Bryan, you're onto something important by talking about the AI intermediary layer. When pondering this topic, I think a lot about a specific (perhaps very modest) case of intermediation: setting up appointments with one other person and with groups of people, especially over email. I use Google Calendar and reclaim.ai to help manage my time and calendar. One of the most annoying jobs I do constantly is negotiating appointment times over email. I've tried everything from doodle.com and when2meet.com to reclaim.ai scheduling links—and I daydream about setting up "rbotyee," my own assistant AI bot that would do the negotiating for me (and how rbotyee would negotiate with both humans and other bots). Even in that specific case, despite searching extensively for AI mediation that would solve the problem to, say, a 98% level, I still haven't found what I'm looking for.
My larger point here is that if we're still not that close to a solution for this specific, bounded problem, then full-blown AI intermediation is not going to happen tomorrow. If AI can't reliably handle something as straightforward as "find a meeting time that works for everyone," how realistic are the grander visions of AI intermediaries managing complex negotiations across multiple domains and stakeholders?
My dear Raymond, I have many thoughts in response to your fine comments:
1) (I said this on Facebook) I agree about the scheduling challenge. That *should* be easy but it hasn't been solved yet. This might be the killer app nut to crack.
2) As such, the agentic AI layer might not be what we have tomorrow. Today we're using the basic generative tools to build up interpersonal, intermediary materials, and how we do this gives us clues about what we'll do next.
3) Bigger picture: some of us will prefer the intermediary layer for non-technological reasons. In the US, increasing polarization and separation might make the layer a more attractive, safer option. As the demographic transition works away, we will have to get used to lives without so many people.
Great post, Bryan. It kind of blurs the line between inteligente speculation and prophetic dystopia.
I think that line is where I live these days.
Bryan, you're onto something important by talking about the AI intermediary layer. When pondering this topic, I think a lot about a specific (perhaps very modest) case of intermediation: setting up appointments with one other person and with groups of people, especially over email. I use Google Calendar and reclaim.ai to help manage my time and calendar. One of the most annoying jobs I do constantly is negotiating appointment times over email. I've tried everything from doodle.com and when2meet.com to reclaim.ai scheduling links—and I daydream about setting up "rbotyee," my own assistant AI bot that would do the negotiating for me (and how rbotyee would negotiate with both humans and other bots). Even in that specific case, despite searching extensively for AI mediation that would solve the problem to, say, a 98% level, I still haven't found what I'm looking for.
My larger point here is that if we're still not that close to a solution for this specific, bounded problem, then full-blown AI intermediation is not going to happen tomorrow. If AI can't reliably handle something as straightforward as "find a meeting time that works for everyone," how realistic are the grander visions of AI intermediaries managing complex negotiations across multiple domains and stakeholders?
[This response was written with help of machine intelligence, claude.ai, specifically: https://claude.ai/share/851179e0-a7df-462c-af5d-77bf749f96e1]
My dear Raymond, I have many thoughts in response to your fine comments:
1) (I said this on Facebook) I agree about the scheduling challenge. That *should* be easy but it hasn't been solved yet. This might be the killer app nut to crack.
2) As such, the agentic AI layer might not be what we have tomorrow. Today we're using the basic generative tools to build up interpersonal, intermediary materials, and how we do this gives us clues about what we'll do next.
3) Bigger picture: some of us will prefer the intermediary layer for non-technological reasons. In the US, increasing polarization and separation might make the layer a more attractive, safer option. As the demographic transition works away, we will have to get used to lives without so many people.
4) I'm grateful for your citing AI at the end.