Herbie Bradley argues that expectations of widespread engagement loops with GPT-4o-style models are overstated within AI circles and predicts only a small fraction of the general public will sustain them
Séb Krier agreed while Roon discussed future specialized agents.
@herbiebradley @sebkrier do u think they will remain text only interfaces for long? we will need sales agents and sell-side investing agents and much much more
Herbie Bradley@herbiebradley@tszzl @sebkrier I think this idea has some probability but is strongly overrated in the AI community, because this community is highly selected for people who like to talk to AI I would bet only a small fraction of the public ever fall down a 4o-style engagement loop
@herbiebradley @tszzl +1
Herbie Bradley@herbiebradley@tszzl @sebkrier I think this idea has some probability but is strongly overrated in the AI community, because this community is highly selected for people who like to talk to AI I would bet only a small fraction of the public ever fall down a 4o-style engagement loop
IMO a key driving factor is that however much people might like scrolling tiktok, they have on avg a deep preference to talk to humans because they are humans
empirically I could also say that text is a very constrained interface relative to multimodal avatars, and it seems to have been very hard so far to make these avatars convincing, perhaps for similar reasons to why it's hard to get good quality writing out of the models.
I think eg enterprise sales will be handshakes by humans even post-AGI in many cases, while sell-side investing seems more like something that will diminish in favor of more text-based trading agents
roon@tszzl@herbiebradley @sebkrier do u think they will remain text only interfaces for long? we will need sales agents and sell-side investing agents and much much more