4 ms·

Sometimes unreliability can be worked around with human supervision. You wouldn't normally want to write a script that just chooses the first Google result, but

Sometimes unreliability can be worked around with human supervision. You wouldn't normally want to write a script that just chooses the first Google result, but that doesn't mean search engines aren't useful. The goal when improving a search engine is to put some good choices on the first page, not perfection, which isn't even well-defined. The AI generators work similarly. They're like slot machines, literally built on random number generators. If you don't like the result, try again. When you get something you like, you keep it. There are diminishing returns to re-running the same query once you got a good result, because most results are likely to be worse than what you have. Randomness can make games more fun at first. But I wonder how much of a grind it will be once the novelty wears off?

Karrot_Kream 4y ago
I think that will be the hard part of productionizing these models. Some domains absolutely need high reliability, like AVs, and applying these ML models is fraught because of how imprecise their outputs are. Problem domains where unreliable results are useful either through human mitigation (e.g. expert search engines, customer service chatbots) or because they're just not that reliable anyway (like image generation) will be the domains where these models will have sticking power.
christkv 4y ago
The fact that they produce valuable content is more a reflection of the mediocre output of many people. I hold it up against the response I would expect from a human and in a lot of cases I’m not sure the human wins.
- pixl97 4y ago
  Yea, I see a lot of "AI models need to be perfect", but man, that is really neglecting that humans on average suck at a great many things.
- skybrian 4y ago
  Often when you have a question, you don't have access to a qualified expert you could ask for help. So there's less competition than you might think.

Karrot_Kream 4y ago

I think that will be the hard part of productionizing these models. Some domains absolutely need high reliability, like AVs, and applying these ML models is fraught because of how imprecise their outputs are. Problem domains where unreliable results are useful either through human mitigation (e.g. expert search engines, customer service chatbots) or because they're just not that reliable anyway (like image generation) will be the domains where these models will have sticking power.

christkv 4y ago

The fact that they produce valuable content is more a reflection of the mediocre output of many people. I hold it up against the response I would expect from a human and in a lot of cases I’m not sure the human wins.

pixl97 4y ago
Yea, I see a lot of "AI models need to be perfect", but man, that is really neglecting that humans on average suck at a great many things.
skybrian 4y ago
Often when you have a question, you don't have access to a qualified expert you could ask for help. So there's less competition than you might think.