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.
@redbullfuturist just use sonnet, it’s more than capable for most tasks, and only use opus for really complex tasks or automated code review. you’ll never run out of your limits
@redbullfuturist in my experience: he does the coding for you so you don’t need to manually edit code / run tests. super convenient, too easy to over-rely on. we must chew on problems
@viktaur27 @Teslarati The rate of improvement from original GPT to GPT-3 is impressive. If this rate of improvement continues, GPT-5 or 6 could be indistinguishable from the smartest humans. Just my opinion, not an endorsement. I left OpenAI 2 to 3 years ago. Am a neutral outsider at this point.
first class of 10t models trained on blackwells.
spud will follow. gemini next month. grok a few months after.
you need to prepare yourself psychologically for one obvious fact.
this does not stop.
more compute. more efficiency. more breakthroughs. more intelligence.
now we’ve entered a point, the new models will accelerate all of these accelerants.
the government won’t act in time so we are in the hands of demis, sam, dario, elon, and mark.
may they be kind.
Does Reinforcement Learning Really Incentivize Reasoning Capacity in LLMs Beyond the Base Model?
"we uncover that RL-trained models excel at low k (e.g., pass@1) but are consistently outperformed by base models at high k (e.g., pass@256)."
"RLVR enhances sampling efficiency, not reasoning capacity, while inadvertently shrinking the solution space."
NEWS: New Tesla Model Ys are now autonomously driving themselves about 2 miles from the end of the production line at Giga Texas to their designated lanes near the outbound lot with no humans in the cars.
Thx for the vid @JoeTegtmeyer