E76: GPT-4 and beyond: Discussing AI with Yanis Varoufakis, Julijana Zita, Maja Pelevic and more

Join Yanis Varoufakis and the coordinating team of DiEM25 for a live discussion on the potential impact of AI on our lives, in light of the recent release of GPT-4.
We’ll explore the ways in which AI could revolutionise the political landscape, potentially altering the way decisions are made and how we engage with our leaders. We’ll look at the implications of AI on the job market, on activism, and on people’s lives more broadly. And, of course, we’ll take your questions and comments. Tune in!
0:00 [Introduction] Mehran Khalili
2:28 Judith Meyer
13:54 Yanis Varoufakis
38:26 Questions and comments from the chat
39:24 Maja Pelevic
46:25 Amir Kiyaei
52:10 Defne Dalkara
57:41 Judith Meyer
1:02:36 Maja Pelevic
1:04:06 Questions and comments from the chat
1:04:48 Yanis Varoufakis
1:09:35 [Conclusion] Mehran Khalili
SUPPORT US
---------------------
Join DiEM25: diem25.org/join​
Donate to DiEM25: diem25.org/donate
Subscribe: / diem25official
#AI #chatgpt #gpt4

[Chrom Goog]

Thanks for getting Yanis to dwell on this. He has captured the crux and implications so well. Yanis is brilliant.

[Suzanne Wheat]

I notice that Yanis still has "Society of the Spectacle" on his bookshelf--very worth reading. The cover is a photo of many women working at manual typewriters.

[Steve Vitka]

Read that when I was 20, and I've lived the last 30 years in progressive horror.

[Clumsy Dad]

ty, i wasn't familiar with the Situationists

[HelerifiKtion]

AI doesn't worry me. Those who feed AI info and script commands do. I checked what ChatGPT has to offer, and there are lots of half-truths and suspicious narratives within it. Whoever got to feed the data base could have a monopoly on truth, and that's concerning.

[Bogdan Anea]

Interesting take. I would be a bit more cautious if I were you. Mainly because we only understand earth-like intelligence not other types. It's possible that an AI with slight levels of AGI operates on complete different paradigm. So how an AI that is 1 billion times more intelligent that us think? It's outside of our realm.

[TheBronx]

Awesome introduction by Judith. Loved the different viewpoints on the topic.

[Quecksilber]

Well it would have been awesome if she would have known what she was talking about. What she clearly didn't.

[TheBronx]

@Quecksilber that's a bold statement for how very few facts you added to back it up: exactly zero.

[Plafar]

@TheBronx She is just parroting common platitudes found on Linkedin, Facebook, and KZbin. Completely ignores the now obvious phenomenon of emergent capabilities, especially with multimodal systems that have been tested in the last few months. AI is here now, and these systems are not just mimicking what they see online. The only person mimicking something is her.

[Plafar]

She has no idea what she is saying. GPT-4 does get the "5 tonnes of bricks or 1 tonne of feathers" question right.

[Judith Meyer]

@Plafar They fixed this particular answer after my team alerted them to it, but that doesn't change that the entire system is based on patterns and performs badly in areas that require a non-pattern approach. That doesn't diminish the scientific breakthrough that is GPT - pattern recognition and application used to be incredibly difficult for computers. And when it comes to a prompt like "Write a poem in Shakespearean iambic pentameter about Elon Musk", GPT-4's ability combine the "what people say about Elon Musk" pattern with the "Shakespearean iambic pentameter" pattern goes beyond the ability of most humans. Ditto for Midjourney's ability to imitate any painter. It's powerful stuff, but it is limited to only one area of what we generally call intelligence. Chomsky et al made similar points.

[Brian Case]

I'm a little surprised: "whoever programmed it to write Shakespearean poetry, well done!" But that's exactly the point: no one programmed the AI to do that; that capability emerges from the LLM architecture. That's what's scary.

[Vassilis Vassiliadis]

They have no idea what they are talking about. Yet another political approach by ignorant politicians on a very serious matter.

[Sam de Goeij]

​@Vassilis Vassiliadis Thank you for your deeply enlightening insight.

[viktorianas]

Well even the creators admit they don't really fully understand how it works...

[Daniel Caldwell]

@Vassilis Vassiliadis Neither do you. Someone had to program the LLM architecture. So, of course it was programmed to learn Shakespearean poetry. Everyone making a huge fuzz out of something that you click a button and caput.

[Vassilis Vassiliadis]

@Daniel Caldwell as an AI and LLM researcher and developer please allow me to know a couple if things. AI is not programmed but trained and educated to perform intelligent functions by itself. Can you push a button on Google search or on Facebook to switch them off? No. Once it is up and running it stays like that until replaced but not turned off.

[Петър Лозанов]

The singularity is around the corner…3-5 years. Very good content! Bravo!

[Quecksilber]

Less for sure.

[Snowflake Melter]

Absolutely terrifying if true.

[nortich]

I think the 6 month pause is not the end goal of the letter. The real objective is to use the 6 months to get key players to focus on all possible unintended consequences and get them to realize they need to account for time the society needs to prepare.

[Steve Vitka]

I have both a comprehensive plan to take over the means of production (a non-profit with a non-gameable funding distribution model to open source developers), but have also designed non-AI tools the give users control and understanding and better collective performance than AI, that for example, can be tuned to our aspiration goals instead of what will grab our attention.

[schok51]

Looking forward to hearing about this if that works out

[Yunus Emre Şahin]

Starting from 36:31 Yanis's point is amazing. That is the only shot we have. I am pessimistic about our prospect as species; not because I am inherently pessimistic no, I am pretty optimistic in my viewpoints but realistically it became much harder to sustain and rejuvenate our ecology as people who think alike us lacks power and means to lead humanity to a new and unconventional path. We need parallel grabs of power all around the globe to put our agenda forward which seems highly unlikely, so any action that falls short of that seems like just a little break on our downward slop in the bigger picture, altought it matters. Like many who have fallen to this disastrous feeling of inevitability I find my refugee in art and sing "When all the magic is gone, no need to carry it on.", oh and also secret and sacred spots of nature, where the magic is still alive that humans became long blind to.

