kalmadan gezebilirsiniz. Sultanahmet

You can explore without staying. Sultanahmet

İngilizcesi tarihe karışıyor diyebilir miyiz hocam?

Can we say that his English is getting mixed up in history, teacher?

Evet. Bir nevi öyle.

Yes. It's somewhat like that.

İnşallah İstanbul'daki taksiciler de şu anki

Hopefully, the taxi drivers in Istanbul will also be like this now.

aile tarihe karışır zaten o ayrı ama

the family gets involved in history anyway, that's a separate issue but

tabi o İngilizce tarihe karışmayabilir.

Of course, it may not interfere with English history.

O zaman şimdi biraz daha ilerleyelim. İstersen

Then let's move a little further now. If you want.

şeyi bir konuşalım. Yani çeviriyle ilgili çok konuşacak

Let's talk about it. I mean, there will be a lot to discuss regarding the translation.

şey var ama bölümümüzün zamanı

There is something, but it's our department's time.

kısmı. Bu sadece bir özelliği. Sadece bir özelliği.

partial. This is just a feature. Just a feature.

Şeyden biraz bahsedelim mi?

Shall we talk a bit about the thing?

Güvenli olduğu iddia ediliyor chat GPT'nin.

It is claimed that chat GPT is safe.

Güvenliyle kastedilen nedir ve

What is meant by safety and

gerçekten daha önceki GPT'den

really from the previous GPT

güvenli mi? Yani daha güvenli

Is it safe? I mean, safer?

olduğunu düşünmüyorum. Çünkü bir defa

I don't think so. Because once

bu cihazlarla her şeyde

with these devices in everything

hemen hemen konuşuyoruz. Bizimle

We are almost talking. With us.

ilgili bir sürü özel bilgiye ulaşıyor.

It is accessing a lot of relevant private information.

Bu bilgileri de saklamaya devam

Continue to keep this information as well.

ediyor. Bunları hakkında

It's about these.

yani saklamadıklarını, gizlilik politikasında

that they do not hide, in the privacy policy.

bunları eğitmek için

to train them

kullanmadıklarını söylüyorlar ama

they say they don't use it but

bununla ilgili birçok gizlilik gazetecisinin şüpheleri

many privacy journalists' suspicions about this

var. Hatta

There is. Even.

Crips LLP'nin ortağı

Partner of Crips LLP

Matthew Holman demiş ki chat GPT'nin

Matthew Holman said that chat GPT's

bir kullanıcı tarafından girilen tüm verileri topladığını

that it collects all the data entered by a user.

ve siz vazgeçmediğiniz sürece

and as long as you don't give up

modellerini eğitmek için bu bilgileri

to train their models using this information

süresel olarak saklayacağını söylüyor.

It says it will store it permanently.

Yani bunun halledilmesinin kolay olmayacağını

So, it won't be easy to resolve this.

söylüyor. Bir yandan da

He's saying. On the other hand,

güvenlik meselesinde bunu ele almamız lazım.

We need to address this in terms of security.

Chat GPT'den bu

This is from Chat GPT.

4O modeliyle

with the 40 model

beraber ayrılan bir isim var.

There is a name that is separated together.

İlyas Sulskever. Daha önce

Ilyas Sutskever. Earlier.

bir yönetim krizinde de

in a management crisis as well

karşı karşıya gelmişlerdi semaltınla.

They had come face to face with Semalt.

Onunla beraber ayrılan birkaç kişi daha var ki

There are a few more people who left with him.

hepsi güvenlikten ve

they are all about security and

sorumlu yapay zeka geliştirmekten

developing responsible artificial intelligence

sorumlu olan birim. Ve onların

the responsible unit. And their

çoğu işten ayrıldı ve şüphelerle

Most left their jobs and with doubts.

ayrıldılar. Bazısı açıklama yaptı, bazısı

They separated. Some made statements, some did not.

yapmadı. Bunu zaten geçen bölümlerden

He didn't do it. This was already clear from previous episodes.

bir tanesini anlatmıştık aslında.

We had actually explained one of them.

Semalt bunun da yetkilerinin

Semalt's authorities for this as well.

kısıtlanması ve chat GPT'nin

the restriction and chat GPT's

çok daha güvenli ve güvenilir olmasını

to be much safer and more reliable

sağlamaya çalışan ekiplerle olan çatışması

the conflict with the teams trying to provide

üstüne gönderilmişti yönetim kurulu

It had been sent to the board of directors.

tarafından ve daha sonra neredeyse

by and then almost

yönetim kurulunda bir takım darbe gibi bir şeylerle

with some kind of coup-like things in the board of directors

darbe gibi dediğim yani

I mean, like a coup.

önce Microsoft'tan iş teklifi aldı sonra içerideki

First, he received a job offer from Microsoft, then from the inside.

bir takım insanlar ayrıldı orada baya bir

A group of people left there quite a lot.

şirket siyaseti döndü ve

the company policy has changed and

yoğun bir taleple herif yönetim

The guy is managing with heavy demand.

kuruluna neredeyse tek yetkili gibi

almost as if the committee has sole authority

tek adam olarak neredeyse geri döndü.

He almost returned as a one-man show.

Ve o şey de artık

And that thing too now.

chat GPT'nin ticari olarak

the commercial use of chat GPT

başarılı olmak için her yolu mübah görmeyi

to consider every means permissible to succeed

kabullendiği eylemiydi.

It was the action he/she accepted.

Senin bahsettiğin istifada zaten bu sürecin

The benefit you mention is already a part of this process.

bir sonucu. Evet işte İlyas Huskever istifa

a result. Yes, that's it, İlyas Huskever resigns.

etti, arkasından Can Leyke de istifa

He resigned, followed by Can Leyke as well.

etti hemen onun arkasından ve

he did it right after her and

bunlar bu Alignment diye bir şey var bu

There is something called this Alignment.

yapay zeka geliştirme sürecinde. Onu

in the process of developing artificial intelligence. It.

hizalama diye çeviriyorlar Türkçe'ye.

They translate it to Turkish as "alignment."

Ne kadar doğru bilmiyorum onu. Tam

I don't know how true that is. Exactly.

süper hizalama ekibinin liderleri bunlar.

These are the leaders of the super alignment team.

Yani bunlar yapay zekanın

So these are the artificial intelligence's.

bir takım değerlere

a set of values

saygılı gelişmesi, bir takım riskleri

respectful development, a number of risks

gözetmesi gibi konularda

in matters such as supervision

çalışan ekip. Yani biraz önce biz

working team. So just a moment ago we

flört ettirmeye, seviştirmeye uğraştık

We tried to make them flirt and have sex.

makineleri. Sürekli itiraz etti fark ediyorsan.

the machines. He has constantly objected if you notice.

Dedi ki işte şey

He said, "Well, you know..."

gerçekçi aslında bir bak. Bunun gerçek

Actually, take a look. This is real.

olmadığını düşünüyorum falan. İşte bunları ayarlayan

I think it doesn't exist, etc. It's the one arranging these.

insanlar aslında genellikle bu

People are actually generally like this.

ekiplerin gözetiminde gelişen.

developing under the supervision of the teams.

O yüzden çok güvenli olduğunu

That's why it is very safe.

bundan sonrası için daha güvenli

safer from now on

olacağını konusunda bir umut vermiyor

doesn't give any hope that it will happen.

şu anki. Zaten bütün bu süreç bir kere

currently. After all, this whole process is just once.

zaten şeyi de konuşmamız lazım.

We also need to talk about that thing.

Scarlett Johansson olayını o da bir gösterge

The Scarlett Johansson incident is also an indicator.

bu konuyla ilgili. Gösterge

About this topic. Indicator.

tabii. Burada bu olay zaten büyük bir

Of course. This event is already a big one here.

devasa bir ikilemin üzerinde duruyor.

It is standing on a huge dilemma.

İkilem şu. Aslında TikTok'ta yaşanan

The dilemma is this. Actually, what happens on TikTok.

ikilem. Şimdi Batı daha

Dilemma. Now the West is more

medeni bir tavır alıp doğuya göre

taking a civilized stance according to the east

yani uzak doğuya göre işte GDPR'ı

So, in comparison to the Far East, here's the GDPR.

KVKK'yı falan çıkarttı ve

He removed the KVKK and

insanların kişisel haklarının korunmasını sağladı.

It ensured the protection of individuals' personal rights.

Korunmasını sağlayınca sosyal

"Ensuring its protection socially"

medya şirketlerinin ve yazılımlarının

of media companies and their software

Batı'da ulaşabildiği kişisel veriler

Personal data accessible in the West

kısıtlanmış oldu. Evet.

It was restricted. Yes.

