HACKER Q&A
📣 tmaly

State of intelligent agents for content curation?


I was just reading another post on HN about William Gibson's latest novel. It just dawned on me that I did not even realize it was released in January.

I experience this same idea with a few tv series that I follow. I never quite keep track of when a new season is released.

Years ago, there was this idea of intelligent agents that would track and curate content for you.

Is there any great system that does this now?


  👤 PaulHoule Accepted Answer ✓
It is not a problem of technology, it is a problem of business models.

So long as it is a "cloud based" service that does "collaborative filtering" is it is going to be the same "almost works" system that we have that is really about the needs of advertisers. Make it "free" and it is guaranteed to be crap.

If it is a service where you maintain your own database, make your own judgements, and feed them back into the system that's a different matter. Trouble is it takes thousands of judgements to make something like that good, most people seem to get bored around judgement #5. Also if it isn't free you will get lots of messages from people who will complain that you're asking for money.

I've built two systems that do this, one of which filtered hacker news, one of which filtered job listings (some of which came from hacker news.)

The main issue is that ELMO and BERT and other shiny stuff that people talk about will make a minor improvement over a decent bag-of-words classifier; what you really need is thousands of judgements to teach the system what your opinion is.