Pinned
Consciousness is crazy man
I'm this little guy in the pilot seat, making all the decisions
7oponaut
5,582 posts
- Replying to @hourly_shitpostIf nobody has 4 kids except Indians then there's no counterexample
- Replying to @TheERDoctorHow does one reintroduce a disease into a vaccinated country?
- Seems like the MCTSr authors did use ground truth information in the MCTS refinement process. They use the LLM for determining the rewards, but the search terminates when the output is equal to the GT. While a similar method could be used as an RL environment to train agents
Replying to @teortaxesTexNothing surprising tbh, I've tried it 2 days ago before noise The thing is that it's literally bruteforcing answers from llm. What if we don't know the ground truth? It will fail miserably at those tasks. - Replying to @bryancskThe stick person would obviously slip to the bottom and then it looks exponential. Follow me for more math tips.
- Months of worldbuilding casually destroyedMira Murati says the AI models that OpenAI have in their labs are not much more advanced than those which are publicly available
00:00 - Replying to @s_streichsbierIt converted a natural language description of the code into python. It could do this because the code already existed in the first place.
- I am told 350°F/175°C is close to the point of discomfort. Only issue: the method is imprecise
- Replying to @karpathyAnother difference is that RLHF doesn't do proper exploration: it mostly learns to exploit a subset of the pretraining trajectories. In contrast, when doing proper RL the discrete action distribution is usually noised by adding an entropy term to the loss function.
