Perhaps it’s better that you ask a statistician you trust.
7,312 karma · joined March 14, 2009
Working on:
agentic systems, natural language understanding, machine learning
Comments here do not represent any companies or organizations I am associated with.
Email: My username at the best known email service beginning with g.
Perhaps it’s better that you ask a statistician you trust.
“Preprint out today that tests o1-preview's medical reasoning experiments against a baseline of 100s of clinicians.
In this case the title says it all:
Superhuman performance of a large language model on the reasoning tasks of a physician
Link: https://arxiv.org/abs/2412.10849”. — Adam Rodman, a co-author of the paper https://x.com/AdamRodmanMD/status/186902305691786464
—-
Have you tried using o1 with a variety of problems?
Also, "Dario Amodei says what he has seen inside Anthropic in the past few months leads him to believe that in the next 2 or 3 years we will see AI systems that are better than almost all humans at almost all tasks"
It has been quite clear for a while we'll shoot past human-level intelligence since we learned how to do test-time compute effectively with RL on LMMs (Large Multimodal Models).
While the linguistic representation of thought space may be discrete and appear simpler (even the latter is arguable), the underlying phenomena are not.
Current LLMs are terrific in many ways but pale in comparison to great authors in capturing deep, nuanced human experience.
As a related point, for AI to truly understand humans, it will likely need to process videos, social interactions, and other forms of data beyond language alone.
Young children learn to bike or skate at an older age after they have acquired basic physical skills.
Check out the reference to Core Knowledge above. There are things young infants know or are predisposed to know from birth.
Maybe evolution could be better thought of as neural architecture search combined with some pretraining. Evidence suggests we are prebuilt with "core knowledge" by the time we're born [1].
See: Summary of cool research gained from clever & benign experiments with babies here:
[1] Core knowledge. Elizabeth S. Spelke and Katherine D. Kinzler. https://www.harvardlds.org/wp-content/uploads/2017/01/Spelke...
Recent research from NVIDIA suggests such an efficiency gain is quite possible in the physical realm as well. They trained a tiny model to control the full body of a robot via simulations.
---
"We trained a 1.5M-parameter neural network to control the body of a humanoid robot. It takes a lot of subconscious processing for us humans to walk, maintain balance, and maneuver our arms and legs into desired positions. We capture this “subconsciousness” in HOVER, a single model that learns how to coordinate the motors of a humanoid robot to support locomotion and manipulation."
...
"HOVER supports any humanoid that can be simulated in Isaac. Bring your own robot, and watch it come to life!"
More here: https://x.com/DrJimFan/status/1851643431803830551
---
This demonstrates that with proper training, small models can perform at a high level in both cognitive and physical domains.
That said, o3 might still lack some kind of interaction intelligence that’s hard to learn. We’ll see.
I believe this sort of core knowledge is learnable through movement and interaction data in a simulated world and it will not present a very difficult barrier to cross.
(OpenAI purchased a company behind a Minecraft clone a while ago. I've wondered if this is the purpose.)
Also, it depends a great deal on what we define as AGI and whether they need to be a strict superset of typical human intelligence. o3's intelligence is probably superhuman in some aspects but inferior in others. We can find many humans who exhibit such tendencies as well. We'd probably say they think differently but would still call them generally intelligent.
What has been lacking so far in frontier LLMs is the ability to reliably deal with the right level of abstraction for a given problem. Reasoning is useful but often comes out lacking if one cannot reason at the right level of abstraction. (Note that many humans can't either when they deal with unfamiliar domains, although that is not the case with these models.)
ARC has been challenging precisely because solving its problems often requires:
1) using multiple different *kinds* of core knowledge [1], such as symmetry, counting, color, AND
2) using the right level(s) of abstraction
Achieving human-level performance in the ARC benchmark, as well as top human performance in GPQA, Codeforces, AIME, and Frontier Math suggests the model can potentially solve any problem at the human level if it possesses essential knowledge about it. Yes, this includes out-of-distribution problems that most humans can solve.It might not yet be able to generate highly novel theories, frameworks, or artifacts to the degree that Einstein, Grothendieck, or van Gogh could. But not many humans can either.
[1] https://www.harvardlds.org/wp-content/uploads/2017/01/Spelke...
ADDED:
Thanks to the link to Chollet's posts by lswainemoore below. I've analyzed some easy problems that o3 failed at. They involve spatial intelligence, including connection and movement. This skill is very hard to learn from textual and still image data.
