open source will win, in the end. in the meantime, labs should focus on pushing the frontier as far as possible. packaging and distributing ai to the world for the sake of maximizing access is a secondary, while important, goal. in the unlikely case where the gap between the public and private model grows over time, instead of shrinks, as it is now, responsible parties should publish research to close the gap.

Some reflection on what today's reasoning launch really means:

New Paradigm
I really hope people understand that this is a new paradigm: don't expect the same pace, schedule, or dynamics of pre-training era.
I believe the rate of improvement on evals with our reasoning models has been the fastest in OpenAI history.
It's going to be a wild year.

Generalization across Domain
o1 isn't just a strong math, coding, problem solving, etc. model but also the best model I've ever used for answering nuanced questions, teaching me new things, giving medical advice, or solving esoteric problems.
This shouldn't be taken for granted!

Safety by Reasoning
The fact that our reasoning models also improve on safety behavior and safety reasoning is very much non-trivial.
For years (a decade?) the boogeyman of the AI world was reinforcement learning agents which were incredibly adept at game playing but completely incapable of reasoning or understanding human values!
This is a strong point of evidence against this.

Scaling inference-time compute can compete with scaling training compute!
The fact that o1-mini is better than o1 on some evals is very very remarkable. The implications of this I'll leave as an exercise for the reader.

Multimodal Reasoning
It's kind of crazy that reasoning improves on multimodal evals as well! See MMMU and MathVista: these aren't small improvements.


To be clear I'm not one of the contributors to the o1 project: this has been the absolutely incredible work of the reasoning & related teams.
The rate of progress has just been faster than anything I've ever seen: it's absurd how fast the team has climbed the scaling OOMs just after discovering this paradigm.

i think people are misunderstanding gpt-4o. it isn't a text model with a voice or image attachment. it's a natively multimodal token in, multimodal token out model.
you want it to talk fast? just prompt it to. need to translate into whale noises? just use few shot examples. every trick in the book that you've been using for text also works for audio in, audio out, image perception, video perception, and image generation.

announcing... starlinkmap dot org
real-time map of every starlink satellite. tracks upcoming launches, other constellations, orbital updates, etc.
finally launching this after a while! more details below. starlink is, imo, one of the most exciting technologies of our generation.
today, only 65% of the world has access to the internet at all (and far fewer have high-speed internet).
with direct-to-cell coming, soon every device, anywhere on Earth, will be connected together.

I ask DALLE-3 to generate a Pepe but each time I tell it to make it "more rare." "make it more rare"

DALLE-3 is the best product I've seen since GPT-4, super easy to just get sucked in for hours generating images. No need for prompting since GPT-4 does it for you.
Let me know if you have requests for prompts below. Here are some examples of what it can do:





It's shockingly good at styles that require consistent patterning like Pixel Art, mosaics, or dot matrices.

FIGMA-OS: The first Turing-complete Figma file.
SPECS: 8-bit architecture, 512 bits of RAM, 16 bytes of Program Memory, MISC instruction set of 16 OPCODES, 10HZ clock speed, 4 fast access registers, binary-tree RAM/ROM memory.
MOTIVATIONS: For the meme.
HOW: Explained below. FIGMA-OS has every feature that any modern, enterprising technologist could possibly need:
► A stunning and detailed user manual.
► Useful pre-installed programs like: Fibonacci Numbers.
► An award-winning graphical user interface.

do you have any hobbies?
yeah making computers out of things that shouldn't be computers. watch me be the first to bring turing completeness to figma
(edit going to build this tonight so scroll for my live tweeting of a computer) https://t.co/a07l9Ib0Qntwitter.com/i/web/status/1…

https://twitter.com/figma/status/1671563763789725698

ok simple clock working seems promising. add/sub/mult/div already implemented for numbers already by figma, seems like there might be more ops for other types which is great

I'm excited to announce Alexandria: Project Tenet, a community project to embed all human belief.

Today, we're open-sourcing the embeddings for 10+ major religious texts (over 15m tokens, 20 billion vector dims) and launching multiple live products with the data.

Let's dive in. Last time, we released the embeddings for the Arxiv (600m+ tokens, 3.07B vector dims). We did this because we saw immediate usefulness in helping improve & accelerate research efforts; and we were right!

We see the same with working on human belief.

https://twitter.com/willdepue/status/1661781355452325889?s=20

Today, I'm announcing Alexandria, an open-source initiative to embed the internet.

To start, we're releasing the embeddings for every research paper on the Arxiv. That's over 4m items, 600m tokens, and 3.07 billion vector dimensions.

We're not stopping here. twitter.com/i/web/status/1… A significant number of the world's problems are just search, clustering, recommendation, or classification; all things embeddings are great at.

For example, finding research papers via keywords is hard when there's 10 words that mean the same thing. Embeddings makes this easy.

⚡️Introducing WebGPT⚡️

Just this month, Chrome announced WebGPU's release. What does this mean? Near-native GPU speeds, from the web!

I took the opportunity to build WebGPT: a package to run GPT models entirely on the browser.

Here's why this is a big deal: twitter.com/i/web/status/1… WebGPT is just four Javascript files. No installations, no libraries, just ~1000 lines of vanilla JS!

Soon, every device in the world can run models... just by clicking a link. No friction, amazing interoperability.

But why does this matter?

I am extremely worried about AI pornography and chat bots.
Birth rates are already bad: undeniably correlated to pornography, unrealistic expectations, and introversion.
This gets much worse: AI generates custom content of incredibly attractive people + you can talk to them.
Cont In a sci-fi manner, I worry if bots can’t engineer faces that are vastly more attractive than any human.
But realistically, they already can produce images of people in the top 0.1% in beauty, tailored to anyone’s preferences.
Below image is AI generated. How are we this far.

One of the greatest one-page guides I've read this year. Anyone got other similar guides, in same/different subjects?

https://twitter.com/algekalipso/status/1609588936912900096

chessengines.org