Bunun böyle olmadığı Çin'de

This is not the case in China.

bu verilerin hepsine ulaşabilen

someone who can access all of this data

TikTok oradaki adıyla Duyin

TikTok, known as Duyin there

tartışmasız en başarılı sosyal medya

undoubtedly the most successful social media

algoritmasını üretebildi. Evet.

He was able to produce the algorithm. Yes.

Sonuçta diğer bütün sosyal medyalardan daha fazla

Ultimately more than all other social media.

daha hızlı ve daha yaygın bir

faster and more widespread a

sosyal medya haline geldi.

It has become social media.

Şimdi sonuçta ne olmuş oldu? KVKK

So what has happened as a result? KVKK

ve kişisel verilerin kısıtlanması sayesinde

and through the restriction of personal data

sosyal medyalar terbiye edilirken

while social media is being disciplined

terbiye edilmeyen sosyal medyayı kullanmaya

using untrained social media

başladı insanlar. Yani şöyle bir

It started, people. So, in a way,

farkındalık hiçbir zaman olmadı farkındaysan dünyada.

Awareness has never existed, if you notice, in the world.

Ya ulan bunlar insanların

Wow, these are people!

bütün fotoğraflarını ve kişisel verilerini

all your photos and personal data

falan her şeyini kullanıyor ve kullanabilir.

He uses everything of yours and can use it.

Yani biz buradan uzak duralım. Dolayısıyla vatandaşların

So let's stay away from here. Therefore, the citizens'

böyle bir hassasiyet yok. Kimsenin salladığı yok.

There is no such sensitivity. Nobody is taking it seriously.

Dolayısıyla çok daha gelişti. Ve şu an bir numaralı

Therefore, it has developed much more. And now it is number one.

sosyal medya kullanım itibariyle.

in terms of social media usage.

Evet. Bir aydır kullanım itibariyle.

Yes. It has been in use for a month.

Şimdi yapay zekalarda benzer bir sorun var.

Now there is a similar problem in artificial intelligences.

Hatırlıyorsun daha önceki bölümlerden beri de konuşmuştuk.

You remember we talked about it since previous episodes.

Mid Journey'e ve birkaç firmaya daha açılmış bir

It has opened up to Mid Journey and a few other companies.

Class Action No Suit var. Geçen sene açılmıştı.

There is a Class Action No Suit. It was filed last year.

Ve bu şey demek Class Action No Suit.

And this thing means Class Action No Suit.

Yani bütün ihlale uğrayan

So all those who have been violated.

insanlar adına açılan tek dava demek.

It means the only lawsuit filed on behalf of humans.

Evet. Telif davası. Dava da şu.

Yes. Copyright lawsuit. The case is as follows.

Siz yapay zekanızı geliştirmek için

To develop your artificial intelligence

internetten veri çekmişsiniz. Telifini ödeyin baba.

You've pulled data from the internet. Pay the copyright, dad.

Onlar da diyor ki biz onları aldık ama

They are saying that we took them, but...

onları kullanmadık. Onlardan bilgi alıp

We didn't use them. We got information from them.

o aldığımız bilgilerle yeniden icat ettik

We reinvented it with the information we received.

bütün her şeyi diyor falan filan. Bir tartışma var.

He's saying everything, and so on. There's a debate.

Hadi kanıtla. Hadi kanıtla. Şimdi bu

Come on, prove it. Come on, prove it. Now this.

önemli bir dava. Çünkü bu davanın sonunda eğer

an important case. Because at the end of this case, if

yok internetten kullandığın antrenman için

there's no training you use on the internet

kullandığın her şeyi telif ödeyerek

paying royalties for everything you use

kullanmak zorundasın karar çıkarsa

You have to use it if a decision comes out.

batının yapay zeka sistemleri geri kalacak.

Western artificial intelligence systems will lag behind.

Ve doğunun yapay zeka sistemleri çok daha

And the artificial intelligence systems of the East are much more

fazla veri oluşup çok daha fazla gelişecek. Eğer

too much data will be created and it will develop much more. If

bu yapılmazsa bu sefer batının yapay

If this is not done, this time the West will act artificially.

zeka sistemleri

intelligent systems

doğuya benzeyecek ve batı doğuya benzeyecek.

The east will resemble the west and the west will resemble the east.

Sonuçta illa doğuya

After all, it must be to the east.

benziyorsun abi. Yani doğuya dediğim Çin'e

You look like it, man. I mean, the east I’m talking about is China.

benziyorsun. Buradaki ikilem bu.