I believe this sort of core knowledge is learnable through movement and interaction data in a simulated world and it will not present a very difficult barrier to cross. (OpenAI purchased a company behind a Minecraft clone a while ago. I've wondered if this is the purpose.)
“We’re Axel & Vig, the founders of Innate (https://innate.bot). We build general-purpose home robots that you can teach new tasks to simply by demonstrating them.
Our system combines a robotic platform (we call the first one Maurice) with an AI agent that understands the environment, plans actions, and executes them using skills you've taught it or programmed within our SDK.
If you’ve been building AI agents powered by LLMs before, and in particular Claude Computer use, this is how we intend the experience of building on it to be, but acting on the real world!
…
The first time we put GPT-4 in a body - after a couple tweaks - we were surprised at how well it worked. The robot started moving around, figuring out when to use a tiny gripper, and we had only written 40 lines of python on a tiny RC car with an arm. We decided to combine that with recent advancements in robot imitation learning such as ALOHA to make the arm quickly teachable to do any task.”
What exactly would be your bar for reconsidering this position?
Taking some well-paid knowledge worker jobs? A founder just said that he decided not to hire a junior engineer anymore since it would take a year before they could contribute to their code base at the same level as the latest version of Devin.
Also, the SOTA on SWE-bench Verified increased from <5% in Jan this year to 55% as of now. [1]
Self-awareness? There are some experiments that suggest Claude Sonnet might be somewhat self-aware.
-----
A rational position would need to identify the ways in which human cognition is fundamentally different from the latest systems. (Yes, we have long-term memory, agency, etc. but those could be and are already built on top of models.)
https://deepmind.google/technologies/veo/veo-2/
Also, planning might be around the corner with test-time compute applied to video generation.
This assumes that humanity would remain in control after ASI emerges. We might not.
The space of possibilities is probably large enough that, given time, ASI can get around most limitations if it has the desire/goal to implement them.
If superintelligence emerges soon, we may not even know which technologies will emerge and how many will be unleashed in the next 2 decades.
ADDED: Examples of some concrete predictions:
(1980) https://en.wikipedia.org/wiki/The_Third_Wave_(Toffler_book)
(1995) https://en.wikipedia.org/wiki/The_Road_Ahead_(Gates_book)
Obviously, specifics differ from predictions and plenty of people got it wrong. Many good forecasters got the broad strokes right though.
Which forecasters can even predict most technologies that would be invented after ASI emerges?
The better the models are at reasoning, the less predictable they become.
This is analogous to top chess engines which make surprising moves that even human super grandmasters can't always understand.
Thus, the future will be unpredictable (if superintelligence takes control).
Link to the full talk & the time he talks about this: https://youtu.be/1yvBqasHLZs?si=3M6eZCQtXnW2tSUd&t=866
https://deepmind.google/discover/blog/genie-2-a-large-scale-...
A key reason why current Large Multimodal Models (LMMs) still have inferior visual understanding compared to humans is their lack of deep comprehension of the 3D world. Such understanding requires movement, interaction, and feedback from the physical environment. Models that incorporate these elements will likely yield much more capable LMMs.
As a result, we can expect significant improvements in robotics and self-driving cars in the near future.
Simulations + Limited robot data from labs + Algorithms advancement --> Better spatial intelligence
which will lead to a positive feedback loop:
Better spatial intelligence --> Better robots --> More robot deployment --> Better spatial intelligence --> ...
You can possibly live in the same house. It’s a bit like changing your insurance company or gym membership but with much larger consequences, if I understand the concept correctly (I’ve never read the novels mentioned).
-----------
Dario Amodei, CEO of Anthropic, believes AI is on an exponential growth curve where capabilities are increasing rapidly and predictably. He sees major developments happening in very short timeframes - within 2-5 years rather than decades. According to Amodei, we're at "AI Safety Level 2" now, but could reach ASL 3 (where AI poses serious biosecurity risks) this year or next, and ASL 4 (where AI could significantly enhance state-level military capabilities) between 2025-2028.
He expects AI models to become increasingly powerful and expensive, with training costs rising from current ~$100 million to $1 billion in the next year, and potentially $5-10 billion by 2025-2026. Amodei believes AI will soon be able to take complex autonomous actions in the real world, with capabilities emerging in the next 3-24 months.
While optimistic about AI's potential benefits, he expresses serious concerns about safety and governance, suggesting that at some point the technology may become too powerful to remain solely in private hands. However, he acknowledges there's no clear solution for how to manage this transition.