Diogo Almeida, a former OpenAI researcher and co-inventor of ChatGPT and RLHF, has launched a new AI model called Jev through his company TypeSafe AI. This model is designed as a "System One Model" optimized for automation, diverging from traditional LLMs by not generating text. Instead, Jev takes world state and typed questions to provide calibrated decisions with probabilities, aiming for lower latency and cost. The training method is Reinforcement Learning for Calibrated Decisions (RLCD), which focuses on epistemically honest probabilities, and it utilizes a parallel sampler to generate all outputs in a single query. AI
IMPACT This model's focus on calibrated decisions and automation could shift how developers integrate AI into applications, reducing the need for complex parsing and validation layers.
RANK_REASON New model release from a company founded by a claimed co-inventor of ChatGPT and RLHF. [lever_c_demoted from frontier_release: ic=1 ai=1.0]
- ChatGPT
- Daniel Kahneman
- Diogo Almeida
- OpenAI
- RLHF
- RLCD
- System One Model
- The Register
- TypeSafe AI
- Vercel
- X
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →