Hacker News Reader

Comment by madaxe_again | Hacker News Reader

The current wave of AI stuff is likely not indicative of how AI will be used in consumer-facing businesses in the future.

At the moment, it’s there for all to see, be it as a chatbot or whatever - but the real applications will be under the hood, not directly interacting with the consumer, doing everything from market microsegmentation to tailored recommendations to diagnostics, credit scoring, criminal profiling, fraud detection, and all of the things that currently rely on sub-optimal human-crafted algorithms or even hand-cranking.

So all in all, what consumers think of AI is almost neither here nor there, as they are not the primary market - business and government are.

View on HN

> The current wave of AI stuff is likely not indicative of how AI will be used in consumer-facing businesses in the future.

wut. the first real deployments we've seen of LLMs are chatbots, esp. those aimed at consumers. hence all of the "sell me a Chevy for $1" memes.

> At the moment, it’s there for all to see, be it as a chatbot or whatever - but the real applications will be under the hood, not directly interacting with the consumer, doing everything from market microsegmentation to tailored recommendations to diagnostics, credit scoring, criminal profiling, fraud detection, and all of the things that currently rely on sub-optimal human-crafted algorithms or even hand-cranking.

That's already a thing, and is what FB and GOOG (or banks like Ant Financial, etc.) have been making money off of for years.

In other words, its job will be perpetual-bias-perpetuator! I cannot fucking wait.

Could be. One of the annoying things is that when the AI has a bias people agree with, they say "algorithms can't be biased!", yet when the AI has one they do disagree with "it's so unfair they've made it woke/racist [delete as appropriate]".

Reality is difficult, AI is difficult, statistics is difficult, everything is difficult. Accidental bias is almost inevitable no matter what, and that's one of the things GDPR is there for, to make sure errors can be fixed.

The EU AI act that went into force today also spells out some limitations.

I think it would have been fun to require companies using AI to prove/show exactly why it makes a determination or decision, upon request, if it's being used to make decisions where we say that such biases should not be perpetuated.

That this is burdensome or not yet possible with current LLM limitations is a problem to solve (or "tough shit") for the companies creating or using this technology; a necessary reckoning that they should have been forced to face from the onset.

Reply on news.ycombinator.com