Adam Farence

Warm take: The reason why AI hype is slowing isn't about LLMs, it's about humans. 2 main reasons why: 1) Expectations. When GPT-4 came out, it literally blew everyone away. The "demos" of its capabilities were everywhere - every single person I knew was amazed. We were promised the future. But as more people began using these systems, they started seeing the flaws - the hallucinations and limitations. They were disappointed, so they left. Their high expectations weren't met. 2) The blank canvas problem Every walk into a restaurant that has 100 items on the menu? Impossible to pick something right. Inversely if they have like 5 things on the menu, and 1 is labeled "the most popular" you'll likely just get that. Same thing applies to AI. The fact they can do so much that actually makes it hard for everyday people to figure out how to really use them. After trying the basic stuff like "write me paper" or "draft me an email" most can't think of anything to ask anymore. These systems are insanely powerful, but it needs a power user. They need an intent-driven power user to fully utilize them Combine these two factors and you get a drop in users and hype This is actually a product problem - chat interfaces are likely not the answer until we have low latency voice -> text. We need better UIs and ways to teach capabilities. Once everyone has their "Aha!" moment (beyond basic examples), we'll start to see the hyper growth of AI.