After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI? I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev • 20-200x faster • 40-400xShow more
After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI? I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev • 20-200x faster • 40-400xShow more
Heaviside-1 predicts the EM response in a few hundred milliseconds and grades the response. Jev met the full spec in 5 decisions, averaging ~230 ms per decision. Astra, fable and opus did not solve this within 10 decisions.
Jev was 53x faster than opus, 20x faster than astra, but still was the only model to solve the physics. Jev isn’t acc doing the physics, to be clear here. Atlas Fields is running Heaviside-1 for that, our team has a whole writeup on that here:arenaphysica.com/publications/h…. WithoutShow more