Viewing profile — p1esk

If Siri is using a 3T model in high reasoning mode to answer your question you will.

I stopped writing code completely about a year ago, stopped reading code completely about 8 months ago, and I feel like I stopped thinking hard about anything at work about 6 month…

Is there any degradation with INT8 weights quantization? Why would anyone want to apply ConvRot to do 8 bit weights? Note the paper [1] focuses on 4 bit weights and 4 bit activatio…

Zero. These were open problems.

This blog post

It’s still at GPT-1 level, but GPT-2 moment feels imminent.

replace they key-query-value mechanic by just dropping it while making the entire context the latent space. What do you mean by this? Like concatenating all token embeddings into o…

I wonder how difficult it would be to convert these to 3D

I’m guessing other models would probably stop in this situation and ask the user for instructions.

I think they meant that Opus 5 had to find and set up a vision model to process the image

[flagged]

[flagged]

It’s nice to be rich I guess

where do you see "twice the memory bandwidth"?

There’s noticeable accuracy degradation when they switched from fp8 to mxfp4

That’s how you get skills

Do a decentralized p2p one

We’re experiencing gpt-2 moment in robotics now. This means in about 2-3 years they will do useful work (cooking, repairs, cleaning, etc).

It’s got to be similar to Fable, which I experienced for 3 days, and which impressed me (compared to Opus 3.8)

VLAs are new LLMs. Give them 5 years to develop. But even good old LLMs are still improving every six months.

Unfortunately if your manager thinks something is your problem, it becomes your problem.

You realize llms as a field is barely 5 years old? Give it at least another 5.

Many people here spent a lot more than $300 on headphones long before AirPods appeared.

100×10^15 km

Do you feel that recent advances in AI can speed up such rare disease research?