OpenAI o1 — our first model trained with reinforcement learning to think hard about problems before answering. Extremely proud of the team! This is a new paradigm with vast opportunity. This is evident quantitatively (eg reasoning metrics are already a step function improved)
RESEARCH
-
Reasoning tokens: tracking, feedback, and model consistency challenges
By
–
Having another token type would add complexity. Do you track and store reasoning tokens? If the initial reasoning is off, and you provide it feedback, how well will it guide the next reasoning, especially if the model drops those tokens (or so it seems that way based on how I’m
-
OpenAI Unveils ChatGPT Reasoning Feature
By
–
OpenAI Unveils New ChatGPT feature That It Claims Can Reason #AI #AGI #Strawberry
-
Processing Versus Thinking: Semantic Distinctions in AI Systems
By
–
Take with a grain of salt because English is not my first language but my perception is that "processing" can be an action of a program while "thinking" is mostly attributed to humans (would be weird to say that Google is "thinking" when you run a search query)
-
Powerful and economical reasoning model, OpenAI IDE speculation
By
–
In short: They have released a specialized model for reasoning tasks (math, code etc.) with crazy performance on these tasks, and especially 80% cheaper and faster. My little finger tells me that OpenAI will soon launch its own IDE… like cursor 😉 ……..
-
Transformative AI Closer Than Expected Says Industry Leader
By
–
This is remarkable. Transformative AI is closer than we thought just a few years ago.
-
OpenAI Project Strawberry: New AI Step-by-Step Reasoning Approach
By
–
OpenAI has revealed details of Project Strawberry, what it claims is a more general approach to teaching to AI solve problems with step-by-step reasoning. Outside experts say it is an important new direction for AI but note that challenges remain:
-
Partner with IWF upfront to prevent illegal content in AI datasets
By
–
Partner with IWF upfront, not when problem arise! LAION knew about the risks of illegal content (e.g. child abuse) being in their dataset. It would have been sensible to partner with professional organizations like IWF to check files *before* publishing the dataset.
-
Fair Use Licensing for AI Datasets: Legal Compliance Strategy
By
–
Use an appropriate license compatible with Fair Use. LAION's creation of the dataset required one or more Copyright Exceptions with strict terms. Picking a license that respects the terms of these exceptions would reduce legal risk significantly, and show good faith.
-
User Generated Content Exclusion from AI Training Datasets
By
–
Exclude social media and UGC websites. Popular websites that host user generated content or personal profiles could have been excluded from the dataset. Platforms don't have full rights to their users private information, and avoiding that kind of content reduces legal risk.