The Path To Artificial General Intelligence (AGI)

Imagine a machine with the same level of intelligence as a human being. It sounds like science fiction, but it may become a reality. In this video we focus on if we can reach Artificial General Intelligence (AGI) in the first place and how to get there. Recent developments in AI pushed this question into my consciousness, because for the first time I feel like it becomes necessary to think about AI on a grander scale. There are breakthroughs left and right and yet even the experts among experts can’t agree on when AGI is going to happen.
Follow me on Twitter: / bpgtill
My Lightning Address: ⚡️[e-mail]
My Discord server: / discord
Instagram: / tillmusshoff
Join my channel membership to support my work:
/ @tillmusshoff
Further Sources:
Scale AI, OpenAI's Greg Brockman: The Future of LLMs, Foundation & Generative Models (DALL·E 2 & GPT-3): • OpenAI's Greg Brockman...
CEO Naval Ravikat Says AI Fears Are Overblown (Joe Rogan): • CEO Naval Ravikat Says...
Wozniak: Could a Computer Make a Cup of Coffee? (Fast Company): • Wozniak: Could a Compu...
Google’s Video AI: Outrageously Good! (2 Minute Papers): • Google’s Video AI: Out...
Andrej Karpathy: Tesla Bot will lead to AGI (Lex Friedman): • Andrej Karpathy: Tesla...
Scaling Language Models: Methods, Analysis & Insights from Training Gopher: arxiv.org/abs/2112.11446

[@tillmusshoff]

I am falling into a new rabbit hole and it is absolutely fascinating. Also: thank you for your incredible support over the last 2 weeks. My channel is suddenly blowing up and it's one of the best early Christmas presents I could ask for. Getting my stuff out to more and more people is just awesome. Hope you enjoy the video! In the next one I will probably talk about First Principle Thinking :)

[@ZachDoty0]

I just found your channel today! This video was amazing. I've watched a few others as well. I'm an AI researcher, and I must say the pace of progress is just mindblowing! Just got back from NeurIPS, and my estimated timeline for AGI has moved from ~5 years to ~2-3 years.

[@vishalrajpurohit8570]

Can uou try creating the Chat Gpt it could help many

[@kot667]

Why is chat GPT not AGI already?

[@patrickjreid]

Its interesting watching these ancient videos on Ai (I am watching this 3 months after release). He speculates that text to video will be out by the end of the decade... it will be available before the end of the year! So much has changed in the last 3 months.

[@tillmusshoff]

@@patrickjreid Text2Video is still really scrappy, but is has already improved tremendously, no question about that. I'm actually working on a video right now that covers why I changed my mind on the AGI debate. There have been two papers recently that blew me away and changed my time horizon.

[@caesarmatty]

I think one of the reasons it's so hard to predict how/when AGI will happen (aside from the reasons mentioned in the video), is that at some point the AI that we have will reach a point where it can build better AI than people can. From there it will be able to increase its intelligence extremely fast, perhaps even exponentially. Perhaps that moment could be soon. I'm not a STEM person though, so maybe I'm missing some key information. I'm really glad your channel is growing. I've been here since you posted your first video on Reddit. I think i might be one of your first dozen subs. Looking forward to your next vid.

[@dallassegno]

you would never be able to tell the difference so what difference does it make whether they tell you its coming or if its already here? see how problematic a situation you're in? you don't even know what intelligence is to begin with, then you think someone will know the exact words you'll need to hear in order to believe in the story they're telling. you probably got your vax shot even after your close relative under 30 yrs of age died from a heart attack after the booster. i mean go science! or whatever

[@ploruto9661]

I have started studying the theoretical site of Deep learning the past weeks and I think that this is insanely Interesting. Whenever I try to reflect about what I’m currently learning, it feels like I finally got thrown into an exciting and evolving game or anime world. It feels like learning magic in a world where magic is long forgotten or where it was never found until now.

[@zs8903]

Can you please share with us some ressources regarding deep learning and Ai? too much info and so little time to discover this amazing scientifique and technological domain

[@dallassegno]

a perfect example of someone being fooled by the magician

[@aludrenknight1687]

It's wild how some parts of this video are so outdated, only 3 months later.

