True!

Proof of walk in London!

🤪

😳

😵‍💫

After visiting a Generative AI conference in London last week I genuinely think there is misallocation of funds in big corps. There was a presentation on how to ensure AI agents stick to protocol and don’t hallucinate or come up with creative responses that are unexpected. To solve for this one had to build incredible amounts of test data, heuristic quality gates, non LLM based judge criteria etc. at every step of an agent pipeline. Listening to the presentation, I was convinced their whole pipeline could have been built and run successfully just using those heuristic tests alone, and that sometime inside BigCorp, a few million was earmarked: “How can we do this with AI?” and not “Should we do this with AI?” It felt a bit like listening to a session on how to ensure a genius multi disciplinary artist would excel at accounting.

#Proofofwalk