Machine Learning Street Talk (MLST) Aiden Gomez - CEO of Cohere (AI's 'Inner Monologue' – Crucial for Reasoning)
40 snips
Jun 29, 2024 
Aidan Gomez, CEO of Cohere, dives into tackling AI hallucinations and reasoning improvements in this lively conversation. He discusses why Cohere avoids using GPT-4 for training and shares insights on the unique challenges enterprises face with AI, from onboarding to legislation. Aidan highlights their commitment to crafting robust, industry-specific solutions while addressing the societal implications of AI advancements and regulatory needs. Get a glimpse into Cohere’s ethos and their strategic vision for the future of AI technology.
AI Snips
Chapters
Transcript
Episode notes
Model Differentiation
- Many LLMs sound similar because they are trained on similar data, including other LLM outputs, leading to a homogenization of their personalities.
- Cohere differentiates its models by excluding outputs from other models like GPT-4 during training, resulting in a more unique "feel."
Reasoning in LLMs
- Aidan Gomez believes reasoning is crucial for intelligence, and LLMs can reason.
- LLMs are becoming increasingly robust reasoners, moving from weak reasoning to more sophisticated planning and task breakdown.
Improving LLM Reasoning
- To improve LLM reasoning, expose them to data that demonstrates the "inner monologue" of problem-solving.
- Create synthetic datasets with step-by-step reasoning processes, mistakes, corrections, and retries, as this data is scarce online.