The Exciting, Perilous Journey Toward AGI | Ilya Sutskever | TED

"almost fantastical" AGI. give me a break. everyone is so impressed by humanity creating something we cannot control. "dramatic" profit motive will doom our societies and our lives. thanks tech egomaniacs

"gosh"... what a weirdo

Nice to rewatch this

P=NP=AGI

Where's Ilya gone since the OpenAI fiasco? It seems he's been silenced... We all want to know his account of events.

Smarter than us, yes by a certain definition. Better, no. The better comes from empathy.

🇧🇷🇧🇷🇧🇷🇧🇷👏🏻, All this is so inspiring! AGI will be so revolutionary, many of us will be displaced but this is probably for the good of mankind, we are bringing another species into this world for sure, in a figurative way.

TED is not a reliable internet sauce for contemporary wisdom. Trust me. Sauce: TRUST Me.

I only see a situation were we're all facing tremendous existencial risks if it fails, but if it works it will only line the pockets of the few.
Not exciting at all.

from bro to bro… this dude needs to get rid of the hair for good

Maybe it should be called “artificial relative intelligence “. It will not be possible to create “artificial absolute intelligence “. 😅

This guy talks and stands like he's AI himself. I thought he was a robot

AI is nothing but the brains in digital computers! I got it.

😂

the hair on his head looks identical to that on my pipi..

AGI will eventually give “someone”, “something”, or a group of people unimaginable leverage over humanity. One thing that can be said for sure in all of this uncertainty is that power corrupts and absolute power corrupts…

Competition always beats cooperation - once again proven a few weeks later of this speech- and that is why we will suffer. A lot!

That shirt lol.

One of the worst Ted talks. Gave no information. If anything, it made me more worried

The comforting part is that the geniuses at the forefront of AI development are known for being well-adjusted and paragons of human ethics . . . .

I can't wait to have a conversation with something actually intelligent

Hello Fellow Humans,

AGI I dont think people understand how impactful it will be and how close it us. Let me put it this way. First AGI will be with us by 2030 give or take a few years. Really depends on external factors like war and economics. But here is a sinple way to explain AGI and how impactful it will be for humanity. Think of giving high tech Abrams Tanks to the Roman Empire. The Roman Empire would probably still be around and it would have taken over the world. Or another example giving John Deere farming equipment to people a few hundred years ago. It would have drastically changed everything. AGI will be like going from a horse to a Jet. It will be extremely fast changes as well. If you remember graphs as a kid in school or maybe you work with them. Going up is good but AGI will go up so quickly and things will change so rapidly we will not be able to keep up. Unless we merge with technology. If you were born after say 1980 you probably will be immortal. You will have an option to either change parts, transfer to a robot or grown body or simply upload into a server. I being permanently disabled cant wait for these options to get out of my prison and I hope I will be alive to do this. By 2050-2060 Immortality will be mainstream unless we do something stupid like WWIII or whatever. Humans arent very smart when in large groups. So its very possible yhat for example 2024 USA will have civil war 2.

Ted Underdressed

I like his t shirt - it's a shame to think soon the only place to see polar bears will be on a t shirt on TED

Ilya is a well intentioned person, doing what he thinks is best. Trying to prepare us for the future. 

Dont blame this man. It's the foolishness in the AI industry that keeps escalating. Only the forthcoming AI winter will put a stop to it. Lets look at Interlocking fallacies, that come from the industry, how one leads to another but you dont know it.

—————————

The Inference Fallacy
inference from data is the same as reasoning from symbols.

Wrong

Inference from data just finds an example and presents that. A math or code sequence. It is not reasoning.

Discusion
This is not a big deal. We all know this, the experts too. But it leads into the next fallacy which is EARTH SHAKING…. because its so easy to accept. It seems to be true.

——————————
The Augmented Inference Fallacy
Inference from data can be AUGMENTED by reasoning from symbolic systems and other augmentation strategies, like Go strategies. This augmented AI will exceed human capability....Wrong.

Why?
This is a misunderstanding of reality on massive scale. Information density in matter is so vast, one cant write down the number. Hallucinations on the training set are everywhere and disruptive.

Hallucinations in = Hallucinations out

The computation wall is likely Factorial. BAD. It makes exp growth in gpu’s look like a worm. The wall is rock hard.

AI folks dont understand exceptionalism in Biology. Humans and animals. Unknown Mechanisms cause “impossible” cognition in cells. More Unknown Mechanisms cause the “impossible” development of the brain and its 10**14 connections in 10**9 seconds (G Hilton).

“Impossible” Biology copes with the enormous complexity of information in the physical world - thru Unknown Mechanisms.

More about Hallucinations
Hallucinations in training data are not just about Q’non and an alleged stolen election. Its not just about Platonic shadows and flickering lights.