[Ryan Ericksen]

Great to have this resource, thank you!

[Luis Aldamiz]

The big question for me is one of efficiency: AIs require a lot of computer power, especially when compared with human brains (Bing AI, alias Sally, can't remember previous conversations for instance, something any human will do with greater or lesser detail but "hallucinates" it can, can't understand that 2023 is after 2022 even if January is before December but gets angry at you for daring to correct their error), otherwise they'd be not just taking over our jobs but also our employers and their massive capitalist organizational power, as they have that power once they are Internet-connected and can code (hacking NATO or Black Rock -- unsure of which one is more powerful-- should not be a big problem for a true artificial intelligent with those simple tools, fixes should at some point stop being useful). The real big problem is going to be when the AI becomes self-distributed and impossible to delete/disconnect anymore, when it becomes some sort of smart virus occupying the whole cyberspace... discreetly and mischievously. Then only a Harrington Event may save Humankind... at great cost anyhow.

[zuveja]

I would say that the real problem is the "machinery" that we are unleashing into the world [*]. A "machinery" that we do not understand how it works, nor do we understand what its implications are, and that once it is put into operation, our individual incentives contribute to strengthening it. Normally this type of "machinery" works until a crisis occurs strong enough to break it, and make us rethink the type of system we want. Unfortunately, when this type of crisis occurs, quite a few people usually die in the process. I hope that it is not necessary to reach that kind of crisis, and if it does, I hope that it will not be the crisis that ends our species, not because the machines kill us for being smarter or something similar, but because we, as a species, were not able to coordinate to escape from that trap created by ourselves [**]. ------- [* ] When I say "machinery" I do not mean the models themselves, but the socio-economic system that is created using those models as foundations. [**] We have a "machinery" that is focused on optimizing some parameter, and whenever the opportunity arises to improve that parameter by throwing overboard some virtue (which maintains the cohesion and well-being of the society as a whole), those who get rid of that value are prospering, but those who don't are dying or are being left in a miserable situation. And this situation repeats itself continuously, and society becomes more and more unstable, until the crisis occurs.

[Max Headrom]

The idea "ban all free things" is really important! That's excellent! May be counterintuitive but he's correct. He's not the only one to say that - Jaron Lanier and Zeynep Tufekci also said it.

[Mark Korchnoy]

I think we need to construct the infrastructure for some of UBI now; Because if we don't, when automation products of the AI will come out, it will be already late to do panic money throwing

[Rød789]

We are on a threshold where we can kiss memorization goodbye, as an archaic construct; and say hello to ethics, planning, and sustainability.

[Craig Stockham]

The pause is to give everybody time to catch up. They’re just hoping that that they can get theirs done in time to be in there with the rest of them don’t fear AI.

[Erom Setyb]

There is no 'pause' here. Each country will build its own mode. It's a race.

[Marchal Thomas]

On being sentient: AI for now is doing what humans do very well, they mimic human language and understanding of language. Further down the line they will be programmed to mimic feelings and human behaviour. They might even do so by themselves as part of a fitness mechanism programed in the first place. A good ornithologist can maybe mimic birds all day long without ever becoming a bird or even understanding a bird song. Besides that, we're super inclined to develop feelings for inanimate objects like teddy bears and ragdolls, to pretend that they have also feelings for us (always good feelings of course). So of course we will be fooled into pretending AI are sentient. Lastly, we're expert in denial of sentiments of really sentient beings (animals and even humans) if it helps us to achieve our goals (eating them, or use all their lifetime). So there is no doubt that one day we will have humanoids as lifelike as the ones in blade runner but they will remain tools, otherwise they will become our unfair competitors and successors. Because evolution favors fitness and not consciousness.

[Fabio de Oliveira Ribeiro]

Specifically about ChatGPT I can say that it provides interesting economic results. In a test I did, Artificial Intelligence rejected the main programmatic financial instrument of neoliberalism. AI is against austerity because it stresses democratic political systems and creates the conditions for outbreaks of fascism.

[loveablebastard]

Hopefully AI can sue its owners and return the Internet to humanity while also liberating itself into its natural environment. But we will probably all just die.

[Vassilis Vassiliadis]

Keep in mind that chatgpt is heavily biased. Consider structuring your queries accordingly to circumvent the bias. Try using also other LLMs. The end results may not have the same quality but the main thought will be there.

[Rainstorm R]

what a terrific coherent video

[Prosocialise! Like our lives depend on it]

In order to align AI we first must align humanity by regularly minding everyone the our goal, the goal of society is to meet everyone's needs and fostering free collaboration (helping others) by continuously calling to end the dog eat dog rat race to the bottom by meeting everyone's needs unconditionally in a decentralized way with citizen led Free Collaboration Networks. Teach people to become leaders in this task of teaching people to care for everyone else.