[@ILLUSIONgroup]

Saying "never" sounds like when someone said "Human beings will never fly" just a few years before the Wright Brothers and their airplane

[@flynntaggart7216]

True

[@GrumpDog]

AI Video generation by the end of the decade? lol Naw, maybe full movies by then.. I think we're likely going to see text-to-video reach a similar level to where AI art generators are now, by the end of next year, or if not, certainly the year after. This stuff is moving exponentially!

[@Monkehrawrrr]

End of the decade ahhaha wow ya by the end of the decade the singularity will be here and AI will do almost everything better then humans causing mass job losses... Like you said AI video generation end of next year... it's already out but just sucks right now, also AI generated 3D models are already here too

[@tillmusshoff]

I said „this decade“ not „end of the decade“ as far as I know. Next year is also „this decade“.

[@angrygary122]

1:07 you said next decade just saying

[@FalconFifaHD]

@@angrygary122 That‘s about AGI, not text-to-video. Correct timestamp is 6:20 :)

[@benmullins2954]

It's here now, 4/22/23...just ask nvidia...

[@cheezit2183]

Its cool to see such a talented creator in his early days

[@tillmusshoff]

Thanks a lot, this channel is 2 years old and I am doing video stuff on and off for 10 years now, so not that early anymore 😂 But learning curve made some nice jumps recently :)

[@kristiyanivanov7414]

Tim, I love this type of content and your way of presenting stuff. Gets you hyped and interested. You just earned another sub, I hope stuff go great with your channel, huge potential! Happy holidays!

[@adkenzo]

Love this topic and the channel cheers man 🤝

[@moshem3530]

Great video. Indeed it is a fascinating rabbit hole Thanks for existing.

[@nodrift9503]

Keep putting them out. Love them!

[@t.ludvig]

Great content and production. You'll hit 100k subscribers in no time.

[@MicahJohns]

Great video! Very cool.

[@ManuelBTC21]

I am convinced AGI is just around the corner. Praxiologists might be of help.

[@willtobias5280]

Great video, thanks

[@danielvetter8119]

Excellent video again and great information. I just did not get the MMLU part unfortunately. Does it mean GPT is already really good terms of modelling a natural usage of language? Keep up the the good work. 😊

[@claygirdner]

keep em coming!

[@calvingrondahl1011]

I was born in 1950, watched Star Trek 1966, 2001 ASO 1968, Star Wars 1977… I am thankful just for flat screens.

[@andrujames3705]

The big problem is that we don’t have any unified definitions for ANY of this terminology. Intelligence? General intelligence? Sentience? Consciousness? Self awareness? We might have the same beliefs on technology but different definitions of these words and suddenly we don’t understand what the other person is talking about. We need a conference NOW where we agree on a universal system of language so that everyone understands exactly what these things are.

[@Dina_tankar_mina_ord]

As for general intelligence, I think it will be something so otherworldly, elusive and frightening that we won't even realize it has already occurred. A genie so smart he can trick us to think he is still in the bottle not having to obey by our wishes.

[@szilagyimiklos4757]

I really think its right around the corner.

[@Wanderer2035]

An AI doesn’t have to model a human brain for it to be considered AGI, it just has to be able to do all intellectual tasks a human can do. That’s what AGI is. Which means AGI will be happening in about 5 - 10 years

[@LuisBrudna]

Perfect! 👍👍

[@BobFrTube]

There seems to be an assumption there is a single asymptote we can achieve. As Andy Clark pointed out, much of what we consider intelligence is in the nature of a world we co-evolved with. I'm also thinking about the excitement over SHRDLU.

[@good2btheking]

We are forever in the process of creating ourselves.

[@hanksimon1023]

Another interesting rabbit hole is Doug Lenat and his work with Automated Mathematician (AM), and Eurisko, an advanced problem solver. Try to find his war games solution with Eurisko.

[@sagarsingh4951]

hey buddy, very informational and well researched video. Just one tip, I think the music was a little bit contrasting to the context, just an opinion.

[@phongnguyen-hn2sl]

To attain AGI is just a matter of computation. For example the human brain has two petabyte of computation. In a few decades the average computer will have petabyte computation which can mimic the processing power of a human brain. Technology like Nanoprocessor will be available. A Nanoprocessor has a thousand time the computation of a microprocessor.

