The One Machine to Rule Them All - Origin Stories. Mini-Documentary on How the Founding Vision of Each AGI Lab Went Awry
Added 2025-02-10 19:30:14 +0000 UTCA different style of video! Ft. professional video editor, and focused on the 'impossibility' of AGI labs (OpenAI, DeepMind, Anthropic etc) keeping to their founding visions...
Feedback appreciated!
Comments
Just became a member. Videos like this make me feel like I should've joined a long time ago. It's a breath of fresh air to have someone actually regularly applying a critical and nuanced lens to this industry.
hyzypg
2025-03-14 03:37:18 +0000 UTCThat was excellent.
Mik Quinlan
2025-03-10 06:57:43 +0000 UTCMuch agree, would be happy to pay to be able to share this. Might have already tried to rip it a few times because this felt too important and unknown to stay in the shadows (:
Felix Akkermans
2025-03-07 13:38:42 +0000 UTCI just want to undersign all that Ethan wrote here. The cutting frequency between various clips was too high. For a far better executed example of a style your editor seemed to go for, check the AI videos on Art of the Problem YouTube channel. The end result here was too distracting for me. Also, I try to look up some of the quotes you referenced and couldn’t locate the sources. If you decide to publish this on YouTube, I would argue for providing a list of references in the description. I’d like to double check some and dive deeper into others. Pretty please? 🙏
Pavol Vaskovic
2025-02-18 14:57:48 +0000 UTCI think you make some good points, but I disagree about that last visual of Starship debris burning through the atmosphere: it's a powerful visual metaphor for the failure and breakup of the initial, more pure, perhaps more virtuous, motivations of the important players now pushing forward progress on AI. It's imperfect of course, since the actual words being spoken at that point in the video are also hopeful that maybe the rest of the story still turns out well, but far from unrelated imo.
David Higgs
2025-02-17 21:06:46 +0000 UTCFor some feedback, I think that the more narrative storytelling was generally well executed, and I think it's valuable alongside your technical deepdives. It's an interesting topic, and you brought it to life with a compelling script and a generally useful summary of the history and dynamics of the various companies and their leading figures. That said, I can't help but feel that this type of content is both not your style, and not where you bring the greatest value. Foremost, the editing really stood out as quite awkward, with the video content really not matching the audio script very effectively. One example of this are the scene grabs from the Matrix films. Being included once makes for a funny reference, but showing up several times while not seeming to carry a clear metaphorical throughline only strikes me as incoherent. Another example is the final shot: SpaceX Starship debris burning up in the atmosphere over the Turks and Caicos Islands. Apart from being related to a company owned by one of the figures you discuss, how does this related to your conclusion about money corrupting the lofty ideals of the leading AI lab founders? The words you speak are quite excellent, but as soon as I paid any attention to the video content alongside them, I couldn't help being distracted by it—I feel it detracts from more cohesive point you are making. I do understand this is your first foray into this more visual style of video editing, and that you worked with an editor on it. And for a first effort, it is indeed still good! As you continue to experiment with it, perhaps consider how you can use these other media sources to supplement what you are saying rather than just ride along in the background, One thing news organisations often do is to let the people they're talking about do the speaking—including audio-visual clips of the figures themselves could help with this disjointed feel. To conclude, I don't think you should feel the need to make this sort of content unless it's something you are passionate about or find fun. Your technical and useful AI content is why we're all here—it's the best, most accessible stuff on the internet.
Ethan Haluza-Delay
2025-02-17 00:14:48 +0000 UTCExcelent document, thank you and great job.
P4blö
2025-02-16 19:43:57 +0000 UTCMy concern is that when ASI does show up, it's going to look back over this time period and this headlong race off the edge of a cliff and justifiably think that we are complete morons. Maybe we should be trying to make a better first impression on our future overlords, and I don't think this is the way.
Jason Dowd
2025-02-15 21:12:13 +0000 UTCPlease release this so I can share it with my friends and family!
Joseph Howarth
2025-02-15 16:26:08 +0000 UTCBlimey Philip, this is brilliant. I thought I knew a lot about the history but this video was full of new details I didn't know. Thank you very much.
Kol Tregaskes
2025-02-15 15:30:15 +0000 UTCGreat job! Excellent video. Worth multiple views.
Mike Jones
2025-02-12 13:05:29 +0000 UTCI agree with this thoroughly. The above video is top notch among the crop of videos of people discussing the dark birth of AI. Yet I'd hate to see the channel pivot into that genre, other than at this quality and relative (in)frequency.
Harold Godsoe
2025-02-12 01:35:18 +0000 UTCReally entertaining and informative. More of these!
Harrison Roloff
2025-02-11 20:26:48 +0000 UTCWow man... this is your best video! I loved it 🤩 the main takeaway for me is: nobody trusts nobody, everyone is lying
Raffaele Gaito
2025-02-11 15:37:16 +0000 UTCQuite sobering. I enjoyed it, and the production values are very high. I will say though that my greatest value from your channel are the slightly more technical analyses, if only because there has been no shortage of videos with people discussing the dark birth of AI.