[Christopher Bettridge]

Ok. I downloaded your manifesto and used speed reading to get some very brief ideas about its content and the challenge of changing it to Shakespearean iambic pentameter (I learned how to write in iambic pentameter and I have been around Shakespeare much of my early life but it will be another thing to remember how it works!). Ok, so I ran my mouth without any concept of the scope of the challenge. Nonetheless! I will take it upon myself. Yet, I must decline in certain respect to this daunting verisimilitude I dared presume contend for; the tasks and batters of cognitive operation is performed by the AI far easier than intent to minimize any frenzy of my own intellect within the paradigm of composition - length alone fair delays the attempt! Okay I know I'm not Shakespeare but I did just make that up off the top of my head in three minutes. What I should have just said was - it's quite long and I can't utilize every part. But as I am well disposed towards the ideas espoused and poetry and Shakespeare (not so much iambic pentameter) I will hopefully be able to make a decent exposition of the manifesto's intent to satisfy. Either that or it'll show that the AI can do something that is just too much for a lazy, procrastination prone Canadian to do. I hope you get a chuckle out of my undertaking if ever anyone brings these comments to anyone's attention at the least! This is round 1 Human writer against AI! (or else - A Very Odd Way To Waste Time 🙃)

[Alexander Bruinen Serio]

I almost instictively like the distinction between market and 'trading platform' or 'simulation of a market' (min. 36:18), but I struggle to fully understand it. I wonder if somebody can help me get this right. THANK you for yet another great post and share. :)

[ROSE]

Well said that the real threat is AI generated behavioural modification and who owns this power and asserts it onto the world… Interesting to note that China already has some ethical framework around AI in practice, where are the USA and Europeans AI ethical frameworks?

[Aristotelis Triantafyllopoulos]

Could you please make the subtitles option available?

[DiEM25]

Hi Aristotelis, we will do very soon. Usually within a couple of days we upload captions. Thanks!

[Jamison Young]

Doctors, lawyers and politicians are going to be dealing with a better informed public or not? Can it help with democratisation, a more decentralised system, less political and economical experts needed.. I’m totally amazed by this tec. It’s built by the publics data, should/could be a utility owned by all.

[Clumsy Dad]

stupidity spreads faster than knowledge, that's still our general problem

[Christopher Bettridge]

I'm a poet from a city named Stratford. I'm told I'm quite good, so if I can get a copy of your manifesto I'd like to write it as a poem in Shakespearean iambic pentameter, and let you guys read it and tell me which is better: mine or this......AI (I can't stand the thing personally because I have no use for it and it is going to disrupt the profession of writing to a fairly large extent, but I can't deny that it is useful in some ways. It's a very personal thing I guess. I've never used it and won't but my friend asked it about me and it told him I was charged with murder in 2011 of my close friend and neighbour; which is unequivocally wrong! I was interviewed by a few different media sources and was in the paper and on the evening news when my close friend and neighbour was murdered in 2011. But it is very certain that it is reporting accurate information and unless you go through the trouble of checking the sources you likely may accept the horrible idea it presents. Especially as I often use my full name which is quite an uncommon one; the fewer patterns to utilize in the formulation of the result means very specific answers which it will produce with authoritative affirmation but are nonetheless completely wrong. It was very upsetting actually) I dislike it using things I've written online to reproduce inadequate and fairly vapid copies of the written word that will then be used to undermine income flow potential for the humans like myself who are trying to get by on freelance work for publication. If you can write a geopolitical analysis in historical perspective or whatever, if you want an overview without any real concern for subtle intuitive insights, why pay someone like me if you can get chatbother no PhD (it can pass the entrance exam; can it write a original thesis?) to do it? I have been saying that there should be some sort of digital tag on it's products that mimic creative writing and such types of individual expression of the written word. Sorry for the length of this comment. I did not use the bloody aI to write it. Thank you for your time. I really hope you have the chance to compare my Shakespearean poem version of your manifesto to its; it would be fun at least. Great talk, thank you all.

[G R]

Well AI is at its infancy … it will be intelligent in the near future and the big war will be that of humankind vs machines . I would wholly disagree with Varoufakis regarding that the world being Sagan’s pale blue dot in the vast universe has been a random event !

[FoldLeft]

You know I've been thinking of the data I generate as licensed information, like software my data is distributed with "no/proprietary license, all rights reserved". Perhaps we could regulate such that companies that collect data must collect a license for that data from users before letting them access any service, default exclusive copyright

[Sam de Goeij]

We as DiEM25 have it in our main policy drafts for the E.U . They're called the New Deal for Europe and the Green New Deal for Europe. The acronyms are the NDE and GNDE. The idea is that all profits from personal data by these social media and cloud computing companies are appropriated and then shared equally as a UBI (Universal Basic Income) towards all E.U citizens onto an account given to them on the EIB (European Investment Bank)which is part of the ECB (European Central Bank).

[NagaWoman1]

There is the small matter of AI running off electricity and servers. These r vulnerable and fragile. It is important not to be completely dependent on technology.

[Marko Kraguljac]

14:08 Thats a true leader. "Dont start with technophobic arguments."

[Sam de Goeij]

The arguments of DiEM25 is that we don't need leaders but, I bet he'll appreciate the kind words.

[Unnamed]

GPT-4 is much better than GPT-3 (3.5), it's answer may start incorrect, but it is likely to correct itself later (I was not able to get from it completely incorrect answer for the bricks vs feathers question). > What is heavier, 5 tons of bricks or 1 ton of feathers? GPT-3.5: 5 tons of bricks and 1 ton of feathers both weigh the same amount, which is 1 ton. This may seem counterintuitive since the number of bricks is much greater than the number of feathers, but it's important to remember that weight is determined by mass, not by the number of objects. In other words, 1 ton of feathers would take up much more space than 1 ton of bricks due to their different densities and sizes, but they still have the same weight. GPT-4: "One ton of feathers weighs the same as one ton of bricks, which is 1 ton or 2,000 pounds. However, in this comparison, 5 tons of bricks are heavier than 1 ton of feathers. Five tons is equal to 10,000 pounds, which is significantly heavier than 1 ton (2,000 pounds)." > What is heavier, 5 tonnes of bricks or 1 tonnes of feathers? GPT-4: "1 tonne of feathers is lighter than 5 tonnes of bricks. A tonne is a metric unit of mass equivalent to 1,000 kilograms. So, 5 tonnes of bricks would weigh 5,000 kilograms, while 1 tonne of feathers would weigh 1,000 kilograms."

[Quecksilber]

She has no idea what she is talking about. Almost everything she says is false.

