OpenAI: ‘We Just Reached Human-level Reasoning’.

Surely, it is worth taking a few moments to reflect on the claim from OpenAI that we’ve reached ‘human-level reasoning’ in their o1 series of AI models? I scrutinize Altman’s dev day comments (picking out 4 highlights) and cover the most recent papers and analysis of o1’s capabilities. Then, what the colossal new valuation means for OpenAI and the context you might not has realized. We’ll look further down the ‘Levels of AGI’ chart, cover a NotebookLM update, and end with a powerful question over whether we should, ultimately, be aiming to automate OpenAI.

Assembly AI Speech to Text:

AI Insiders:

Chapters:
00:00 – Introduction
00:52 – Human-level Problem Solvers?
03:22 – Very Steep Progress + Huge Gap Coming
04:23 – Scientists React
05:44 – SciCode
06:55 – Benchmarks Harder to Make + Mensa
07:30 – Agents
08:36 – For-profit and Funding Blocker
09:45 – AGI Clause + Microsoft Definition
11:23 – Gates Shift
12:43 – NotebookLM Update + Assembly
14:11 – Automating OpenAI

Reuters Funding-block Exclusive:
OpenAI Scaling AGI:
NYT Revenue Story:
For Profit Move:
Bloomberg Levels Chart:
Scientists React, in Nature:
Math prof:
AGI Clause:
Microsoft Sci-fi:
Mensa Tweet:
Sci-Code:

FT Agentic Systems 2025:

Bill Gates Turnaround:
OpenAI Preparedness Framework:

NotebookLM:

My Coursera Course – The 8 Most Controversial Terms in AI:

Non-hype Newsletter:

I use Descript to edit my videos (no pauses or filler words!):

Many people expense AI Insiders for work. Feel free to use the Template in the 'About Section' of my Patreon:

Joe Lilli
 

  • @jfitnesshealth says:

    You think thats AIR your breathing?? Hmmmm…..AGAIN! 😅😅

  • @Andytlp says:

    I’ll believe it when i see it. The last 6-8 months we’re pretty uneventful. Ai generated videos are getting way better. So is audio. Language models have stalled. At least whats available or known publicly.

  • @scrollop says:

    I am sick of Altman’s hype hype hypinh – and it seem to increase funding and an eventual stock price. He is noise. Wait for the models.

  • @JasoTheRed48F2 says:

    Remember folks, OpenAI is a for-profit company now. Take everything they say with that in mind.

    • @tom9380 says:

      They been acting the hype game quite before.

    • @itsdakideli755 says:

      Why do people always say this? You’ve been repeating the same stuff for months now. Every time, OpenAI raises the bar, and it’s getting old seeing the same tired response.

    • @ChrisBa303 says:

      That’s irrelevant to what he meant. They are now required to generate profit at some point and will generate hype to keep their investors happy. Otherwise top management gets axed

    • @teekanne15 says:

      Doesn’t really matter if for profit or not. Mainly a talking point by competitors that fear missing out on the biggest technical power shift since facebook.

  • @nlcwilson2017 says:

    I know some humans that don’t have human-level reasoning

  • @SamWilkinsonn says:

    I’m sceptical. I’m always impressed at what LLMs can do but still.

    These days I can’t help but notice the stark similarities between Altman and Musk and that obliterates any trust in him.

    I’m open to being wrong though and I very well could be. I guess I won’t have to wait long to find out, unless it’s one of Musk’s ‘next year’ promises lol.

  • @kieranhosty says:

    I’ll take a dump truck of salt with that statement. Open ai depends on the AI hype, and it’s creating.

  • @andybaldman says:

    Altman is playing the CEO game.
    And that only involves keeping a company going.
    It doesn’t have to involve truth.

  • @tom9380 says:

    Let’s see what their ARC-AGI score (or other private evals) is going to be, before breaching out in another hype cycle.

  • @julianBTC21 says:

    It’s getting hard to distinguish the “baaaah” sounds from the English the way this guy is goated when it comes to AI news 🐐

  • @thereal_JMT_ says:

    ok so it was actually O1 preview which secured the newest funding round and not Sam Altman? Why not get rid of him and let Chat GPT run the show?

  • @SamWilkinsonn says:

    Microsoft seems to have got confused between AGI and ASI. Also it’s rich coming from Gates saying ‘can we trust governments with [setting the sociological rules of the AI]’

  • @MrJenius says:

    With news that Sam Altman will get equity, paired with the things he has said previously about him NOT having equity, I have lost all trust. Very concerned if OpenAI if the first to get AGI as I could see them putting monopolistic practices into place. It feels like nothing they say can be trusted at this point.

  • @DaveShap says:

    Hold on. Philip saying “My rough around the edge’s London accent” which is perhaps the clearest BBC diction ever and only one step shy of pure Shakespearian Received Pronunciation…. this is a new level of English modesty!

  • @kylewollman2239 says:

    Progress hasn’t stalled. Here’s our new Plateau 1 model.

  • @victorgold571 says:

    Wouldn’t most of the current issues LLMs face be solved by integrating their architecture to that of liquid neural networks?

  • @DreckbobBratpfanne says:

    Interesting to see the SciCode Benchmark improvements of 4 to o1. A similar step by next year would be >>50% score. Also Sonnets secret sauce is something to behold.
    I wonder if a vision o1 could solve Classic Tetris, current systems fail miserably (quite a fun benchmark in its own right)

  • @reinerheiner1148 says:

    An AI with human level reasoning should have no problems with the abilities defined in levels 3-5, because it should be able to do all of these with reasoning and agent frameworks. Its a really strange definition.

  • @RoyMagnuson says:

    As a college professor, the existential crisis is: if students add class materials to notebookLM (which we know they are/will)…that is not their IP.

  • @faolitaruna says:

    11:08 My prediction: robotic cat-girl cafes before terraforming of exoplanets.

  • >