Oh the chatbots absolutely have the potential to create tons of value. The question is if its creator can harvest it/how much will be left with the consumer vs the producer. I suspect that it will still be very difficult to get people to sign up for subscriptions and advertising will be less effective than current search advertising so the chatbot creator will have difficult harvesting the value they create.

It’s the innovators dillema. Google is selling the answers to questions for millions of dollars per day. If their bot answers with ads, people will use another bot.

why settle for advertisements or user subscription when they could have the Chat AI make transactions for the user and take a cut from the seller. They could publish an API spec for many transaction categories and providers would be in line to sell products and services.

The key is that AI needs to have a fiduciary duty to its users: it should always suggest the most economically advantageous deal for the user, not what is best for the AI if those are not aligned.

Make a marketplace for targeted discounts to incentivize changed behavior.

Many companies that sell consumables are happy to give a first sample free in hopes of repeat purchasing.

If I ask AI for a good lavender scented hand soap refill, it should have a fiduciary duty to give me the best economic choice for me. HOWEVER, AI could solicit discounts from this marketplace platform at the point the AI is making the decision: a new layer above and around Amazon and Walmart and every other established player. Hand soap manufacturers can compete with deals like "people who order hand soap at least twice a year can get 60% off their first purchase" and pay a commission to the AI platform for facilitating the discount. Users can trust the AI with their data because of the fiduciary responsibility and so become willing participants in keeping it accurate because it increases the ability of the AI to serve them.

I disagree with your premise. I do T want the cheapest thing automatically, especially if, as it often is, the difference is crappiness is more than the difference in price.

A fiduciary duty doesn't mean choosing the cheapest thing, it just means the AI can't base decisions off how much of a kickback the AI would receive for making/suggesting that choice.

> The key is that AI needs to have a fiduciary duty to its users: it should always suggest the most economically advantageous deal for the user, not what is best for the AI if those are not aligned.

This generally means you get what the provider can prove seems like the most sensible choice to a regulator, which would mean not being an AI because you can't show how that algorithm works.

Fully automated self-driving puts the brainpower equivalent of a squirrel at the wheel of a 2-ton vehicle.

Automated mapping sends drivers down railroad tracks and impassible abandoned roads.

We'll eventually figure that AI chatbots similarly degrade the social and intellectual commons. Enjoy your AI advisor, the bar and professional boards exist to protect the public, however imperfectly, from exactly the unpredictable charlatanism embodied by your jack-of-all-trades AI assistant.

FYI the assistant is currently working for pennies per second. It costs a few cents of compute (I think including training) to respond to the average query. There are a few orders of magnitude to go.