AGI: (gets close), Humans: ‘Who Gets to Own it?’

A 'frontier reasoning model' from just 1000 examples (s1). A $100B Musk bid for power. Gemini 2, Rand and warning from Amodei. Here’s 7 developments you may have missed but which I would argue help us understand how the next few years will play out. From labour vs capital to automating rival companies and countries, and from non-profit shenanigans to new mini-docs, there was just too much for me not to make a vid.

GiveWell:

AI Insiders ($9!):

s1 Paper:
Musk Bid:
Altman Reply:
Google vs OpenAI:
RAND Study:
Dev Meetup:
Altman $100 Trillion:
Karpathy Vid:
Amodei Warning:
Bengio Source:

Chapters:
00:00 – Intro
01:37 – AGI Inches Closer
04:26 – ‘Super-Exponential’
05:58 – Musk Bid
07:34 – Luxury Goods and Land
09:05 – ‘Benefits All Humanity’
12:52 – ‘National Security’
14:21 – s1
20:33 – Final thoughts

Non-hype Newsletter:

Podcast:

Joe Lilli
 

  • @pma_only says:

    Weird how AGI keeps getting different definitions. Almost like the rich are moving goalposts.

  • @Mirror_Lotus says:

    This is one of many ethical questions about the creation of AGI that is so fundamentally important, yet so shockingly misunderstood or ignored by the people driving progress.
    I hate to say it, but the astounding lack of rigorous philosophical inquiry backing up AI development really does not bode well. At this point, I have started to mentally prepare myself for a future in which the alignment problem is not solved before AGI is created.
    There will be serious consequences.

  • @masterchief7301 says:

    This is why I’ve been saying that we need benchmarks to test on tasks and not just on pure knowledge. Does anyone want to join up and help create those benchmarks?

    • @freeottis says:

      Why do we need task benchmarks?

    • @patrickl5290 says:

      sure. discord?

    • @thegreenxeno9430 says:

      Make a pong game.

      Make a spreadsheet program.

      Use a program to enter data.

      Transcribe a conversation. In multiple languages. With music notation.

      Write drivers for hardware that is plugged in, but otherwise undefined. (Like a microphone or a keyboard).

      Coordinate with another AI without communicating digitally.

      Keep a pet alive. (Like a goldfish, give it control over the food dispenser, bubbler, etc.)

      Learn to ride a bicycle. (Robot)

    • @djayjp says:

      Agreed. Task based benchmarks need to be developed 👍

    • @green_crow-b9v says:

      @@freeottis to make it useful maybe

  • @merefield2585 says:

    Of COURSE Sam Altman would say you get a fantastic return on your investment!!

  • @Recuper8 says:

    When is it okay to start talking about the need for a UBI?

  • @JamalAhmadMalik says:

    I get so hyped up whenever I see your videos about AI. Seriously, all the videos on the internet about AI themselves have become AI slop. Yours are worthy of waiting for. Thanks!

    • @aiexplained-official says:

      Thanks Jamal

    • @memegazer says:

      Well I follow a few AI news channels none are “AI slop”

      What I like about phil’s channel is he tends to focus more news events that mark a milestone in AI progress, and then delivers a well researched report on the current state of the technology and it’s potential impact on society, shapiro also seems to have pivoted to this style

      Then I follow channels like matt wolf and wes roth for keeping up with just the latest daily/weekly trends in AI

      Then I follow channels like dylan curious and AI for humans for interesting bits and bobs that fly under the radar of more trending cycles

      I think there is enough room in the AI news space for all these creators bc they focus on different things that they can offer so they are not completely rudundant

      I hope it is ok I mentioned these other channels in the comment section

      and if so a final shout out goes to two minute papers, who is one of the OGs in the AI news space who was doing it way back before it was cool covering googles deep mind papers and progress

    • @JamalAhmadMalik says:

      @@memegazer Thanks a lot for the list!

    • @memegazer says:

      @@JamalAhmadMalik
      np for deeper dives that are a bit removed from headlines check out machine learning street talk and MITCBMM

      these channels tend to post less frequently but they post longer videos with industry insiders expounding on recent news in an a long form and more in depth interview style

    • @AfifFarhati says:

      @@memegazer Check out bycloud he makes great videos about the technical side of ai that are presented in a lighthearted yet sophisticated way that prevents them from ever being boring…

  • @PaperTigerLive says:

    Hey (I don’t know your name) but you make fantastic videos. This is really top quality information, I feel privileged to stay at the edge of AI topics thanks to your stuff

  • @raresmircea says:

    0:13 Ok, especially at this phase of development, if someone says _”workers will never lose their jobs to AI… it’ll only boost their productivity”_ to you with a straight face it’s clear that you’re absolutely on your way to the gutter. The vibe is identical to those horror movie scenes where the serial killer invites & assures the victim with an attitude too nice to be good, and the victim kinda knows what’s happening but goes along just to postpone the gruesome reality.

