a16z Investor Sees Diminishing Returns on Frontier Models
a16z partner discusses slowing AI gains and app builder opportunities
Olivia Moore, a partner at Andreessen Horowitz focused on AI and consumer tech, posted that it feels like many tasks show diminishing returns from newer frontier models. She suggested products may no longer auto-switch to the next frontier model upon release and argued this creates opportunities for app-layer builders to bring down costs. In a reply, Moore added that new frontier models will still gain value by unlocking tasks that we never dreamed of being able to do, such as curing cancer as her speculative example. In a reply to her post, Nick Walker suggested language models should publish task-specific performance scores resembling IQ ratings.
It feels like we've hit diminishing returns on intelligence for many tasks We may no longer see every product auto-switch to the next frontier model upon release This is great for app-layer builders (many opportunities to bring down COGS!)
Reactions from ranked influencers
4 posts…and at the same time, I think new frontier models will also just get more valuable as they start to unlock tasks that we never dreamed of being able to do (curing cancer!)
@omooretweets The LLMs should publish stats like IQ when it completes a task. Like, “help me roast a chicken… IQ 105”.
@omooretweets The LLMs should publish stats like IQ when it completes a task. Like, “help me roast a chicken… IQ 105”.
@omooretweets this has also to do with education - gotta show users why fable is so special / what you can uniquely do with it still v early in that regard