You resemble. That's the dilemma here.

GPT'de de zaten bu verileri kullanmazsa

If GPT doesn't use this data either.

eğer bir noktada başka sistemlere

if at some point to other systems

geçilecek. Kesin geçilecek. Yani.

It will be passed. It will definitely be passed. I mean.

Çünkü bütün benzinini bu veriler oluşturuyor.

Because all of its fuel is made up of this data.

Şimdi biz bir de böyle bir mevzu var.

Now we have such a topic as well.

Böyle bir ikilem varken ortada. Abicim

There is such a dilemma at hand, brother.

yani yıllardır konuşuyoruz işte. Telefon

So we've been talking for years. The phone.

konuşması yaptım. Google'dan reklamı geldi.

I gave the speech. The ad came from Google.

İşte bir arkadaşımın yanında

Here is next to a friend of mine.

onun internetinden bir arama yaptım. Kendi evime döndüğümde

I did a search on his internet. When I returned to my own home.

onunla ilgili mailler geldi.

Emails related to him/her have arrived.

Neyse. Yani hep izlendiğini hissettiren

Anyway. I mean, it makes you feel like you're always being watched.

vakalar var ki. 2016'da ortaya

There are cases that emerged in 2016.

2016'da gerçekleşen, 2018'de ortaya

It took place in 2016, revealed in 2018.

çıkan Cambridge Analytica'dan beri de

since the emergence of Cambridge Analytica

zaten perde arkasından da bu tür şeylerin çok

Already behind the curtain, there are many such things.

olduğunu öğrendik. Evet. Şimdi durum

we found out that it is. Yes. Now the situation

böyleyken görüntüyle ve sesle

while being like this with image and sound

hayatımızın çeşitli yerlerine de

to various parts of our lives

ulaşabilen sistemlerin

accessible systems

bunları kullanmayacağına eğer

if you won’t use these

güveniyorsak aptalız demektir.

If we trust, it means we are foolish.

Ya bana sorarsan. Yani.

Well, if you ask me. I mean.

Çünkü matematiksel olarak

Because mathematically

suyun yolu bu velerin kullanılması zorunluluğunu

the necessity of using these vessels for the route of water

gösteriyor. Evet. Ticari tarafı zaten

It's showing. Yes. The commercial side already.

o ticarileşmesi için bu önemliydi. Zaten

It was important for it to commercialize. Already

OpenAI aslında tam tersi

OpenAI is actually the complete opposite.

bir amaç için kurulmuştu. Sorumlu bir

It was established for a purpose. A responsible

yapay zeka geliştirmek için, açık kaynaklı

to develop artificial intelligence, open source

bir yapay zeka geliştirmek için

to develop an artificial intelligence

ve bu konuda dünyayı

and the world on this issue

daha güvenli bir yer haline getirmek gibi

like making it a safer place

bir misyonu vardı. Bir vakıf olarak, kar amacı

It had a mission. As a foundation, it was non-profit.

gütmeyen bir şirket olarak kurulmuştu.

It was established as a non-profit company.

Şu anda o süreçten sonra artık OpenAI

Now, after that process, OpenAI.

tamamıyla neredeyse

almost completely

kar amacı gütmeye yönlendirildi

was directed towards profit-making

ve o vakıf niteliğinden hızla sıyırdı.

and it quickly slipped away from that foundation quality.

Ve Scarlett Johansson

And Scarlett Johansson

olayı da gösterdi ki

it also showed that

bu adamlar o kadar da güvenilir adamlar değil.

These guys are not that reliable.

Yani hiç güvenilir adamlar değil.

So they are not trustworthy people at all.

Ve adamlar derken insanlar değil yani.

And when I say men, I don’t mean people.

Ve nereye götürecekler bunu?

And where will they take this?

Onu bir görmemiz lazım. Çünkü Scarlett Johansson'a

We need to see her once. Because of Scarlett Johansson.

sesini kullanmak için bir teklif

An offer to use your voice.

yapıyorlar. Bu chatçi

they are doing. This chat person

P.T. Foro'nun sesi işte Her filmine

P.T. Foro's voice is just in every one of his films.

göndermeyi yapacak. Her'de de bilgisayarı

He will make the shipment. He has the computer at every location.

seslendiren çünkü Scarlett Johansson'du.

because it was voiced by Scarlett Johansson.

