kuprel

Nod head to change jet's pitch, tilt head to bank the jet. Pinch to fire a missile. Pinch and hold to throttle the aircraft. I've only tried on AVP2, not AVP1

https://artificialanalysis.ai/evaluations/humanitys-last-exa... Not bad

From the image it doesn't look balanced for VTOL when the propellors are vertical. Also are the jets enabled during VTOL?

GPT 5.2 loses at everything but they included that

Why don't they show Grok benchmarks?

Sounds scary haha

A simulation of God's decision tree for the Serenity Prayer: "God, grant me the serenity to accept the things I cannot change, the courage to change the things I can, and the wisdom to know the difference."

This was my first thought. It seems this solar event happened over 1000 years before Younger Dryas

They released a crippled model: https://github.com/SesameAILabs/csm/issues/63

OpenAI doesn’t have voice cloning

I wonder if a language model can be treated as a policy over token-level actions instead of full response actions. Then each response from the language model is a full rollout of the policy. In math and coding, the reward for the response can be evaluated. This is not how DeepSeek works now, right? It treats full responses from the language model as the action if I understand correctly.

This merge request requires approval

I wonder if Cerebras could generate video decent quality in real time

So then 3 in 4 say AI tools do more good than harm?

What if they’re even older than 11.5ky? Neanderthals had larger brains and have been around for half a million years

So maybe the pyramids are older than we thought?

Using int16 and an H100 the speedup is actually 108x over the M2 air

The M2 Air was actually much faster than whatever CPU was on the cloud RTX4090 machine I rented. I chose the stronger benchmark to compare to

This adds PyTorch/CUDA training support to Andrej Karpathy's minbpe. It takes 2min 28sec (148 seconds) on an RTX4090 to train the BasicTokenizer with a vocab_size of 512 on 307MB of Enron emails. The original code takes 2hrs 15min (8076 seconds) on an M2 Air with Python 3.11 to do this. That is a 55x speedup.

chips will have to come a long way for this to be generated in real time, but there's no reason a generated 3D environment can't be interactive

Next they have to add audio, then VideoChatGPT is possible

Why don't they just share the original sample? At least it could offer an existence proof

I didn't realize that the vast majority of the time Earth is in an ice age. What causes it periodically like this?

Was LLaMA2 potentially trained on the raw text from this page? https://huggingface.co/datasets/tatsu-lab/alpaca_eval/viewer...