I previously wrote about Ben Thompson's Aggregation Theory in Natural Orders (and excerpted in the above Walled Gardens article).Β
This has recently been updated with an addendum to project the impact
of LLMs:
"Imagine if Google had an entire collection of system prompts that mapped onto the Topics API (transparently posted, of course): the best prompt for the user would be selected based on what the user has already showed an interest in (along with other factors like where they are located, preferences, etc.).
This would transform the AI from being a sole source of truth dictating supply to the user, to one that gives the user what they want β which is exactly how Aggregators achieve market power in the first place.
This solution would not be βperfectβ, in that it would have the same problems that we have today: some number of people would have the βwrongβ beliefs or preferences, and personalized AI may do an even better job of giving them what they want to see than todayβs algorithms do.
That, though, is the human condition, where the pursuit of βperfectionβ inevitably ends in ruin; more prosaically, these are companies that not only seek to serve the entire world, but have cost structures predicated on doing exactly that.
That, by extension, means it remains imperative for Google and the other Aggregators to move on from employees who see them as political projects, not product companies.
AIs have little minds in a big world, and the only possible answer is to let every user get their own word.
The political era of the Internet may not be inevitable β at least in terms of Aggregators and their business models β but only if Google et al will go back to putting good products and Aggregator economics first, and leave the politics for us humans."
Aggregatorβs AI Risk >>
|