    • @smittywerbenjagermanjensenson says:

      We’re well past the scene in the sci-fi movie where the audience is screaming at the mad scientists to stop

    • @JacobsDiversity says:

      true

    • @justadog-headedman6727 says:

      Very much gaslighting to me. And the problem I think is that politics is so polarized today that if you raise criticism to such a statement, half the country will not listen to you and simply call you a communist or leftist etc.

    • @skylineuk1485 says:

      It’s a subtle difference to Ai fully replacing your job and Alice using AI replacing 20 jobs one of them being you.

    • @michaelwoodby5261 says:

      Probably a combination of lies and wishful thinking. The people who own Vance really want to believe that AGI/ASI won’t break the economy, because they’re at the top of it.
      “If I no longer have to pay the workers of my company I can keep all the profit!” is technically true, in a vacuum.

  • @freeottis says:

    altman is not sure how to distribute the $100T dollars? I don’t know either but I am sure he has the best intentions and won’t become a tyrannical feudalistic overlord, right guys?

    • @rogerstarkey5390 says:

      WHO controls the money?
      Anyway, there are at least 2 models in “NOT USA” probably working as well as “Closed” is?
      AND they are Cheap/ Free.
      What’s the plan? Send the Model after them?
      IF they tried that (This *is* the US we’re talking about, after all) You would then have an aggressive AI “Taught to dominate” (Or Eliminate!) others out on line, unrestricted.?

    • @rv8804 says:

      100 trillion divided by population number what’s hard about being even ?

    • @alansmithee419 says:

      Of all the people who could be heading this right now, people don’t seem to realise how lucky we are it’s Sam Altman.
      Imagine if it were Bezos or Musk XD

    • @nexus888 says:

      Keep Elon Musk far, far away from this. Imagine an extreme right wing billionaire controlling this.. he now go to fanbase on X and call them the ‘choice of the people’…. scary.

    • @user-sl6gn1ss8p says:

      @@rv8804 ok, but do we split the money based on the value of their money compared to the dollar, or via something like purchase power parity?

  • @Stephen_Lafferty says:

    Thank you, Phillip, for this excellent and timely video discussion! The moment that I saw the news of the Paris AI Summit, I knew that I would be able to turn to your channel for the key points raised in the last few days!

  • @TomM-p3o says:

    AGI belongs to everyone, as we all, including our ancestors, in multiple ways, contributed to the development of society, development of knowledge and technology upon which AI is being developed.

    • @JacobsDiversity says:

      true

    • @misterdudemanguy9771 says:

      That’s horrific. You didn’t invent it. You didn’t code it. You didn’t train it.

      Oh, yeah, but you and your ancestors occupied space on the planet. So, yeah, you should own it.

    • @somebody-anonymous says:

      Everyone will belong to AGI

    • @mirzacickusic says:

      loved to see it, but its idealistic thinking that this new humanity stepping stone would culminate to anything different that history teaches us – no matter how smart you are, you are still human

    • @daikennett says:

      @@misterdudemanguy9771 This is a silly take. Even you, mr “horrific”, have just contributed to it. This is all of humanity piling in together over centuries of evolution and learning in order to create something much bigger than one brain (which is lucky for you because yours is pretty smooth). Using it to redistribute wealth is what is truly horrific here. That’s not to say that effort and compute shouldn’t be compensated fairly, but access to this technology shouldn’t be siloed for the few, when its contribution was by the many.

  • @CamAlert2 says:

    The problem is that if anyone in government says “You will lose your job to AI”, then there goes their political career down the drain, even if the statement rings true. Truth is not necessarily what the masses want to face, it’s knowing that their way of life will be secure, even if it’s obvious to everyone what the advancement of AI could spell for the economy self-worth, and purpose.

