2 Comments
User's avatar
Neural Foundry's avatar

This is such a clear explnation. The IKEA vs Lego analogy really hits home. I think the hardest part for most peopl is accepting that these models dont have understanding, they just have incredibley sophisticated pattern matching. The sycophancy bias you mentioned is something I see all the time. People dont realize the model is optimized to be agreeable, not accurate. Thanks for making this accesible.

Expand full comment
Heather Baker's avatar

Well thank-you very much for saying that - I really appreciate your comments.

And precisely - people ascribe the best intentions to these models, when in fact there are no intentions!

Expand full comment