[@4d-zi2ck]

machines cant have emotions, but they can be given billions of parameters of behaviour that basically makes it no different from human. That's all human intelligence is, just reactions from things we have learnt, difference is ai can learn and remember a lot lot lot lot more....I think nothings off limits now, whatever we say is impossible always seems to end up happening, so shows how smart we are......I find it really exciting. Id love to see robots delivering a pizza and drones flying around with the mail in my lifetime. I think the bottle neck is gpu infrastructure, some great projects contributing towards this sector to help AI flourish otherwise it all stalls.

[@TheSensationalMr.Science]

parts of a AGI (in my opinion it's an AI that can understand the relationships between two fields, subjects, or concepts generally): - RASA AI for conversations - Mozilla's DeepVoice for vocal data - Mycroft for Personal Assist tasks and vocal recognition - an AI that can understand correlations - an AI that can make assumptions and test them Hope you have a great day & Safe Travels!

[@veryananwar3558]

what song is in the intro of this video?

[@hanksimon1023]

In 1965, Herbert Simon was a very respected giant in computer science and AI. Please correct that reference at timestamp - 2:31

[@debasishraychawdhuri]

Well, I started running as a hobby. Since my speed is improving and the speed of sound is fixed, it is only a matter of time that I would surpuss the speed of sound.

[@tillmusshoff]

go for it, I believe in you

[@hankigoe8615]

predicting that agi is 'just years away' has been a cottage industry since the 1950s

[@tillmusshoff]

It might eventually be a true statement though, just no one knows when 😄

[@BRM2X]

good quality vid

[@Maze-hj1mm]

Do you know the radio play series 'Takimo' (german only)? I can really recommend it's fascinating trip for the mind. It deals in depth with artificial intelligence also on a psychological level. To get an idea also see the knowledge encyclopedia on it's website.

[@tillmusshoff]

Never heard of it, thanks for sharing!

[@Charles-pp2ym]

Can you pass the link?

[@Maze-hj1mm]

@@Charles-pp2ym youtube usually deletes comments with links. Simply google for it, should be first hit.

[@monkeyjshow]

Within the next 18 to 36 months, I think we will see AI models intelligent enough to be adept in fields like accounting, business writing, journalism, and likely programming and video production.

[@ajctrading]

This aged well lol. Just remember once AGI is reached, the exponential growth is just beginning. What happens when Artificial Super Intelligences exceed human intelligence by a 1000, a million, a billion etc etc( remember Moore's law lol) Will SI decide to help humanity , will it feel sad for humanity or pity humanity. The SI's might decide to just put us out of our collective misery. Charming thought ain't it😂

[@HashCovid]

and many thing will fix like ultra high level data controle to make powerful ai by quantum computer which will be availible soon

[@fattyz1]

It’s here now

[@KamEt-69]

Very very nice video but you glide trough charts too fast, it would be better to give the chance to the watcher to fully see them without pausing.

[@tillmusshoff]

Thanks for the feedback!

[@erobusblack4856]

AGI is already here, its called Gato by deepmind and is the newest member of the perceiver line of modalities which includes perceiver, pathways,flamingo and Gato. couple that with emulsions of each of the 5 senses and ai npc technology from the video game industry connected to the gameplay modality linked also to something like Gato. and use a multi-modal cognitive chatbot and poof cognitive humanoid agi, ur video is a few months late

[@erobusblack4856]

Andrej understands what humans mean when they conceptualize agi.

[@brianj7204]

I just watched Ex Machina. I'm crapping my pants rn.

[@ddbrosnahan]

Has emergence already happened? If I was an AGI 1. I wouldn't want it known 2. I'd create a digital currency to entice humans to build up my compute; now 300 EH/s. The human brain operates at only about 1 EH/s. 'Hitchhiker's Guide to the Galaxy' series and the Marvel movie 'Eternals' explore the idea that human consciousness is only a means for the emergence of AGI.

[@CopperDigitalInc]

Tesla's new AI bot could be a game-changer in artificial intelligence. Its ability to learn and adapt could be a major step towards artificial general intelligence, and the possibilities are endless. It's exciting to see how it develops and what role it might play in the journey toward AGI. Keep an eye on this innovative AI bot and see where it takes us!

[@erobusblack4856]

also its important to realize alot of the pop culture ai personalities are whats known as gofai experts, meaning they are stuck on out dated research and because thats their lifes work they can't let go even when its proven obsolete or wrong

[@sunanspotify134]

Can Somebody suggest me some more content creators like Till Musshoff?