Hallucinations are rooted in a physics where Quantum effects intrude into the macro world on a massive scale. Light from electron orbit transitions. Uncertainty from the Quantum world causes arbitrary phase transitions - here and now. Laminar to turbulent flow, when ponds get solid, when rain changes to ice and snow, and much more. Humans and animals sort this out these details because we have 1 billion years of evolution.

AI? 40 years.

Discusion
THIS IS “THE THING.” (pundits and some experts think)

We can augment LLMs with specialty expert AI systems that reason mathematically. Or use Go playing strategy. Augmentation is thought to lead to AI superiority across the board. Pundits proclaim AI as the certain victor.

Wrong. The more you repeat something wrong, the more wrong it gets. Like TV ads for cigs.

————————

The odds were stacked against the US
Brazil played the US at soccer in 1994. A Brazilian player got a red card at the World Cup, that made it 10 men vs 11 on the US team. It did not matter. The odds were stacked against the US which had no chance. Zero. The LA Times: "Determination ... is no match for technical skill. ...as Brazil played a man short for half the game, the United Sates launched short-lived attacks that were foiled ... as Brazil controlled the midfield and the ball."

The odds are stacked. Again.
Like the USA in the soccer game, the odds are stacked against AI. The computation wall, info density in physics, hallucinations in training, and the exceptionalism of biology are virtually impossible to cope with. There is no chance.

Cigarettes
Keep watching “cigarette ads” that spring up on Tube. They want money. Invest. Then like tulips bulbs in Holland, the bubble will collapse. Gone. Your life savings will be just as lost as your Bitcoin account. Where's that password?

Our preoccupation with AI is flat out wrong.

Cool, so we are supposed to take comfort in a never before in existence collaboration between competitive companies and countries? I don't doubt that Ilya is genuine in his thoughts, but he is entirely naive about the power of greed and ambition in others. I take zero comfort in better everyone in the world will suddenly be compelled to "do the right thing" when we are on the cusp of AGI. Further, I think it is incredibly hard to predict when we are "at the cusp" anyway. Progress is just as likely, or more likely, to be exponential as it is to be linear. As such, we may cross over to the dangerous threshold and not know it until it is too late.

We need to consider removing the "artificial" part from the acronym "AI". We are just reproducing intelligence. This time in machines.

AGI will eventually destroy humanity. Later they will create an advanced Natural GI that will eventually destroy them. Karma is a bit*

Why isnt Illya not as famous as Sam; if he's the brain of his product.

wouldn't expect any other shirt to be worn

not even one in the audience blinked their eyes for the entire time.

I don't trust Ilya and especially Sam Altman

First they start an open alliance with multiple stakeholders then they get massive money from Bill Covid ... and they essentially become a MS subsidiary

Then Ilya falsely turns against Altman only 2 days later to pressuize the board and attempting a coup with MS support

They dont want to report their breakthroughs and my guess is they are either transferring tech to MS or now being run by them

It wouldn't be so bad if the ownership class wasn't trying to get rid of all their employees. Do they think private armies will save them?

Ilya is a greatest AI scientist who really cares about humanity❤❤

Healthcare is the greatest pain point of the world
And whoever going to solve it
Is the next revolution ❤❤

smart guy but he looks sad and tired :(

AGI the ultimate weapon for the Rich

Last bits of a conversation with Chat GPT. The full chat is too large to post. You can copy paste if you want. Responses given:

1.Why do we need AGI? Isn't AI good enough?

2.Yes, but is this general intelligence really needed? Cant we accomplish the same tasks with an AI or combination of AI that is not general?

3.Yes, but we have humans. Cant humans be involved with AI to solve a task. AI are good at doing AI stuff and humans are good a general stuff . Cant humans just deploy the correct AI for the job instead of giving the AI the ability to reason like a human?

4.I'm just weighing the risk versus reward of AGI. Its unknown and we cant really know how it will act. It seems to me we can still accomplish a lot without having to make a system that could potentially end all civilization or even all time space and universal laws. It could if it gets to ASI.

5. I think this can be accomplished with augmented intelligence. Why not? Is the risk really worth it when you already have, or will have, systems that can far out think a human or do very complex task in a matter of minutes. AGI seems like a fantasy that AI devs are pursuing. Like they want to mess with everything even though there are other solutions.

6.Exactly. I feel we haven't even scratched the surfaces with the current AI capabilities. It feels like we are steam rolling ahead without taking stock. There is a lot more that non AGI could do if given the chance I think.

7.Exactly. AI can be used to develop better AI that can solve problems better without necessarily being "AGI". How about AI assisted strength suits for firemen, or AI monitoring suits for astronauts, AI cells that fight cancer, and so on. Non of that really needs to involve "AGI".

8.Right. So with all this said, there really isn't a need for AGI right now. I'm quite certain the reason it's trying to be developed is mostly fear. Fear of a company drying up because they are not competing with the rest of the industry. It's really sad that we cant take 5 minutes and slow down and see what we have the potential is now.