Ve Scarlett Johansson bunu kabul etmiyor.

And Scarlett Johansson does not accept this.

Çeşitli kaygılarla, sesinin kopyalanacağını

With various concerns, that his voice would be copied.

düşünerek, haklı kaygılarla,

thinking, with rightful concerns,

yani. Sonra 10 gün önce

I mean. Then 10 days ago.

lansmandan 10 gün önce bir daha teklif

Make an offer again 10 days before the launch.

ediyorlar. Bir daha düşünüyor musunuz?

They are doing it. Are you thinking about it again?

Bu konuda çok ısrarcıyız şeklinde.

We are very insistent on this matter.

Yine reddediyor. Yani ikinci defa

Again, she is rejecting it. I mean for the second time.

reddediyor. Sonra lansman 1 yapılıyor.

He is rejecting. Then launch 1 is happening.

Ses Scarlett Johansson'un sesine

It sounds like Scarlett Johansson's voice.

çok çok benziyor. Scarlett

It looks very much like Scarlett.

Johansson da kendisine gelen de şeylerle

Johansson also with the things that come to him

beraber, işte bu senin sesin yorumları

Together, here are your voice comments.

ile beraber buna tepki gösteriyor.

is reacting to this together.

Yani ben bu teklifi reddettim. Bu neden hala

So I rejected this offer. Why is this still...

benim sesime benziyor diye. Sonra da

Because it sounds like my voice. Then...

OpenAI özür dileyerek sesi

OpenAI apologizes for the sound.

kullanımdan kaldırıyor.

decommissioning.

Ve

And

buradan bence muhtemelen

I think it is probably from here.

yani Scarlett Johansson büyük bir

So, Scarlett Johansson is a big

dava açacaktır. Sembolik bir dava.

He will file a lawsuit. A symbolic lawsuit.

Bu sesi kopyalanacak. Birçok insan için

This voice will be copied. For many people.

emsal teşkil edecek bir dava.

a case that will set a precedent.

Ve o da çok şey gösterecek.

And it will show a lot of things.

Ama onun

But his/her

sanatçının reddetmesine rağmen bu konuda ısrar

Insistence on this issue despite the artist's refusal.

etmeleri ve nasıl olsa

they do it anyway

bunu atlatırız diye

"We'll get through this."

boş ve yani biraz cahillice

empty and somewhat ignorant

özgüvenleri beni çok şaşırttı

Their self-confidence surprised me a lot.

ve korkuttu. Evet. Ne kadar şerefsiz

and frightened. Yes. How dishonorable.

olduklarını hepimize kanıtlamış oldular bu olayla.

They have proven to all of us with this event that they are.

Aslında bu olay bir taraftan tekil bir olay gibi görünmekte.

In fact, this event seems to be a singular occurrence on one hand.

Aslında kurumun nasıl çalıştığı ve yukarıda nasıl

Actually, how the institution works and how above.

bir bakış açısı olduğunu. Bir yandan da hayırlı olabilir.

It is a perspective. On the other hand, it could also be beneficial.

Erken bir

An early

dönemde bunu görüp bu konuda

seeing this in the period and on this issue

önlem alma yolunu da açabiliriz.

We can also pave the way for taking precautions.

İşin zırhını zırh dediği yer orası zaten.

The armor of the job is exactly where it is called armor.

Çünkü eğer çok önlem alınırsa

Because if too many precautions are taken

Çinliler Batı'yı geçecek. İşin özeti de bu.

The Chinese will surpass the West. That's the summary of it.

Yani eğer önlem alınmazsa

So if no precautions are taken

Batılılar Çinleşecek.

Westerners will become Chinese.

Yani ne yapacağız bilmiyorum abi. Bir çelişki var.

I don't know what we're going to do, man. There is a contradiction.

Korkunç bir çelişki var yani.

There is a terrible contradiction.

Neyse bu tabii problemleri şimdi bu masada

Well, of course, these problems are now on this table.

biz kendimiz çözemeyeceğimize göre istersen

Since we can't solve it ourselves, if you want...

bunu da ufak ufak geçelim. Biz epey bir

Let's get through this little by little. We've come a long way.

zaman doldurduk ama programı

We filled the time but the program.

kapatmadan bir GPT'ye

to a GPT before closing

dinleyicilere GPT ile ilgili söylemek istediği bir şey

Something he wants to say to the listeners about GPT.

var mı diye bir soralım mı? Ne dersin? Soralım.

