I work at Southampton University. We are leading a national project looking at Responsible AI. One of the things that is very clear to anyone involved in AI is that it has picked up on the racisms prevalent in society - and is very much pro-white, and pro-male. AI at the moment is misogynistic and racist. Not sure where Os is getting his stuff from, but it might be great for him to send it to me so I can get it to the people who have found the exact opposite to be the case.
Is the computer science working in collaboration with humanities / social sciences then? Because if not then who is making the judgements about AI being misogynistic and racist? How is this defined and measured? Should computer scientists be doing the former or even the latter? I’m not sure they are but I’d be interested And I can tell you now - if there IS a link with social sciences or the like then there will be many people prepared to throw the findings of this entire project in the bin because of the perceived left wing bias of universities in general but especially in the non-STEM fields Incidentally I was going to bring up: - the point already been made of concerns AI to date is racist against black people (and Google being an over correction it seems) - question what the Google AI is really being used for and ask why it’s a big deal (beyond this nebulous “general agenda” point - point out my experience from the AI used in my work is that it is currently dreadful and has no clue what it’s doing. The best use I’ve found for it is generating NPC prompts for RPG names and also get it to generate more detailed and flowery descriptions that I can do (both in general and on the spot)
One of the best explanations I have had of this is that AI is not a logic system, it is a language system. All it does is ape the language it is picking up from the documents it is given. AI wants to please and will literally make up answers it thinks you want. Never let AI do your work for you as it can just plain lie. I am not an expert in social science or AI - so not much use on the rest. I guess I would say it is like this: Is the internet a fair and balanced place? If not… AI doesn’t have much chance. It is like a baby born to a couple of NF drug addicts.
Which for me is a million times more interesting than discussing whether a large language model has taken a wrong turn into crazytown for the nth time, because that is effectively a monthly occurrence at this point. There is this rush to integrate AI into everything, and it's just generally really bad at all of those things. It's a reasonable con artist -- ask it a question, and it will give a decent facsimile of an answer -- but it should not be given any sort of decision-making capacity, because it is still impossibly stupid. It drives badly and it writes badly and it draws badly and yet many of the biggest companies in the world are looking to outsource entire chains of production to it. And that's a genuine concern. As is enshittification, Cory Doctorow's term for the degradation of global platforms (Facebook, Twitter, Amazon, etc). As more and more of the internet comes to consist of AI-generated, algorithm-chasing slop, it will become harder and harder to find good ideas, good writing, good voices to listen to. And lord knows it's already difficult enough.
Maybe I am wrong, but as someone who identifies as right of centre I don't see any of this as woke. I see it as balance. For hundreds of years black or non white humans have been portrayed as a underclass. If you are not pure white, you are not pure human has been the narrative. Guess what, it makes no difference what colour your skin is, we all bleed the same blood and breathe the same air. We all have the same organs too. I have said it before, look at dogs. They are all sizes, all shapes, all colours, but still all dogs. They don't discriminate against each other as they havent been taught to. The word woke gets my goat as it is just a right wing way of trying to keep the equilibrium of fat rich white guys trying to run the world. Thursday morning soap box now put away.
Google images and the Google AI are different. I don’t know how you access the Google AI or if it is even publicly available. It looks like everyone commenting so far is going off of the article already linked in this chat with the mixed “German soldiers in 1943” etc
A healthy dose of scepticism needs to be applied to anything on the internet. I'm always looking for verifiable sources and give short shift to sensational one liners of a hysterical nature.
I agree with you. What I do see as woke is college campuses, the list of speakers and the courses available compared to courses that aren’t available. Funded mainly by Qatar and the usual communist suspects.
Genuine question - if you worked at, for example, Karachi or Lagos University, would AI have have picked up a different set of data which would have delivered an entirely different outcome ?
I don’t believe so… as AI is built from global data? But it is possible. I don’t know enough about it!
The problem with the internet is, whatever your personal prejudices may be - and we all have them - you can find stupendous amounts of information to confirm them. And your search history will keep directing you towards similar prejudice confirming sources, thus generating a perpetual cycle of self deception.
****, the implication of this is enormous. You work in AI? Does this put doubt on the 25 unbeaten, was it all computer generated? Are you even real?
The problem is (and you are missing the point again - as usual) that this isn’t just a language model problem. It’s not the output that is the problem - (you’re correct here. This can be fixed.) Google Gemini has an input problem. They are trying to overcorrect for previous perceived racism and it’s coming out the other end as more racist. This is a completely valid topic of conversation. We should be equally worried about biases against the white community as about other communities. Simply put; at the moment the social zeitgeist is heavily weighted to ignore anti-white racism
You are wrong. I also believe that race, religion and background doesn’t matter. Equality is something we should all strive for. But there IS a woke agenda and it is damaging for society in various ways.
I don't think there's a "Woke agenda" - I think there's overcorrection, then correction back. That's happening.
Equalities Minister, Kemi Badenoch, has been caught out by a FOI request. Badenoch has claimed to have met with LGBT groups on numerous occasions yet the FOI request has shown that she hasn’t met a single LGBT group in her 18(?) months in the role. The FOI request has, however, revealed that she has met with 2 fringe groups that are anti Trans. Minister for inequality perhaps?