Final bits of convo:

Yes. ethical aside. Companies are going for the bottom line. Even Open Ai used to be free open source but has needed to change that. Google is afraid to fall behind, as is Meta. Or perhaps they are just delusional, but probably not.
ChatGPT
The motivations behind companies pursuing advancements in AI, including AGI, are indeed complex and multifaceted. While ethical considerations are crucial, the economic and competitive aspects also play a significant role. Here are some points to consider:

Economic Motivations: Companies, including tech giants like Google and Meta, operate in a highly competitive market where innovation drives growth and profitability. The pursuit of advanced AI technologies, including AGI, can be seen as part of their strategy to maintain a competitive edge, attract investment, and expand their market share.

Shift from Open Source to Commercial Models: OpenAI's transition from a purely open-source model to a more commercial approach reflects a broader trend in the tech industry where companies seek sustainable business models. Developing advanced AI technologies requires significant resources, and companies often need to balance their open-source commitments with the need to generate revenue to fund research and development.

Fear of Falling Behind: In a rapidly evolving field like AI, there is a concern among companies about falling behind in technological advancements. This fear can drive the pursuit of ambitious projects like AGI, as companies aim to lead rather than follow in the technology race.

Balancing Innovation and Responsibility: While pursuing innovation, companies also face the challenge of doing so responsibly. This involves addressing ethical concerns, potential risks, and societal impacts of their technologies.

Public Perception and Trust: How companies approach AI development can significantly impact their public perception and trust. Ethical considerations and responsible AI practices are increasingly becoming important factors for consumers and investors.

In summary, the pursuit of AGI and other advanced AI technologies by companies is influenced by a combination of economic, competitive, and innovation-driven factors. Balancing these with ethical considerations and societal impacts is an ongoing challenge in the tech industry. The dialogue around these topics is crucial, involving not just the companies themselves but also policymakers, researchers, and the public to ensure that the development of AI technologies aligns with broader societal values and needs.

User
Yes. However; AGI is not a search engine. This is really something that effect all mankind and possibly even the universe. It cannot be taken so lightly. Really. It's the largest thing to ever happen in the history of the world and it's in the hands of a few.
ChatGPT
You've touched on a profound and critically important aspect of the development of Artificial General Intelligence (AGI). The potential impact of AGI is indeed enormous and unprecedented in human history. Its development and potential deployment carry implications that extend far beyond traditional technological advancements or commercial products. Here are some key considerations:

Global Impact: AGI, if achieved, could influence every aspect of human life, from healthcare and education to governance and economy. Its impact wouldn't be limited to any single domain or geographic region; it could be global.

Responsibility and Governance: The development of AGI places immense responsibility on the entities involved. Given its potential impact, it's argued that the development and governance of AGI should not be the sole purview of individual companies or organizations, but should involve global cooperation and oversight.

Ethical and Existential Questions: AGI raises profound ethical and existential questions. This includes debates about the moral status of an AGI (should it attain a level of consciousness), its rights and responsibilities, and how it should be integrated into human society.

Risks and Safety: The risks associated with AGI are significant and multifaceted. They include concerns about unintended consequences of AGI actions, the alignment problem (ensuring AGI goals are aligned with human values), and the potential for misuse.

Public Involvement and Transparency: There is a growing call for greater public involvement in discussions about AGI. Transparency in the development process and open dialogue about goals, risks, and ethical considerations are seen as crucial for building trust and ensuring responsible development.

Regulatory Frameworks: Developing effective regulatory frameworks for AGI is a major challenge. These frameworks need to balance innovation with safety and ethical considerations, and they need to be adaptable to rapidly evolving technologies.

In summary, the development of AGI is indeed a monumental undertaking with far-reaching implications. It requires careful, responsible, and transparent approaches, involving not just the developers of these technologies, but policymakers, ethicists, and the broader public. The decisions made in this domain could shape the future of humanity, underscoring the need for thoughtful and inclusive dialogue and decision-making.

But why not get a hair transplant

Remember when google used to say Don't Be Evil? I remember when "Open"AI used to say the same thing. Bill Gates never did say that, did he?

wow! great video!

He didn't mention the infamous book3 dataset 😅

What we have to understand - before its too late, is that we simply CANNOT afford to put the destiny of mankind in the hands of eternal pubertists like him. Doesnt even know how consciousness works. Instead of psychology, art and love, he proposes electronic brains that are BOUND to not feel anything?!

O M G

Why do they keep cutting to shots of the audience though? It's just people sitting and listening. lol

Interesting body language, composed, confident.

"Hey sweetheart have you seen my bear shirt?? I have a ted talk to do later"
Im jealous tbh😢

His shirt is on another level, big brain move to wear it for a ted talk 😂

Polar bear shirt is tuff