OpenAI O3 is scoring great on all of the coding and AGI tests. It is saturating many of the tests.
OpenAI O3 seems to have solved a lot of advanced reasoning and math.
OpenAI O3 needed to use about $1 million worth of compute to answer the AGI test questions.
Dr Alan Thompson increased his AGI rating for AI to 88%.



o3 scores a 2727 ELO on Codeforces which places it 175th in the global ranking. That‘s better than ~99.9% of humans on the website (who already tend to be far above average). pic.twitter.com/VGXeQ525nL
— Silas Alberti (@SilasAlberti) December 20, 2024


“This is not merely incremental improvement, but a genuine breakthrough”
François Chollet about the just released OpenAI o3 model. pic.twitter.com/8sHWTZAG9t
— Santiago (@svpino) December 20, 2024

Brian Wang is a Futurist Thought Leader and a popular Science blogger with 1 million readers per month. His blog Nextbigfuture.com is ranked #1 Science News Blog. It covers many disruptive technology and trends including Space, Robotics, Artificial Intelligence, Medicine, Anti-aging Biotechnology, and Nanotechnology.
Known for identifying cutting edge technologies, he is currently a Co-Founder of a startup and fundraiser for high potential early-stage companies. He is the Head of Research for Allocations for deep technology investments and an Angel Investor at Space Angels.
A frequent speaker at corporations, he has been a TEDx speaker, a Singularity University speaker and guest at numerous interviews for radio and podcasts. He is open to public speaking and advising engagements.
1 million dollars of compute.
If that follows Moores law, in 20 years it will cost 1000 dollar.
In terms of server farms—dead malls might be a good locale.
There is a Star Trek TOS Enterprise bridge in one—I could see movie sets in the unused areas—for laser tag—with internet traffic paying the bills.
I want to see AI write code a 0 and a 1 at a time—so you can have software become firmware/hardware.
That might be hack-proof.
By the time AGI wants to take over the world—tight coding won’t allow it.
They are very close to the point where they can build new models using most of the programming workforce made of AI agents.
They need optimize cost, as these new benchmarks requires LOTS of computation. Also they will need to fix some problems. But they are very close to be competent programmers replacement so the project could multiply their codding results.
This is getting real. Nvidia is selling a Jetson system with AI acceleration that can run a LLM locally, for about 35 Watts.
If the cryogenic SOI transistors IBM has announced work out, grid insufficiencies will not be a bottleneck for AI training clusters. I wonder if I’ll have a liquid cooled computer that uses N2 for coolant in a few years. I hope it comes with a Stirling refrigeration unit! Love that heat exchange!
I envision server farms of the largest economic Dewars packed densely with something like cuda cores, a few general purpose processors, power electronics, and gallons of liquid Nitrogen. Nothing but conductors for electrical power, optical fiber, and coolant lines penetrate the Dewars.
Sure, refrigeration costs are high, but if you get 1000 times the ops at the same energy cost, it will be a large saving. High temperature superconductors will first be used in the power electronics, and power distribution conductors. I wonder if they will ever be used as conductors on chip? Not only would that lower power dissipation, but the time constant of circuitry would fall since tau=RC. Superconductors rule!
Forgot this link:
https://www.chaindesk.ai/tools/youtube-summarizer/huge-microchip-breakthrough-zero-heat-and-1000x-more-efficient-l9Ic0PnJl3c#Transistor%20Functionality%20at%20Low%20Temperatures
I always do the adventofcode.com tests each year. Not just to solve the puzzles, but to do them in the most pythony way possible. Two years ago I was using chatgpt to learn the keywords of python, and the nuances of lists and arrays etc. This year I just told it how to solve the problem and it wrote all the code for me, far tighter and neater than I would have expected to myself.