Hacker News Reader

Comment by thatguy0900 | Hacker News Reader

I think the problem is Google is also pretty crummy. Ai is in a golden age without poisoned data right now. But wait until Bing gets popular and blackhat seo types start poisoning chatgpt. It will stop being useful, and I suspect in a way that will be unfixable since these language models are so hard to wrangle.

View on HN

There's another side to that medal: At the moment nobody takes any issue with OpenAI doing filtering and curation in deciding what is part of their training data set, aside from perhaps the anti-bias crowd. "AI neutrality" is not yet a topic. Yet.

I've already seen that several times with image generation. Most recently was an article commenting on how the American smile was polluting generated photos. People can't decide what they want. Do they want licensed, curated commercial photos in the database or do they want search engine style neutrality? You really can't have both.

> People can't decide what they want.

You mean, so that you can satisfy everybody at the same time?

No, sorry, the world doesn't work this way.

Musk was on the air this week talking about how the current AI is biased on the left as he promotes his new AI company

Lets laugh and watch him make the "truth social" of AI, hopefully setting fire to another few billion in the process.

People laughed at paypal, tesla and spacex. Now he owns twitter.

Being the richest gives someone a lot of power to make others look foolish

It's always nice to hear Mr. Musk has not given up his cannabis habit.

…or maybe he has!

> without poisoned data right now

It already seems poisoned to me. It's popularity over correctness, because LLMs don't know semantics.

Reply on news.ycombinator.com