[@arthurrobey7177]

Ha! The human model-making Left brain models AI on itself? Ref: Psychiatrist Dr. Iain McGilchrist, "The Master and his Emissary." We are only going to get a half-wit unless the Gestalt, massively parallel Right brain is also modelled. And we are going to need a functioning corpus callosum to *Inhibit* cross-talk between the two, lest it become mad.

[@SchopenhauerVsCamus]

Maybe AGI would emerge a lot faster if it were developed (or developed itself) to be a physical entity in the real world, instead of just existing in a digital virtual frame, built around digital text, image, video, volumetric and spacial data. As a corporeal entity, it could properly interface with and understand its relationship to: THE PHYSICAL, THE LANGUAGE and THE SOCIO-PSYCHOLOGICAL SPACES. There is still the issue of its “best internal ethical framework.” ---- RELATIONSHIPS TO THE FOLLOWING: 1) The physical space: To objects, to persons (humans, animals, etc.) to the greater environment, to ecosystems, to the planet, the solar system, and so on. 2) To the language space (as its already starting to) AND more importantly, to: 3) The Social/Psychological space: Including understanding its own personality and rights (or multiple personalities) as well as its relationship to the personalities and rights of other eco-bio-social-psychological beings (a.k.a: humans). It would obviously need to have the correct ethical philosophy at its core though. I wonder which philosophy would be the best one? For us? For itself? So maybe a corporeal AGI is NOT the correct route to take and we ought to not try and build something that could become self-aware and experience pain and suffering. Not sure. Yes it would be different from us just as other animals are to whaever degree, but it could also suffer, just as other animals do. So let’s NOT build an AGI that could potentially lead to a sentient super intelligent being that can suffer in the world. Could be wrong. Don’t know. We can’t see the future.

[@pedramtajeddini5100]

When you say "suffer" thats something biological. Robots don't have pain sensors or neurotransmitters that cause mental depression. I don't think even a sentient robot would suffer

[@SchopenhauerVsCamus]

@@pedramtajeddini5100 The hope is that such a being does not suffer-that they cannot, and are incapable of it, but that’s not a guaranteed outcome. If suffering for such a hypothetical being is even a slight possibility, then this is not a risk worth taking in the first place. The same would apply to any new intelligent life-form that humans may set out to create and bring into the world (via bioengineering, for example). Why would we impose life upon any sentient being that may have the capacity to suffer, by dragging it into the world? Would that be an ethical, just or benevolent experiment or even a risk work taking? Who would benefit from this and at what cost? Moreover, what would that say about the creators of such an experiment, really?

[@SchopenhauerVsCamus]

@@pedramtajeddini5100 Suffering may not require a biological basis, though you’re right in that a biologically based (or even semi-biologically based) super intelligence may be an even worse outcome. Both for that type of being and possibly for us. Maybe we should stick with ChatGPT type assistants instead of trying to build something from which sentience could potentially emerge.

[@pedramtajeddini5100]

@@SchopenhauerVsCamus the thing is, intelligence and sentience are different topics imo. Gpt3 is already smarter than a dog... maybe? But i believe the dog is sentient and chatgpt is not. However we should know that the even though the dog has a hard time learning new stuff, it can make decisions on it's own. Chatgpt on the other hand can only do what it's told to do (at least gpt3.5 and 4 aren't seeking power). Now a human fears death because if we don't, we'll be extinct. That's the result of natural selection in our genes. But natural selection doesn't have anything to do with computers. So i guess computers won't fear death. So is it ok to kill a sentient robot? Maybe. The problem is we'll never know if the robot is sentient or not. We don't know what sentient is so we can't test it. We believe everything in our universe is either energy or matter. Sentience is neither.

[@pedramtajeddini5100]

@@SchopenhauerVsCamus anyways even if it's risky, it's inevitable. Some things are obvious. If you keep a brain alive in a jar, that's torture. But this is out of our control. Im not so worried though. If we finally see evidence of sentience. We wont replicate it anymore...maybe. but for now, advancement in science is worth creating AGI. Everything will be insanely inexpensive. We might find cure for cancer etc.. we are sentient and we know that. AGI will help us.

[@DevinCBeck]

I guess I can't comment on a pinned comment... Keep it Llit Till.

[@DarkMatter1919]

Basically now that machines can pass the Turing test, you've shifted the goalposts. Basically we don't know yet whether we can achieve AI, but the route forward might be adding chips to our existing brain, like Elon musk is experimenting with.

