Diogo Almeida

Diogo Almeida is a co-inventor of the techniques underlying chatgpt and a key figure in the evolution of Large Language Models (LLMs). He is currently associated with rlcd (Reinforcement Learning from Calibrated Decisions), a framework proposing a shift away from traditional human-preferred text generation toward calibrated decision-making.

Key Contributions & Concepts

• Co-inventor of ChatGPT’s Core: Played a pivotal role in developing the foundational techniques behind ChatGPT chatgpt.
• Advocate for RLCD: Proposes that the industry is shifting from “human-preferred text” to “calibrated decisions” via the RLCD methodology rlcd.
• Critique of Current LLM Paradigms: Argues that the core idea of ChatGPT is being superseded by new training approaches that prioritize decision calibration over text preference rlcd.
• Founder of TypeSafe.ai: Established TypeSafe.ai to develop Jev, an AI model explicitly designed for reliable decision-making with zero hallucinations, contrasting with traditional LLM text generation.
• Development of Jev: Created Jev to address the limitations of current LLMs by focusing on fast, reliable, and hallucination-free decision outputs Jev: TypeSafe.ai’s Fast, Reliable Decision AI with Zero Hallucinations.

References

• Jev: TypeSafe.ai’s Fast, Reliable Decision AI with Zero Hallucinations