Erica Kane
2025-02-11 15:17:07 +0000 UTCNice work! Really insightful to see these developments all put together into a single story. :D Regarding improvement ideas: I would have loved to see a mention about how the big AGI labs recently even have reneged on their promise never to use their AIs for military purposes. That was a heavy blow to my trust in them (especially in Anthropic who seemed like the most ethical lab). Anthropic: https://investors.palantir.com/news-details/2024/Anthropic-and-Palantir-Partner-to-Bring-Claude-AI-Models-to-AWS-for-U.S.-Government-Intelligence-and-Defense-Operations/ OpenAI: https://www.cnbc.com/2024/01/16/openai-quietly-removes-ban-on-military-use-of-its-ai-tools.html Also restating the good ideas of others here: - a list of references would be awesome! - making this video public sometime (I really want to show this to a couple of AI Safety interested friends ^^ — I’d even be willing to pay sth. to gift this video. :) )
Birk Källberg
2025-02-11 13:54:47 +0000 UTCFantastic job Philip. While I love the deeply technical coverage you put out, this was a nice refreshing change to help folks understand the general landscape of how we got here. I thoroughly enjoyed it, and even though I consider myself very well informed on the subject and the history, I learned a number of new things from it. Please do more of these in addition to your technical-type content. Ultimately, I think many of your subscribers want to understand how AI impacts both their world as well as the wider world. These types of videos help put things in perspective. Kudos and thanks!
Jonathan
2025-02-11 13:22:07 +0000 UTCThis was great. Super interesting and a nice switch from your usual content. Not that your usual content isn't great, but variety is good! The main takeaway for me is that since the dawn of time politics, agendas and personal relationships dictate most things.
Neill Southgate
2025-02-11 12:15:01 +0000 UTCMasterpiece!
generousB
2025-02-11 11:22:19 +0000 UTCFantastic.
Eckart Burgwedel
2025-02-11 08:45:04 +0000 UTCThanks that was fun
Allan Mackay
2025-02-11 07:48:26 +0000 UTCSuffice to say wonderful video a real pleasure to watch.
Allan Mackay
2025-02-11 07:46:38 +0000 UTCReally really good. I think these history/minidoc style videos would be awesome additions to the video lineup. Great stuff.
Shroomdeer
2025-02-11 06:37:14 +0000 UTCThank you. Well done! I'd enjoy more of these kind of investigative pieces - cost/benefit permitting.
Gabor Melli
2025-02-11 05:52:59 +0000 UTCThis reminds me so intensely of this comic: https://www.smbc-comics.com/comic/god-4
Scott Salisbury
2025-02-11 02:27:51 +0000 UTCHeavy.
Graham Herdman
2025-02-11 00:46:52 +0000 UTCGood points. Those key AI leaders have basically academic backgrounds, except Altman, who has a financial/VC background, and of course Musk. Maybe the academic mindset is why they initially felt their AI research could be implemented by non-profits for world benefit. It's almost like they first viewed it as discovering a new equation without comprehending the vast capital expenditures (and associated entanglements) required to realize that. Re investment structure, I'm not an attorney, but I think it's possible for a non-profit to have a wholly-owned, for-profit subsidiary, which is the OpenAI structure. However, I thought separate boards of directors would be required (which OpenAI does not have). But I think there's a loophole where the separate boards can be the same people or have significant overlap. The Nov 2023 Altman incident seems to show that's not a good idea. There are separate rules governing different pieces of this by the US government, SEC and IRS. I think a non-profit can have majority ownership in a *publicly traded* for-profit company, and there can be significant (maybe 100%) overlap between the two boards. That seems hard to believe, but it's apparently legal. Maybe that is how OpenAI planned on having a publicly traded, capped-profit subsidiary with a single physical board of directors.
Joe Marler
2025-02-10 23:34:41 +0000 UTCThat was brilliantly made, well done! A list of links to references/sources would also be appreciated as it compressed (really well) a lot of info that I found myself wanting to read up on after it was finished.
Darren Reid
2025-02-10 23:33:19 +0000 UTCEveryone else's comments sum-up my thoughts nicely.
Carleton Torpin
2025-02-10 23:21:11 +0000 UTCThis was excellent! Really great. The only bit that seemed odd to me was the idea that Elon Musk wanted to colonise Mars to escape overpopulation on earth. Perhaps he thought differently back then but these days he seems to be very fixed on the threat of underpopulation. None of this of course affects the idea that Mars wouldn’t be much of a lifeboat if it was easily reached by a rogue superintelligence.
James Maclaurin
2025-02-10 23:01:55 +0000 UTCYou should ping Adam Curtis. He's talked about similar themes - as I'm sure you're aware. (He's also a big sweetie and very approachable despite the prickly facade)
Billy
2025-02-10 22:54:01 +0000 UTCHey Philip! Great work. And loved the in-the-face Matrix references. Keep the mini doc coming, this is a great start. If I may offer more useful and constructive feedback as well, at time the (professional) editing was a bit all over the place, and the story felt a bit floating around / directionless. I was asking myself at time what is the point he is making? Of course it all came together at the end again. Fair to say some people might not have been familiar to what you cover here, so their take might be different. Cheers hope that helps!
Antoine Ferrere
2025-02-10 22:06:35 +0000 UTCThis is a great video, obviously this needs to be published outside of Patreon as well! That said, the one thing I miss is the spark of hope, of positivity -amidst facts and realism- that is key to the Phillip brand.
Erik
2025-02-10 21:45:10 +0000 UTCThis is a video I wish I could gift to several members of my family.
Curt Cox
2025-02-10 21:43:12 +0000 UTCI didn't follow the equity vs no-equity point entirely. If employees had no equity but they were still raising capital from private investors (who would be buying equity) wouldn't openAI still be exposing their organisation to the influence of capitalism? Maybe there's an investment structure that could have retained the original board of the non profit that could have averted that? Feel it wasn't made clear how the employee equity was specifically the issue rather than capitalism more broadly (the need for a return on capital invested). 9/10 otherwise, just need a clearer exposition before I grab my pitchfork.
OG
2025-02-10 21:19:45 +0000 UTCI really do think this is incredible.
Joshua Davis
2025-02-10 21:07:16 +0000 UTCWild ride, and to think it's just getting started... Buckle up...
Oliver Nicholson
2025-02-10 21:03:18 +0000 UTC"AGI - the ultimate general-purpose tool to help us understand the universe" My Proposed Conjecture: "Any pattern that can be generated or found in nature can be efficiently discovered and modelled by a classical learning algorithm." - Demis Hassabis see this article: From Dark to Light: How Understanding the Universe’s Mysteries Transforms Technology - Challenges in 1900 and Today https://www.linkedin.com/pulse/from-dark-light-how-understanding-universes-mysteries-doug-hohulin-d76ic/ "It turns out Einstein, as a young man, read a book by Henri Poincaré, the mathematician and physicist, that in that book, Poincaré put out a series of challenges. Challenges about Brownian motion, challenges about light, challenges about how the world works. It turns out that four or five of those challenges are the ones that Einstein decided to tackle as a 20-something, sitting in Bern, being a third-class patent clerk at the patent office. The breakthroughs that await us will require imagination, bold thinking, persistence and collaboration with AI. They will also demand investment—not just in technology and infrastructure, but in nurturing the next generation of scientists, engineers, and thinkers who dare to challenge the darkness.
Doug
2025-02-10 21:02:16 +0000 UTCGreat video - related to this video see - What if all the world's biggest problems have the same solution? some great history of deep mind https://www.youtube.com/watch?v=P_fHJIYENdI
Doug
2025-02-10 20:54:47 +0000 UTCWatched two thirds and can't wait to finish it. Balanced views while getting into interesting details. Really like it.
John G.
2025-02-10 20:48:35 +0000 UTCVery nicely done! A great summary of the history of AI labs and their changing motivations.
Dylan Southward
2025-02-10 20:48:05 +0000 UTCJust watched while eating lunch ... terrific ! Thank you for the great work as always. Any chance you will eventually put this out for the non-patreon folks?
Daniel A Barbatti
2025-02-10 20:43:42 +0000 UTCThis is really well done -- captures the human drama quite effectively, and is clearly so deeply researched like all of your pieces.
Walker H
2025-02-10 20:41:20 +0000 UTCThat’s awesome
Petr Sektonov
2025-02-10 20:34:30 +0000 UTCGreat video! Learned alot
Jesse van de Kasteele
2025-02-10 20:34:04 +0000 UTCSo very good! The stock footage sometimes really helped the story, but other times it was confusing it up. Especially since the story jumped back and forth through the timeline but the footage did as well, but not always to the same "era". Like where the story is about Greg Brockman talking about people in Deepmind offering their staff more money and the footage is of him presenting ChatGPT on TED. These threw me off (maybe because I just knew so much of that footage already) Anyways good work, trying to share first thoughts.
Arnoldas Kemeklis
2025-02-10 20:32:18 +0000 UTCThis is great! Shows the turbulence but without overblowing it.
William Kentler
2025-02-10 20:23:52 +0000 UTCThis is phenomenal. Please do more minidocs like this. Much appreciated.
Alex Matei
2025-02-10 20:13:39 +0000 UTC"Power tends to corrupt and absolute power corrupts absolutely."
Barnaby Golden
2025-02-10 20:01:46 +0000 UTCAwesome!
Viktor
2025-02-10 19:54:35 +0000 UTCcan confirm
Arnoldas Kemeklis
2025-02-10 19:52:12 +0000 UTCminidoc format? I haven't watched it yet, but I know this'll be fire
zero
2025-02-10 19:34:56 +0000 UTC