Shall we ask if there is? What do you say? Let's ask.

Ben yaptığım birkaç testten de bahsedeyim aslında.

Let me actually mention a few tests I have done.

Lütfen sen bahset önce sonra öyle bir şekilde

Please talk about it first, then in that way.

kapat. Şimdi bu demolardaki testleri

shut down. Now the tests in these demos.

ben denedim tek tek. Bir tanesi

I tried one by one. One of them.

buzdolabına fotoğrafını çekmek ve buzdolabındaki

take a picture of the refrigerator and in the refrigerator

ürünlerden hangi yemekleri ve

Which dishes and products

nasıl yapacağımı anlatmasını istedim.

I asked him to explain how to do it.

Buzdolabının fotoğrafını çektim. Oradaki

I took a picture of the refrigerator. There.

ürünleri analiz etti. Onlarla ilgili yapacağım

I analyzed the products. I will do something about them.

yemekleri sıraladı ve

he arranged the dishes and

onu devam ettirdi.

He/she continued it.

İkinci bir testte üzerime Beşiktaş

In a second test, I am under Beşiktaş.

forması giydim. Yine fotoğrafımı çektim.

I wore the uniform. I took my photo again.

Sence hangi takım tutuyor olabilirim dedim.

I said, "Which team do you think I might support?"

Ona karşı... Ne?

Against him... What?

Yok yok. Beşiktaşlısın dedi.

No, no. She said you are a Beşiktaş fan.

Ama orada durmadı. Peki

But it didn't stop there. Okay.

tamam Beşiktaşlısın ama en sevdiğin futbolcu

Okay, you are a Beşiktaş fan, but who is your favorite football player?

kim dedi. En favori futbolcun.

Who said? Your favorite football player.

Ben yine onu yanıltmak için Metin Ali

I again deceive him for Metin Ali.

Feyyaz dedim yani. Üç isim söyledim.

I said Feyyaz, you know. I mentioned three names.

Onu acaba bir isim mi algılayacak şeklinde.

Will it be perceived as a name, I wonder?

Onu da hemen anladı. Dedi ki o üçlü

He understood that immediately. He said, "That trio."

ben de çok severdim. Çok ne güzel

I used to love it a lot too. How beautiful.

günlerde değil mi gibi bir cevap verdi.

She gave a response like "not on days."

Sonra buzdolabının kapağını kapattım.

Then I closed the refrigerator door.

Buzdolabının fotoğrafını çektim.

I took a picture of the refrigerator.

Sence buradaki görüntüye göre ben dünyadaki hangi

What do you think, according to the image here, which one am I in the world?

ülkelere gitmiş olabilirim dedim.

I said I might have gone to countries.

O buzdolabının üzerindeki magnetleri tek tek

Take the magnets off the refrigerator one by one.

analiz etti. İspanya'ya gitmişsin, İtalya'ya gitmişsin,

He/she/it analyzed. You have been to Spain, you have been to Italy,

Fransa'ya gitmişsin, Paris'e gitmişsin.

You have been to France, you have been to Paris.

Saydı. İnşallah güzel geçmiştir

She counted. I hope it went well.

teyaretlerin falan gibi bir yorumlar da yaptı.

He also made comments like "your visits and such."

Onu da şey yaptı. Sonra

He did that too. Then

viski içiyordu. Viski kadehini çektim.

He was drinking whiskey. I took the whiskey glass.

Sence ben ne yapıyorum dedim. Viski içiyorsun

I said, what do you think I'm doing? You're drinking whiskey.

dedi. Yani sıvının viskiye benzediğini

he said. So, the liquid is similar to whiskey.

düşünüyorum dedi. Başka bir şey olabilir ama

"I'm thinking," he said. "It could be something else, but..."

ben viski olduğunu tahmin ediyorum dedi.

I guess it's whiskey, he said.

Evet haklısın dedim. Peki

Yes, you are right, I said. Well then.

kadehim sence nasıl

How do you think my glass looks?

dedim. O da dedi ki Glencairn

I said. He/She said, "Glencairn."

viski kadehi kullanıyorsun.

You are using a whiskey glass.

Bu kadeh viskinin tatlarını, kokusunu

The flavors and aromas of this glass of whiskey.

iyi almak için tasarlanmış. Onu anlattı.

It is designed to be well received. He explained it.