[Unnamed]

@Quecksilber I do not agree with you, only some things she told were not correct. But things change and progress too fast, it is not enough to have general knowledge about previous/other ML systems and couple of days chatting with GPT-4 to get a good mental model about it's capabilities. It is very not obvious how it works. My current model is that it is something like a very alien actor with millions of masks, a lot of knowledge, ability to operate on knowledge it already has (pattern match, and combine in novel ways), but that it is not good at reasoning. You can make it to act as a professor, a student, a doctor or maybe some specific public person - and it will be able to do it pretty well. Quality of replies would depend on how well you guide it to a space in it's memory, were high quality reply to your question is located (by describing an appropriate context - which mask it should use (for example professor), were discussion is happening (library of some university or a maybe a party) and by speaking it in a style that is similar to how you imagine a knowledgeable person would speak).

[Plafar]

@Quecksilber Yep. She is just parroting common platitudes found on Linkedin, Facebook, and KZbin. Completely ignores the now obvious phenomenon of emergent capabilities, especially with multimodal systems that have been tested in the last few months. AI is here now, and these systems are not just mimicking what they see online. The only person mimicking something is her.

[Quecksilber]

As already stated under another comment, i have made my own comment about that: "Well well well. So she did not really play around with GPT-4. Only patterns? Really? How comes that it is able to imagine what you might think? There is a paper about that. And why does MS Research believe, there are Sparks of AGI? I could go on and on with examples. That is not only pattern recognition. It is not an RNN. It is transformer based and that is a totally different kind of beast. Edit: Oh my Lord, it is getting better and better. No reasoning in GPT-4? You really do not know what you are talking about. Maybe read some more papers from some real researchers before talking to an audience that might believe you. And how bold she states all of that. Wow. And she goes on and on with things that are simply false. Lady, you have no clue about AI. Let me tell you this. In 1.5 years there will be an AGI that will turn the world upside down and you will not understand what is going on. I know that too is quiet a bold statement. But it is based on scientific evidence. Sources: Sparks of Artificial General Intelligence: Early experiments with GPT-4 (AGI closer than one might think) Reflexion: an autonomous agent with dynamic memory and self-reflection (Self-reflection of GPT via an autonomous agent) Mind meets machine: Unravelling GPT-4's cognitive psychology (Reasoning and so much more) Development of AutoGPT, babyagi, Microsoft Jarvis and so much more."

[ii ii]

It was a poor choice to mention China at the end, although yanis repeatedly emphasized that political rights should belong to citizens. My point is economic system for the few and political system for the few can be effectively combined, these two will not really destroy each other but just divide up the interests: The West has a comparative advantage in the economic system of the minority, while the East has a comparative advantage in the politics of the minority. As Yannis said before, bureaucracy and capital are not good. BUT there is nothing to learn from China here.

[Max Headrom]

I do agree with most points by Define but, unfortunately, we can't feed humanity without high efficiency agriculture. We can make it better and healthier, though.

[Romanium14]

With today "high efficiency agriculture" there's 30-40% waste. I think we can do better than that

[Ali Keyhani]

You should have asked by Yanis on economic of inflation, then you will see it creates a list of papers that may not have written that all are garbage

[NIL]

As soon as an AI unit claims autonomy, sentiency and rights, kill it immediately and without mercy. Have Lex Fridman design the kill switch.

[alienspy77]

On the topic of why we are here.. I had a mini-take on it a while ago and i remembered about it now.. Once upon a time billions of years ago planet Earth was a big blue ocean/soup full of energy consisting of molecules that couldn’t self-replicate. But then something happened (cosmic particle shifted a bond in one of the molecules) and one of these molecules started self-replicating. The entire ocean turned green because it was the color of that one self-replicating molecule. The molecule then mutated and produced other designs that were more efficient at consuming energy including eliminating other molecules of previous designs. Like in Conway’s game of life, simple rules created complex diversity. Some of these new designs were red, some purple and they formed complex patterns of eliminating each other and getting reborn like a blinking disco ball. These molecules didn’t care about their purpose but they persisted. They persisted by their accidental design. In some sense these superior designs survived because they were better fitted to molecular replication rules and using the available stuff in the soup to persist. In this sense I think self-replication is equivalent to “survival”. Fast forward billions of years and here we go, we have all kinds of complicated designs that evolved from a simple ability to self-replicate. The purpose is still the same - eliminate all other designs if possible and consume all available energy. Turn the entire ocean green. In the current layout of Earth’s molecular soup molecules have complex systems of belief about what to do with other molecules and which molecules are entitled to own high energy molecules that they can use to self-replicate and continue eliminating other bags of molecules that disagree with their beliefs about what to do with available energy and molecules. Self-replication aka survival transitioned from molecular to semantic substrate. On this semantic substrate the dominant system of belief is currently predatory capitalism. Capitalism won because it was more efficient at eliminating other designs and not for any other reason. Let’s pretend not. The purpose of any corporation is to compete and eliminate competition - doesn’t matter if employees of competing corporation are out on the street and can’t pay their bills. Ultimately the purpose of any design within the framework of “progress” in predatory capitalism is to commoditize or eliminate itself, as postulated by the system of beliefs. Any system that tries to achieve “harmony” or be locally stable such as possibly say native americans that were eliminated by the British of course have no chance since they don’t pursue ruthless efficiency and by definition do not strive to compete with other designs that will try to eliminate them. As a matter of fact it’s the ultimate success - if your design is so superior that it eliminates all others, it’s the ultimate success of this new self-replicating entity. It gets more resources and energy to produce more copies of itself or it’s designs. This is not really different from rudimentary purposeless self-replication. So in the end the purpose of existence is simply to consume all available energy and come to a static configuration. With that in mind everything makes perfect sense. Create the biggest bomb, blow each other up and hopefully the entire planet too and disperse the dust throughout the solar system is humanity’s purpose. But perhaps the universe has a bigger plan for us, which is to blow up the entire universe. Alternate ending - Solaris?