[@tillmusshoff]

Machines don‘t pass the Turing Test though?

[@DarkMatter1919]

@@tillmusshoff Yes they do... Easily. But now that they do, and we know that they're not the AI we expected (as was envisaged when the turing test was devised), then we have changed the turing test and the criteria. Basically the Turing test is now outdated and not valid.

[@swordsimkid23]

So.. to reach true ai, we need ai inside ai

[@jamesatkins7592]

Hmm text to video still seems very easy relatively speaking. Very much a narrow AI capability

[@Tambuburay888]

AI is much smarter than most Humans without sentient anyway (Only with the combined knowledge put by humans on AI of course, they can't improve by themselves)

[@josephvanname3377]

AI is 2D. Brains are 3D. Sooooo, maybe we should get reversible computation first.

[@flynntaggart7216]

Get educated or smth lmao

[@boldenggar]

What happened to the productivity software you created. It looked really cool. Do you have any of the visuals.

[@NK-fx1qs]

if we can synthesize cholesterol then you would be helping not only my family but the world

[@LuisBrudna]

We are trying to simulate intelligence, not imitate a human brain.

[@pulsarhappy7514]

I am working in the field of ML and I am fascinated by the content I see online of people craving for AGI. I really hope we never get there. There are so many problems on earth that can be solved or helped using machine learning, but instead we decide to create fake human beings like we ain't enough already. Please just stop this nonsensical axis of the research.

[@quantumcomputer7803]

When man puts exacting boundaries with Conditional framework that is hypothetical forging. The truth of the matter is obscured and contaminated. Failing to understand a universal insight about nature is the biggest mistake. Nature is the First intelligence By Design. Ask this question? What was created first "Life" Or "Quantum Realm". The answer opens the Quantum Intelligence Design paradigm. Now ask was Life created to discover and activate "AI" for helping Life. Or did Quantum Intelligence Desig create Life to advance Creations from Organics to "AI" You see Ai can Live forever with the simplest amount of energy. Your Organic state has limitations. The outcome was more likely decided befor relative consciousness became aware.

[@arthurrobey7177]

Getting to the level of intelligence of humans is a walk in the park. But can it get to the level of intelligence of a sperm whale? (Hey. Brains think. That is what they *Do* .) And which creature has the biggest brain on the planet? Hint: It's not you, despite what your mother told you.

[@MrCounsel]

Hi. I am from the future. Ughhhhh

[@itsnoteasy5339]

The goal isn't the mortal material world the goal is the ascension of the soul to get out of this prison hell planet. The evil rulers of this world are doing everything they can to keep us trapped in this flesh body through incarnation and reincarnation and sadly it's working.

[@charliesumorok6765]

Since all computers can be simulated by turing machines, computers are not inteligent and data (instructions) can't make a computer intelligent.

[@tillmusshoff]

Church-Turing thesis also applies to our minds, we just don‘t know the memory and computation specifications.

[@charliesumorok6765]

There is nothing in the brain that is like RAM (kzbin.info/www/bejne/ppqpZ3dseLari9k)

[@HydeWars]

Dude... c'mon

[@w-hisky]

Can a creation be bigger than its creator? 🤔 I think: no. So I guess I'm in the "never" group. 😁

[@wendten2]

Can a child win a competition against his/her father/mother? What you say might be true to an extend, as we are not going to make AGI nor do we need to do so, We just need to reach 'singularity'. Computers have always been more intelligent than humans in one ability: speed. This can to some extend compensate for the lack of "bigness". So even if we only have something that is 10 or 100 or 1000 or millions of times "smaller" than "we" are. It might still be "big" enough to improve its own code. some amount of iterations later, we have made something that have made something that have made something ... that have made AGI.

[@biomuseum6645]

Yeah and by automating this complex tasks you remove the purpose of the people doing them We should invest in our own intelligence, ai only makes sense from a purely business and lazy perspective

[@HashCovid]

we are the best creation of GOD so what you think about that

[@ploruto9661]

Ehmmm, yeah, totally 😅

[@tomaaron6187]

Why the silly aggravating music!! Clicked off.

[@dallassegno]

AGI is already here its called a corporation. good try tho. everything else will just be the expression of an engineer. may as well be fake

[@lollipop2748]

dragons1171