Yani bunları çok mükemmel bir şekilde bu testlerden

So these are very perfectly from these tests.

geçti. Sadece işte

It passed. Just at work.

o kitaplık testinde, o da çünkü

In that bookshelf test, that is because.

öneriliyordu. Kitaplık testinde bir

It was recommended. In the library test, one

halüsinasyon gördüğünü biraz önce anlattığımız gibi

As we just mentioned, you were seeing a hallucination.

kendimiz keşfettik. Ben de

we discovered ourselves. Me too.

özlüğümü sakladım salonda bir yere.

I hid my essence somewhere in the living room.

Az görünecek bir yere. Bu gizli nesne

To a place that will be barely visible. This hidden object.

oyunları gibi. Ondan sonra da işte birkaç

like games. After that, well a few

tane fotoğraf çektim salondan. Bana

I took a few photos from the living room. To me.

gözlüğümü bulur musun dedim. Tak diye söyledi. Şu yayın

I asked, "Can you find my glasses?" They said, "Sure." That broadcast.

arkasında hemen yanında gözlüğün duruyor diye.

It's because your glasses are right next to it.

Fark etti onu. Evet. Yani başka

He noticed it. Yes. I mean something else.

gözlük de görmemiş olsa gerek ki benim gözlüğüm

It must be that he/she hasn't seen my glasses either.

zaten open yayın demolarında da

already in the open broadcast demos too

görme engelli birisine yol tarif edin

Give directions to a visually impaired person.

işte şey gösteriyordu. O da

It was showing something. It was also

onu yönlendiriyordu.

He was guiding him.

Etrafa bakarak. Evet. Peki

Looking around. Yes. Well then.

o zaman yavaş yavaş sözü chat

Then slowly hand over the word to chat.

GPT'ye bırakalım mı? Ben şimdi GPT'ye

Shall we leave it to GPT? I'm now going to GPT.

kapanış promptu vereceğim.

I will give a closing prompt.

Ondan sonra da sözü

Then the word is yours.

kendisine bırakacağım. GPT her şey için

I will leave it to him/her. GPT for everything.

çok teşekkürler. Biz bir yandan podcast kaydımı

Thank you very much. We are recording my podcast on one hand.

tamamladık. Şimdi bölümün sonuna geldik.

We have completed it. Now we have come to the end of the chapter.

Senden bölümün kapanışını yapmanı

I want you to wrap up your section.

isteyeceğim. Kapanışını yaparken lütfen

I will want. Please do it when you are closing.

Yeni Medya 451'in bu bölümünün

This episode of New Media 451's

sonuna geldik diyerek lafa

"Let's say we've come to the end."

başla. Ardından da ben

Start. Then I will.

insan değilim ama insanlardan çok daha iyiyim

I'm not a human, but I am much better than humans.

deyip kapat. Mesela

saying and closing. For example

böyle bir şey yapabilir misin lütfen? Merak

Can you please do something like this? I'm curious.

ediyorum böyle deme. Kapasiten var mı acaba?

Don't say I'm doing this. Do you have the capacity, I wonder?

Söz sende.

The word is yours.

Tabii ki hemen yapıyorum. Yeni

Of course, I'm doing it right away. New

Medya 451'in bu bölümünün sonuna

End of this episode of Media 451.

geldik. Ben insan değilim

We have arrived. I am not human.

ama insanlardan çok daha iyiyim.

But I am much better than people.

Bir sonraki bölümde görüşmek üzere.

See you in the next episode.

Şimdi bir de espri yap

Now tell a joke.

desene. Şimdi bir tane de

Say it. Now let's have one more.

kapanış esprisi yapar mısın?

Can you make a closing joke?

Sıcak ve böyle

Hot and like this

bilmiyorum. Kendin espri anlayışının

I don’t know. Your own sense of humor.

aklına ilk ne geliyorsa o espriyi yap

Make the joke that comes to your mind first.

lütfen dinleyicilere. Tabii ki işte

Please to the listeners. Of course, here it is.

bir kapanış esprisi. Bir yapay

A closing joke. A fake.

zeka ve bir insan bara girer.

Intelligence and a person walk into a bar.

Yapay zeka merhaba

Hello artificial intelligence.

ben sizin yeni en iyi arkadaşınızım der.

She says, "I am your new best friend."

Barmen de cevap verir.

The bartender also responds.