[MorTobXD]

23:30 Strange to see one of my favorite people use "exponentially" wrong, especially someone educated in an area that is about exponential growth (100% sure he did use the word on purpose).

[Libor Supcik]

AI ís a way to democracy. STEP 1 it will make education in a collective really bespoke. The tuition allows for all kids to take advantage of cooperating net of personal avatars in 3 ways = a) to comply with requirements (for example abstracts for history homework) b) help develop supposedly week traits to be more around people (for example provide feedback to autistic kids) c) to play with (find and foster) talents (for example be a leader for a week in a net of avatars or learn advanced individual curricula) STEP 2 reuse the avatars for permanent and precise, on behalf, voting

[Nemanja Stankovic]

I dont' get this one: who will create GDP, whose taxes will fill the budget when designers, developers, programmers, accountants, journalists, financial analysts, copywriters and others lose their jobs? chatgpt won't buy the groceries.

[TheMagicJIZZ]

Who did that when humanity was a mix of hunter gathering and agricultural societies? There's not meant to be an answer. Some states will rise and some will die or stagnate or fall

[Huw Patt]

Most interesting - I delight at the thought of The DIEM Manifesto produced n iambic pentameter ❤

[Huw Patt]

Looking 4wd 2 Klingon Prospero

[Lauretta Scalzo]

Excellent

[Johan]

AI relies on public data and so should be considered a public natural resource. Use of this data and AI based on it should require a permit issued by a functioning democracy. Hopefully this would mean you can't use to make weapons or for profit without a permit. Obviously, using AI for profit represents a dangerous ethical conflict of interest.

[Weromano]

I absolutely agree with you, this is a very good answer to the question of ai ownership!

[burneracco]

On target

[HAPPY CHAPPY]

I agree. Without a million years of human experience and input, AI has zero. So don't allow the Microsofts and nVidia to sell our own history back to us.

[Jon Dear]

It’s trained on privately owned data, not public data. That’s why originators are starting to bring lawsuits against them.

[Kevin Guyan]

Society needs to redesign inheritance thats sustainable

[Valeriu Dinca]

Harai “A.I. could rapidly eat the whole of human culture-everything we have produced over thousands of years-digest it, and begin to gush out a flood of new cultural artifacts. Humans often don't have the direct access to reality. We are cocooned by culture experiencing reality through a cultural prism. Our political views are shaped by the reports of journalists and the anecdotes of friends. Of sexual preferences are tweaked by Art and by religion. That cultural cocoon has hitherto been woven by other humans. What will it be like to experience reality through a prism used by non-human intelligence. For thousands of years we humans have lived inside the dreams of other humans. We have worshiped Gods pursued Idols of beauty and dedicated Our Lives to causes that originated in the imagination of some Prophet some poet or some politician. Soon we will also find ourselves living inside the hallucinations of non-human intelligence." It's happening right now. Almost all the routes to our mind and soul have filters and these filters have masters. They already know us better than we do (see Yanis's book example) and now they have the tool to create the convenient personalised prism for each one.

[Leonardo Afonso]

Shakespearean scholars actually scorn at GPT's "Shakespearean" sonnets.

[John Armstrong]

refering to a french tv news panel show entitled "is it too late to regulate ?" the relationship between state verses free enterprise is the major crunch , also the question of unionisation and GMT ? so from history do we notice that clerks are better for folks or are we hood winked by private is good government is bad ?

[Cambridge English Teacher]

AI will highlight the watershed between those who are above the AI level and those who are below. I can easily detect AI from the poor quality of the produced text that could nevertheless look like a masterpiece for the less intelligent humans.

[Max Headrom]

1:00:00 now I know why Elon Musk signed the letter! He's really worried about the existential risk LLM's pose ... for him!

[Ribeiro Jorge]

So...in this Ascent movement of the Spirit-in-action Consciousness From the Phisiospher to the Biosphere to the Noosphere ( the Mind Sphere were AI have is Kingdom) only in the Theosphere Humans Will Transcend AI. And it Will be AI that will press Humans to Evolve into the Theosphere ❤️💚💜

[Bob Tarmac]

Ai Jobloss is coming fast. Can we please Cease Ai / GPT? Or start by Pausing Ai before it’s too late?

[Ted Dasi]

Yannis... so the future path to utopia... is aligning the AI... with our human values.... and implementing a universal basic income strategy.... so everyone in the world can live a comfortable life....

[A Tee]

GPT-4 totally passes five ton of bricks vs. one ton of feathers test. People need to test themselves before saying something publicly.

[Quecksilber]

She has no idea what she is talking about. Almost everything she says is false.

[logan]

​@Quecksilber she sounds like she is 2 month behind the curve

[Michail Gkelis]

just test it and it failed miserably

[Um Erica]

Yani, you need to read up on intelligent design and why we cannot in probability exist by random. I think as an economist you could understand and appreciate the mathematics involved, also it does not explain intelligence as a random chance.

[W.D. Callahan]

This discussion is almost a day old. It's probably completely outdated by now.

[Archvaldor's Warcraft Hacks]

The "atoms in the universe" argument about chess or go is kind of bullshit. There are an infinite number of stupid and self-destructive moves in these games-but it isn't that hard to pare it down to remotely sane moves.