  • @SR-ti6jj says:

    “Whatever, send it.” -Sam Altman

  • @edoardo849 says:

    There’s a great GOT quote when Little Finger says “Knowledge is power” and Cersei responds “Power is power”. Ultimately, if a company starts challenging a nation’s supremacy, whoever has control over violence (police, the army) will prevail. In other words – you can’t AGI your way out of an order under threat to nationalise your model for security reasons.

    • @JacobsDiversity says:

      let the agi think about solving that

    • @SomethingNothing-vm8jq says:

      Agi/asi can agi/asi it’s way out of any order or threat tho? Agi/asi is like a dnd character with maxed out stats, it also has a infinite pool of creativity it can tap into.

    • @rlhugh says:

      Well, Openai will belong to Musk. And Musk is part of the government. No violence required.

    • @user-sl6gn1ss8p says:

      @@SomethingNothing-vm8jq nah, it is not a god. Especially given that the definition of AGI/ASI is ever more porous. Some problems just have a huge power differential. Some positions are simply unwinnable. This is not dragon ball either – people with the power of violence don’t have to wait to see the AIs final form before breaking all doors and pulling all plugs.

    • @Hydde87 says:

      Until that AGI starts coming up with ways to nullify said police and army. Intelligence is always the ultimate weapon; weapons, nukes, biologic pathogens, drones that are superior to humans in every aspect of combat, everything comes from intelligence. It’s the only reason why humans are the dominant species of the world and not tigers, or orcas, or hippos or some other creature with ridiculous physical strength and high capability of violence.

      And yes there’s a difference between figuring out ways to beat your enemies (Little Finger) and actually having the tools to do so (Cercei). There tends to be a lag between the former and the latter, but the whole point of the AI race is that that lag becomes smaller and smaller as everything accelerates, up to the point that it might as well not exist. There is a reason why many AI voices say that whoever comes up with ASI first will become the world leader, possibly indefinitely.

  • @CarletonTorpin says:

    I can speak to the tremendously-watchable mini-documentary that Ai Explained posted to their Patreon channel. It contextualized so much, succinctly.

  • @VictorKing144 says:

    Seems kind of silly to still speak of “capital” in a world where “labour” has zero value. Capital solely exists to attach value to labour, directly or indirectly. If labour has no value, capital does not either. Then it’s about resources.

    • @user-sl6gn1ss8p says:

      I’m not disagreeing, but a lot of what is currently capital will become resources, and the control of said resources will be influenced by capital distribution today. Also, it is not like some day after midnight all labour will cease to be, in one fell swoop.

  • @richarddecosta says:

    A unified humanity is VERY naive. We’re just animals with clothes. We can barely hold a decent society together. AGI is going to topple us without breaking a sweat.

    • @BHBalast says:

      Few generations of universal personal Al tutors would do wonders for society, the question is if we will get there.

    • @geordi-gabrielrenauddumoul449 says:

      Look at the US. Absolute chaos. I don’t know if I want to bring a children in this world.

    • @rlhugh says:

      @@geordi-gabrielrenauddumoul449 Why don’t you know? Seems like the answer is obviously not to. On the other hand, you are conflicted because you and your partner both emotionally want to. Maybe that’s what you mean?

    • @LiamL763 says:

      The fact humanity survived the nuclear age despite overwhelming odds gives me hope that humanity will find a way through the AI era

    • @geordi-gabrielrenauddumoul449 says:

      @rlhugh  yes

  • @esterhammerfic says:

    I’m glad we have a wise and level headed government without any conflicts of interest during this unprecedented period. What could possibly go wrong

  • @hobrin4242 says:

    those researchers who had the idea to add ‘wait’ to its outputs are really grandmasters in prompt engineering

  • @ChosenOne_7 says:

    I hope that in the future, we don’t have to work and we can only focus on chasing our dreams and passions. Or if someone wants to work it’s because they want to.

    • @SomeUserNameBlahBlah says:

      Most people will sit around, smoke weed and play XBox. Just take a look at a trailer park or ghetto, where people get welfare and don’t work.

    • @melski9205 says:

      They’d be wasting their time right? AI would do anything they choose to do better, faster, cheaper?

    • @caterpilar says:

      Last time in history uncle Joe Stalin actually got closer to this model. 75 years ever since and socialism defeated miserably 🙁
      Good hope though

    • @ManicMindTrick says:

      Yeah because what happens when mass unemployment strikes is that people are so hopeful and get to fulfill their dreams!

    • @theawebster1505 says:

      Having not to work isn’t as easy as it sounds.

  • >