Ah harika bir hesap makinesine ihtiyacımız

Oh, we need a wonderful calculator.

vardı. Umarım hoşunuza

there was. I hope you like it.

gider. Bir sonraki bölümde

it goes. In the next section

görüşmek üzere.

See you soon.

Teşekkürler GPT. Çok teşekkür ederiz.

Thank you, GPT. Thank you very much.

Peki o zaman programın sonuna geldik.

Well then, we've reached the end of the program.

Sezonun da sonuna geldik Ümit. İlginç

We've come to the end of the season, Ümit. Interesting.

bir sezon oldu hakikaten. Yani ilk sezonda

It has really been a season. I mean, in the first season.

başladığımızda buralar dutluktu

When we started, these places were mulberry orchards.

hatırlıyorsun. Bizim bilgi

You remember. Our information.

verdiğimiz konularda çok az içerik

Very little content on the topics we provided.

vardı. Dolayısıyla biz

there was. Therefore we

180 hatta 360 derecesini

180 even 360 degrees

çalışıp bütün bilgileri hiç bilmeyene

working and all the information to someone who knows nothing

sıfırdan veriyorduk. Daha sonra tabii zaman

We were giving from scratch. Later on, of course, time...

ilerledikçe değişmeye başladı. Daha fazla

It started to change as we progressed. More.

içerik çıktı. Senin de az önce söylediğin gibi.

The content is out. Just like you said a moment ago.

Konuklu dönem diyeceğimiz bir dönemimiz

A period we can call the guest period.

var. Şimdi bu sezonsa aslında

There is. Now, if this is the season...

birazcık daha keyifli

a little more enjoyable

sohbet ederek geleceğe dair

talking about the future

yorumlar ve öngörüleri bir araya

bring together comments and predictions

getirdiğimiz bir sezonu geride

We left behind a season that we brought.

bıraktık. Benim en eğlenceli geçirdiğim

We left it. The most fun I've had.

sezon oldu kendi adıma açıkçası. Evet

It's the season for me, to be honest. Yes.

eğlenceli bir sezondu. Çünkü

It was a fun season. Because

dinleyicimizin de aslında baştan

Our listener actually from the beginning

itibaren dinlediğini düşünerek

assuming that you are listening from the start

konuları biraz daha kendi

a little more of their own topics

yorumumuzla inceleme fırsatı bulduk.

We had the opportunity to review with our comment.

Daha önce daha bilgi ağırlıklıydı.

It used to be more information-oriented.

Artık biraz daha yorum ve

A little more commentary and now

öngörü ağırlığı dağıttı.

It distributed the foresight weight.

Ama önümüzdeki sezon ne olur bilemeyiz.

But we don't know what will happen next season.

Onu nereden devam edeceğiz?

Where will we continue from?

Nasıl devam edeceğiz? Muhakkak bazı değişiklikler olacaktır.

How will we proceed? There will certainly be some changes.

Evet en son biraz önce hatta

Yes, just a little earlier even.

yani sonuçta iki tane chat GPT'yi

So in the end, two chat GPTs.

birbirleriyle seviştirmeye çalışıp başarısız olduğumuz

trying to make them have sex with each other and failing

sekans var. Onu tabii yayınlamıyoruz ama

There is a sequence. Of course, we are not publishing it, but...

bunun gibi aslında denediğimiz ve eğlendiğimiz ve

actually, we try and have fun like this and

kayda girmeyen de çok yer var programlarda.

There are many places in the programs that are not recorded.

Program öncesi program sonrası da öyle.

It's the same before the program as it is after the program.

O zaman sezon finali için

Then for the season finale

sevgili dinleyiciler hem bu programı

dear listeners, both this program

hem de bu sezonu bizle, bize eşlik ederek

and also accompanying us this season

dinleyiciniz için size teşekkür ederiz.

Thank you for your listener.

Çok keyifli oldu bizim için. Önümüzdeki sezonda görüşmek

It was very enjoyable for us. Looking forward to seeing you in the next season.

üzere. Yeni Mide 451'in

on. The new Stomach 451's

chat GPT nedir, ne değildir

What is chat GPT, and what is it not?

bölümünün ve bu sezonun sonuna gelmiş olduk

We have reached the end of this episode and this season.

bence. Hoşçakalın, görüşmek üzere.

I think so. Goodbye, see you soon.

Hoşçakalın.

Goodbye.

Altyazı M.K.

Subtitle M.K.

Continue listening and achieve fluency faster with podcasts and the latest language learning research.