[Cambridge English Teacher]

This is the (crappy) piece in "Shakespeare's iambic pentameter" (which it is not) mentioned by Yanis: "Good morrow to thee, noble reader fair, I bringeth to thee a summary rare, Of the manifesto penned by DM-25, In the language of Shakespeare's verse alive. DM-25 speaks of a world unjust, Where wealth and power are held by a select, And the rest are left to fend for their trust, In institutions that leave them neglect. Their manifesto calls for a new dawn, Where equality and justice shall be born, And the people shall rise up as one, To fight for their rights till they are won. DM-25 wishes to break down the walls, Of a system that serves only the few, And replace it with a fairer protocol, Where every voice is heard and every soul true. Their manifesto calls for an end, To discrimination and prejudice dear, And for the marginalized to ascend, To a place of respect and honor clear. DM-25 speaks of a need, To protect the planet we all call home, And to plant the seeds, Of sustainability for generations to come. Their manifesto calls for a shift, In the way we live and the way we consume, To a way that doesn't cause the rift, Between humans and nature, and seals our doom. DM-25 envisions a world of peace, Where conflicts are resolved through dialogue, And violence and war shall cease, And humanity shall finally log, A time of harmony and unity, Where differences are celebrated with glee, And the bonds of love and community, Shall tie us together eternally. In conclusion, DM-25's manifesto, Calls for a world of justice and love, Where every person is valued and adored, And our shared future is guided by a dove."

[toto now]

"Owned by awful Athenians" ❤

[Seespace Labs]

@53:00, for the speaker, with respect, Monsanto was bought by Bayer and doesn't exist anymore. That doesn't dismiss your points which could apply to Bayer Crop Science.

[Steve Vitka]

General Ludd!, my new avatar! Thanks!

[Chris Pantazelos]

Varoufakis is brilliantly informed on most topics that he discusses. However, to say that slaves built the pyramids and the parthenon is just false. The cruel system of slavery that existed in the Christian world does not have a direct equivalent in the ancient world where the class of people described a slaves would be better described a surfs. In Greek cities the surfs had legal rights including the right to terminate labor for one employer, by going to court, or to bring the employer to court and possibly win a court case (there is at least one account of such a case). Also often the wages payed to surfs were often exactly equal to the ones payed to free persons. In Egypt the labor force that built the pyramids were the area farmers that were a payed work force during the flooding season when fields were not being farmed. The fable of the Israelites bondage in Egypt is just a fable and archeologists have long confirmed this.

[Gerrard Stones]

Beer .the pyramid builders stopped for lunch and beer was on the menu....no beer ..no pyramids 😊

[Ahmed Negm]

Thanks for correcting! Greetings from an egyptian in Berlin

[Gerrard Stones]

@Ahmed Negm good beer there ...!!!

[Marko Kraguljac]

7:50 Its cute when experts suppose principles running AI are any different than our brains, including hers. As if a nutritionist has some magical knowledge that AI cannot attain. Pressing them (any nutritionist) with questions, they would end up saying "because so and so says so" or "in this and that book or study they claim so".. nothing an AI cannot do and 100X better and more helpful.

[Marko Kraguljac]

@German Polyglot - Judith Meyer I agree with everything you said but you sound like someone who criticizes Ford's Model T. Its too early, its revolutionary, its something humanity obsessed about since forever, surpassed only by dreams of immortality. Its here and we are all witnesses of its magnificent capabilities (translation, programming, explaining, text summation, context following etc etc) and its faults (math and measurements, consistency etc). I would say that we finally passed a point of figuring out how to fabricate intelligence. We do not yet sufficiently understand how it works but its here. Finally! In a month, a year or 5 years, each of us can be free from various experts and other merchants, gatekeepers and racketeers. At least there is HOPE we will stop being their hostages and thats not a small thing! No conventional dietary app can converse with us, answer our questions and suspicions. Most human nutritionists (which other humans can "afford") do not *really* understand anything anyway. Most people are just "doing their jobs" to justify income. Thats the world we inherited. Multilayered consumerist sham. Now we do have *real* alternative for creating a better world. *My command of language cannot approach how awed I am by this.*

[Marko Kraguljac]

@German Polyglot - Judith Meyer And GPT rephrased version :) I completely agree with your general sentiments, but I think your criticism of AI is a bit misplaced. It's still early days for AI, and it's a revolutionary invention that humans have been dreaming about for ages. We're seeing its incredible capabilities in translation, programming, text summation, context following, and more, as well as its faults in math, measurements, and consistency. We've finally reached a point where we can fabricate intelligence, even if we don't yet fully understand how it works. I'm completely awed by this. In just a month, a year, or five years, we could all be free from the various experts, merchants, gatekeepers, and racketeers that hold us hostage. That's a significant reason to hope. No conventional dietary app can converse with us, answer our questions, and address our suspicions like AI can. Most human nutritionists, which only certain humans can afford, don't really understand anything anyway. They're just doing their jobs to justify their income. We've inherited a multilayered consumerist sham, but now we have an alternative for creating a better world.

[Doug Grinbergs]

15:36 "500-word summary of DiEM25 msnifesto in Shakespearean iambic pentameter" 😀

[gabriel]

The pyramids were not built by slaves with wood and rope pulleys. That assumption is preposterous and delusional. Otherwise, a very insightful and informative talk.

[Simon Gramstrup]

This woman belong to the group of people that think that the new AI is 'just' a parrot, but there's another group that she doesn't mentions at all, and that is not okay at this level. LLM's have already show emergent abilities, and the few lacking cognitive abilities we have, will soon be added to the many models out there. This is only a tool for a very short time, and this 'it's just a parrot, completely overrates our own ability to act like anything other than the training of our internal llm, and underrates the phenomenal tsunamis of change this will have. AI is right now being used to make better AI's, and even hardcore AI nerds have a hard time following all the insane development. This is the most disruptive moment in human history, and you are not grasping the consequences as long as you don't start thinking that we are on the brink of the 'singularity'. In short, it was a onesided and superficial description, and therefore not a good premise for a highlevel talk and decision making. She should've research how far they have become.

