/MachineLearning

[Julia Turc @juliarturc]

This is totally happening.

[/MachineLearning @slashML]

@polynoamial @OpenAI So nothing has been meaningfully happened internally that would indicate progress?

[Can Bölük @can1357]

guys you do know you can just disable thinking, and instead give it a "deep_think" tool, and it will call it with internal CoT reasoning format right?

gl fixing that

[/MachineLearning @slashML]

@kellerjordan0 What do you expect if you're only interested in "make number go up" papers?

[/MachineLearning @slashML]

@tszzl @brij So you're saying the new set of signatories will be the new generation of losers of the AI race?

[/MachineLearning @slashML]

@BlancheMinerva @scaling01 I do not see how restriction of access was ever an option when it's obvious that more than one country can produce this technology.

[/MachineLearning @slashML]

SSSI - Straight Shot to Super Intelligence

[SSI Inc. @ssi]

We are announcing a long-term strategic partnership with NVIDIA. NVIDIA is making a substantial investment in SSI that will let us 10x our compute in the next 12 months. We reached the point where our research is worth scaling and with this partnership we will be able to. We are honored by NVIDIA’s conviction.

[/MachineLearning @slashML]

@scaling01 In what universe do people think artificial intelligence will not proliferate?

[/MachineLearning @slashML]

@suchenzang Should poisoning output tokens (human or AI authored) be illegal? 🤔

[levent @__alpoge__]

hello there the jacobian conjecture is false thanx to my close friend akhil for asking about it and my other close friend fable for working during the world cup final

((1+xy)^3 z + y^2 (1+xy) (4+3xy), y + 3 x (1+xy)^2 z + 3 x y^2 (4+3xy), 2 x - 3 x^2 y - x^3 z): \C^3\to \C^3, has jacobian determinant -2, and sends (0, 0, -1/4), (1, -3/2, 13/2), and (-1, 3/2, 13/2) to (-1/4, 0, 0)

[/MachineLearning @slashML]

@boazbaraktcs The issue is the community has provided ample feedback. Have you read his responses? Mostly attacks and insults.

[/MachineLearning @slashML]

Self-serving statement from OpenAI employee about open source: Open source won't concentrate enough power into a handful frontier labs to build out the next model.

[/MachineLearning @slashML]

Crazy that there are people in this world that think only Anthropic and OpenAI can produce state of the art models.

[Jukan @jukan05]

All of these debates ultimately boil down to one point.

If the money flowing into OpenAI and Anthropic dries up, this entire investment cycle will eventually grind to a halt.

If OpenAI and Anthropic can no longer produce frontier, state-of-the-art models, the Chinese open-source models that distill them will stagnate as well. Do you see, then, how naive it is to claim that hyperscalers can simply take Chinese open-source LLMs and resell access to them through APIs?

We have already boarded a train we cannot get off—one that leaves us no choice but to keep pouring money into OpenAI and Anthropic as it races forward.

Ultimately, someone has to buy vast numbers of GPUs and build AI the expensive, inefficient way—not the Chinese way.

[/MachineLearning @slashML]

@BlackHC @stanislavfort Why are people overthinking this? Distilling without logits is just as efficient as just training on human annotated high-quality traces. Clearly this works...

[Zephyr @zephyr_z9]

Crazy big boi numbers from Kimi K3

[/MachineLearning @slashML]

@bakkermichiel Is it possible that the frontier is trying to produce the most economical model and not the most capable?

[Hongyin Luo @lhyTHU]

J-Space introduced by Anthropic is pretty cool. Our paper in ACL 2019 (https://t.co/GvYQbMDB0F), however, showed that the hidden states of intermediate layers in language models contain the information of not only the next token, but the following sub-sequence based on semantic dependency.

[/MachineLearning @slashML]

@kanjun Open source incentivizes standards which commoditizes hardware.

[Xin Chen, Cynthia @XinCynthiaChen]

Excited that our ICML position paper has been accepted as an Oral 🎉!

When a model looks like it "deceives" or "resists shutdown," how do we know it isn't role-play, instruction-following, or just task-completion pressure? Often the current evidence can't yet tell them apart, and those claims are starting to inform deployment and regulation.

We map where the evidence gets thin across four stages and propose a shared standard to strengthen it. Here's the map.

🧵
https://t.co/gzM8mLGB5P