Hacker News Reader

Comment by 0xbadcafebee | Hacker News Reader

Is it malfunctioning? Or is it just being generative AI? Maybe it's time people accept that generative AI is good at making shit up, and so we should just use it for that. Maybe we should stop pretending that "probabilistically predicting the next in a series of words" counts as intelligence.

I say, let the generative AI be racist. Let it hallucinate and come up with all kinds of crazy shit. And with that, stop using it for things that matter. Use it for art, entertainment, experiments, explorations. Use it to mine the depths of the human soul and reflect back at us what we are. Don't use it for law, health care, directions, or anything humans depend on. Acknowledge that it's just a stupid program with output that "looks impressive", and that we've all been fleeced into thinking it would be anything more. Regardless how many billions we sink into it, it only appears intelligent when we conceal how batshit crazy it can be.

View on HN

It is malfunctioning in the sense that it can be conditioned to not be like this. If you ask Midjourney to generate a picture of a Viking warrior, it doesn't spit out a central African guy wearing a Viking armor. It is not malfunctioning in the sense that it was conditioned to generate this kind of "2024 inclusive ad", where every group of people portrayed has an equal number of racial representative.

> Use it for art, entertainment, experiments, explorations. Use it to mine the depths of the human soul and reflect back at us what we are. Don't use it for law, health care, directions, or anything humans depend on.

you could, but the companies building these things very much want to sell it to other companies that do "things humans depend on", because that's a much larger market than just the entertainment industry.

Exactly. There’s no way the VCs who were dumb enough to invest in AI will capitulate so soon. It will happen eventually, but we have to put up with the baseless hype for a while longer.

I agree with everything you wrote, it's the mindset I wish people would have.

One note, though, in this case, it wasn't the generative AI that was racist, it's the people who added the extra configuration of not allowing (or at the very least, strongly discouraging) white people in their images. The racism wasn't a bug (like in the past with the gorilla incident), this time it was a feature that was deliberately put there.

From a game theory perspective, Sundar's approach is actually probably the safest approach. If they didn't add the "over the top" diversity and inclusion filter, Gemini would probably not be given a second chance if the twitter-verse was able to generate anything remotely objectionable as presently the diversity and inclusion crowd has somewhat captured social media, etc. and is quite vocal and experienced w.r.t. canceling people/companies/things.

If Google adds the diversity and inclusion filters and the social media influencers/stock market want it to be less filtered, then Gemini gets to live another day. The Gemini team can go in retro-actively and remove the diversity and inclusion filters and Sundar can write an apology letter.

Then if the Gemini team goes too far in the opposite direction, there is precedence to apologize again as they fine tune it.

That is actually excellent PR. I would also add there was a more minor benefit from causing a reaction on social media in the first place, which indirectly advertised their product.

>Don't use it for law, health care, directions, or anything humans depend on

This is kind of wishful thinking based on the idea that humans are much better. I've used it on law and health and ok it's not as reliable as a good professional but it will answer fairly well in a second or so whereas the professionals will be busy for a month, bill hundreds and you may still get incompetence. I'm also waiting two months for human solicitors to do a standard loan agreement, £2k in fees so far and no loan agreement yet. They are crying out to be replaced by AI as soon as it's a little better.

> it only appears intelligent when we conceal how batshit crazy it can be.

this can be applied to a broader aspect than just AI. in today's post-fact world, this is absolutely applicable as well to pretty much anything in life. we have people believing in the wildest ideas and not as a joke or just for the lulz, but wholeheartedly believing them as fact. people believe the earth is flat, that vaccines make you magnetic, or any of the other well known tropes.

all of that content has been used to train these "AI" systems. why would we ever expect it to not make shit up when a non-insignificant amount of its training data is made up shit.

thats not at all what is happening here. it looks a lot more like Gemini is following direct rules from a kind of DEI organization than anything else.

Reply on news.ycombinator.com