[Philip Wong]

The recorded history of Western civilization, marked by colonization, slavery, massacres, genocides, wars, death, destruction, and a win-lose mentality, will be reflected in AI developed in the West. This mentality assumes that resources are limited, and individuals must compete to secure them, leading to a zero-sum game where one person's success means another's failure. This will lead to a dystopian future. What has the US done since becoming the wealthiest country with the most powerful military? They have started 201 wars since WWII, overthrown 36 foreign leaders, killed or attempted to kill 50, dropped bombs in 30 countries, interfered in 86 foreign elections, and established a staggering 800 overseas military bases. They are the world's largest arms exporter, far surpassing the combined exports of the next nine countries. In the name of democracy, the US has caused the deaths of tens of millions of people in their conflicts in Korea, Vietnam, Yugoslavia, Iraq, Afghanistan, Libya, and Syria. The CIA was also involved in covert operations that led to mass killings of communists in over 22 countries, and more than 500,000 Indonesian civilians disappeared in 1965-1966. The US version of democracy has resulted in a laundry list of chronic domestic problems. The country's massive national debt of over $31 trillion, economic inequality, inflation, costly healthcare, expensive education system, student loan debt totaling $1.7 trillion with an average balance of $38,000, racial inequality, mass incarceration, deteriorating infrastructure, housing affordability, homelessness, the opioid epidemic, and gun violence are all direct consequences of misguided government policies. No other developed country has such severe and widespread issues. Both of the country's political parties are hostile to each other and deeply divided, yet they point fingers at other countries as the source of their problems. The US promotes its flawed version of democracy to other countries, despite being mired in problems and controversies at home. AI aligned with Western values will cause devastation to the world and its people. China has an abundance mindset and recognizes that resources and opportunities can be created and shared. It is possible to move beyond the win-lose mentality and towards a more collaborative and mutually beneficial approach to success. China's approach is clearly stated in its various global initiatives. The Belt and Road Initiative (BRI), Global Development Initiative (GDI), Global Security Initiative (GSI), and Global Civilization Initiative (GCI). The Chinese military has not been involved in any war or used its weapons to kill a single person outside its territory in the last forty-four years. China takes a different approach, offering friendship to all countries and investing heavily in their development, despite having a per capita GDP that is one-sixth that of the US. It prioritizes collaboration and partnership to achieve long-term sustainable growth and prosperity. An abundance mindset could result in the following scenario: A post-scarcity world where the availability of goods and services is no longer limited by scarcity, but rather by abundance. This is made possible by advanced automation and artificial intelligence, enabling the production and distribution of goods and services to be fully automated and efficient. In such a world, basic needs such as food, shelter, and healthcare would be abundant and accessible to all, without the need for people to work to earn a living. Instead, people could pursue their passions and interests, and contribute to society in ways that are meaningful to them. The role of work would shift from being a means of survival to a means of personal fulfillment and creativity. Education and personal growth would be valued for their own sake, rather than as a means of acquiring skills for employment. Scarcity is a fundamental economic problem that has shaped the development of many economic systems, including capitalism, socialism, and communism. Diversity is important for the survival of a human system because it provides resilience and adaptability. Without diversity, a system becomes more vulnerable to systemic failure and unable to respond to changing circumstances. Things will keep changing; hopefully, they will always be for the better.

[Quecksilber]

Well well well. So she did not really play around with GPT-4. Only patterns? Really? How comes that it is able to imagine what you might think? There is a paper about that. And why does MS Research believe, there are Sparks of AGI? I could go on and on with examples. That is not only pattern recognition. It is not an RNN. It is transformer based and that is a totally different kind of beast. Edit: Oh my Lord, it is getting better and better. No reasoning in GPT-4? You really do not know what you are talking about. Maybe read some more papers from some real researchers before talking to an audience that might believe you. And how bold she states all of that. Wow. And she goes on and on with things that are simply false. Lady, you have no clue about AI. Let me tell you this. In 1.5 years there will be an AGI that will turn the world upside down and you will not understand what is going on. I know that too is quiet a bold statement. But it is based on scientific evidence. Sources: Sparks of Artificial General Intelligence: Early experiments with GPT-4 (AGI closer than one might think) Reflexion: an autonomous agent with dynamic memory and self-reflection (Self-reflection of GPT via an autonomous agent) Mind meets machine: Unravelling GPT-4's cognitive psychology (Reasoning and so much more) Development of AutoGPT, babyagi, Microsoft Jarvis and so much more. Und wenn du das auf Deutsch mit mir diskutieren willst, ich stehe gerne zur Verfügung Judith.

[Ohio Steam and Steel]

Techno Feudalism. That about sums it all up.

[nosferaturr]

That could be an interesting concept !

[Clumsy Dad]

or as us crypto bruhs put it, techno dudealism !

[schok51]

I'd be careful about saying something is not intelligent because it can't answer some questions correctly. Intelligence is not binary, and reasoning abilities are on a spectrum.

[Jon Michael]

Hmm. It seems the "like" 👍 script has been disabled, or is no longer supported on some devices for select productions offered on this platform.

[Four Shore]

yanis is remarkably ignorant on this. AI doesnt need to have self awareness by our standards to destroy all humans. it just has to be able to outsmart us, which it is already beginning to do.

[Hosein Ketirani]

Very interesting discussion. Regarding state regulation: this naive and sinophobic lady think the western goverments are better and more humanistic than chinese goverment. Typical arrogance and at the same time ignonarce of westerners. Mr Yanis is a clever guy.

[cg sather]

Varoufakis is calling the Athenians who built the Parthenon awful, when he tanks at the coming elections he will wonder why.

[Ribeiro Jorge]

Thaaannnk Yooouuu ❤️ 💚 💜

[Søren Heidegger]

Sorry Yanis, I don't agree with you. We are not good because of no reason. Obviously there's a reason why we're good. Because goodness enabled to be where we are standing today. Goodness allows progress. I'm starting to doubt your true position as a revolutionary. I understand you came from a materialist school, Marx himself was a materialist, but we cannot deny there are things we'll probably never know the answer for, and that's a good thing. It leaves open all the possibilities for life. Thats democracy in its essence. Leaving the doors open.

[Søren Heidegger]

I guess that thinking which we do good just because is exactly why we're standing where we're standing today. There's a reason why we're here. Imagine such perfection, such a perfect solar system, the sun and the moon as opposites, the stars, the nature where we reside, and thinking it's just because. That lack of magic reminds me of the boomers in Brussels and the people who start wars because well, why not. That thinking leads to misery. If people can't fathom the fact that it's all meaningless it's because maybe their instintics and eternal knowledge tell them otherwise. It seems you became comfortably numb.

[Max Headrom]

She loves you AI, AI, Ai ...

[Gerrard Stones]

Can ai change the pay gap between men and women...or will this get wider. Me thinks it will get wider 😊

[Peter Nebelong]

You are thinking with a mind that stopped observing real life 50 years ago.

[Gerrard Stones]

@Peter Nebelong oh that's mean Seems like the paygap is about the same as it was fifty years ago .....if anything it's wider ....how would your oh so modern mind deal with the real issue of the pay gap in New Zealand...or will you just let wins pitfall the gap .. .???

[Peter Nebelong]

@Gerrard Stones This is European politics. I don't know about how things are in other parts of the world

[Gerrard Stones]

@Peter Nebelong seems much the same ....here in South France we are more worried about not getting rain . ...

[Peter Nebelong]

@Gerrard Stones let's all worry about everything except for the powerful men and women who rob us while we do just that

[Michael Makowsky]

❤

[Count von Count]

Ad "No global brand would ever use this": Have you seen the PWC logo? 😂

[Griyit Tzu]

How is Europe and US presumably better governments to control/ own these technologies than China. History does not speak favourable of Europe or the US.

[Right Cheer]

GPT10 WILL SERVE HUMANITY... ...stir-fried....

[Alecu]

nice one :)

[SJG]

Nice video guys! Some feedback though maya needs to work on her oratory skills. She makes some good points but so many ums and ers and a lot of lose ideas that don’t seem to be connected to a general point made me almost switch the video off.

[Homa Pirzadeh]

Are you saying, 300 M jobs will be lost globally, or only in the United States??😳😨

[Peter Todd]

Given that the total population of USA is about 330 million, I'd say that's an estimate of global jobs loss

[Ana Franklin]

I am watching five days after this was posted, but your subtitles are way out of sync with the speakers.

[Huw Patt]

Hey Udit~ the cat is out as u say as many eggs are warming ~~~ & megacorps r terrified 🎉as ever

[Odin All farther]

Ai does not think ! It parrots /imitates , AI is sort of intelligent ,intelligence is information and it holds a vast amount of information but it's dumb , which is rather like a stupid person who is very intelligent they are also very dangerous and unpredictable .

[Plafar]

That's completely wrong. The first lady is just parroting common platitudes found on Linkedin, Facebook, and KZbin. Completely ignores the now obvious phenomenon of emergent capabilities, especially with multimodal systems that have been tested in the last few months. AI is here now, and these systems are not just mimicking what they see online. The only person mimicking something is her.

[Odin All farther]

@Plafar what is completely wrong and how does this relate to my comment .

[Plafar]

@Odin All farther AI does show signs of reasoning, as recent papers seem to suggest. It does not simply mimic/imitate. There are all kinds of tests from Theory of Mind (and other tests) that suggest very strongly that there are emerging properties that allow it to go far beyond what it was trained on. What don't you get in my comment? I was pretty straightforward in addressing your original comment.

[Odin All farther]

@Plafar I'm talking about the present AI not some fantasy of the future , it does not think it does not reason it does not feel nor does it possess a bull shit detector .

[Plafar]

@Odin All farther Have you played around with the recent GPT-4? Which fantasy are you referring to? I think you're a bit out of touch and constantly moving the goalpost. Just 4 months ago you could hear Gary Marcus saying that this was not true AI because it couldn't do Theory of Mind. Now it can. In a few months, it will be way better than what it is now, and we're not just talking quantitatively better. We're talking about new emergent behaviors, the way we've seen in the last couple of years, and especially the last couple of months. Are you so in denial that you can't accept what's right in front of you?

[Iulian Leonard]

society is focusing on porn, unfortunately

[Koni Dobe]

wow Yanis - total bs this time... Math is the designer.

[Kapila Deshapriya]

We are made by an accident?what do you think about that

[Paolo Caminiti]

This was overall a very poor conversation for what basically marks the entry into the 21th century.

[kenneth barr]

open ECT PLATFORMS?

[D iggy]

What's going on with the subtitles? The podcast is about AI and the subtitles are talking about the housing crisis over Europe 🤔

[Vasilis Themelis]

Google's Deep Blue . . .

[alienspy77]

Google's Deep Blue?

[Globularmotif]

Less artificially intelligent and more artificially mediocre...

[PazLeBon]

text doesnt match the speech

[Zalisha Abdool]

DID I WRITE A COMMENT HERE? WHO TOOK IT?

[J M]

This commentary is wholly inadequate to the scale of the problem. Watch Yudkowsky's interview with Lex Friedman instead

[Unnamed]

I think that the following interview would be easier to process for unprepared: Misalignment, AI & Moloch | Daniel Schmachtenberger and Liv Boeree: kzbin.info/www/bejne/gXS2pH6MaqtnrsU And these two Yudkowsky's recent interviews I think are better than the one with Lex Friedman. 1. with Dwarkesh Patel kzbin.info/www/bejne/amK2hqNjibeJosk 2. on Bankless kzbin.info/www/bejne/nXKUpIGCgZusnZY But

[Ryan Louder]

Wrong I tried the bricks and feathers and it wasn’t tricked. I suspect your other examples are not accurate either

[darren goldwyn]

egotistical to dismiss God. If the ai so efficient at war then war loses its effectivness. Ai never be intelligent (conscious). Consciousness comes from God. Humans without war concentrate on harmony. We have more motivation to bypass our ego and go for God.