Open Letter: Stop the Uncritical Adoption of AI Technologies in Academia
-
@BrKloeckner @ErikJonker there is a French version of the open letter too! https://aeporreca.org/blog/lettre-ouverte-adoption-ia/
@olivia @ErikJonker The english version is fine already, what made me hesitate to sign is the specific address to dutch universities.
-
@olivia @ErikJonker The english version is fine already, what made me hesitate to sign is the specific address to dutch universities.
@BrKloeckner @olivia the technology will not go away but we are in control where to use it or not, how etc. Personally I don't believe banning and not-promoting is the way to mitigate the large risks and problems associated with GenAI. So banning, discouraging in a general sense is not the best strategy in my view, but I understand many people disagree with me.
-
@BrKloeckner @olivia the technology will not go away but we are in control where to use it or not, how etc. Personally I don't believe banning and not-promoting is the way to mitigate the large risks and problems associated with GenAI. So banning, discouraging in a general sense is not the best strategy in my view, but I understand many people disagree with me.
@ErikJonker @olivia "we are in control" Are we? My articles are scrapped, some students send me AI slop to be corrected as their work, searching info on the web has been made very difficult.
The main risks are ecological; loss of autonomy by the normalization of gen AI production levels; wealth and power given to a handfull of fashbros. I don't see how to mitigate those without discouraging use of gen AI.
Our disagreement could be on what the main risks are, but I think the three above are pretty high, or that we xould mitigate them while encouraging using gen AI, but I don't see how. -
@ErikJonker
> "the large risks and problems associated with GenAI"Assuming you're not one of those who believe in the AI apocalypse, the real issues are:
+ AI is a financial bubble, concentrating money and power in the hands of a US-based tech oligarchy;
+ AI is a class weapon, aimed at degrading the conditions of the working class;
+ AI is an instrument of war, being used to advance US-and-friends' imperialism and supremacism worldwide;
@BrKloeckner @olivia -
@ErikJonker
> "the large risks and problems associated with GenAI"Assuming you're not one of those who believe in the AI apocalypse, the real issues are:
+ AI is a financial bubble, concentrating money and power in the hands of a US-based tech oligarchy;
+ AI is a class weapon, aimed at degrading the conditions of the working class;
+ AI is an instrument of war, being used to advance US-and-friends' imperialism and supremacism worldwide;
@BrKloeckner @olivia@ErikJonker
+ AI is a skill sink, making (young) people less skilled in critical thinking and language-related tasks; it's very different to have top mathematicians using AI as an aid for what they already know how to do, than to have young people never even learning how to do something because the tool always did it for them;
+ AI is a socio-environmental risk, with data centers and mineral extraction causing harm to communities in Global South countries and elsewhere.
@BrKloeckner @olivia -
@ErikJonker
+ AI is a skill sink, making (young) people less skilled in critical thinking and language-related tasks; it's very different to have top mathematicians using AI as an aid for what they already know how to do, than to have young people never even learning how to do something because the tool always did it for them;
+ AI is a socio-environmental risk, with data centers and mineral extraction causing harm to communities in Global South countries and elsewhere.
@BrKloeckner @olivia@ErikJonker
> "discouraging in a general sense is not the best strategy in my view"So I guess you also don't approve of discouraging smoking and drinking and cheating on exams and shooting heroin. All of these things "will not go away", "we're in control where to use it or not" and have "large risks and problems associated". But they are harmful to society, so we ban them, or we regulate them. China is big on AI regulation, the EU is somewhat trying, and the US don't care.
@BrKloeckner @olivia -
@ErikJonker
> "discouraging in a general sense is not the best strategy in my view"So I guess you also don't approve of discouraging smoking and drinking and cheating on exams and shooting heroin. All of these things "will not go away", "we're in control where to use it or not" and have "large risks and problems associated". But they are harmful to society, so we ban them, or we regulate them. China is big on AI regulation, the EU is somewhat trying, and the US don't care.
@BrKloeckner @olivia@ErikJonker
The letter is against "uncritical adoption", meaning "think before you act". Think of the harms and benefits for your community. AI companies are obviously benefitting from AI adoption, so unless you're a paid shill, you've got to consider whether this is really in your best interest and that of those around you. The fact that companies already got students hooked up on their stuff is not something to be celebrated. It's sad, and terrible, and must be countered.
@BrKloeckner @olivia -
@olivia @ErikJonker The english version is fine already, what made me hesitate to sign is the specific address to dutch universities.
@BrKloeckner @ErikJonker please sign anyway and consider connecting with local groups if you're at a university in France and addressing it to your university too. There's a differently worded French one too, not anything to do with me, you also may want to look into.
-
Open Letter: Stop the Uncritical Adoption of AI Technologies in Academia
It's been a year & people tell me they've changed their mind, feel safer now to sign, etc. So I'm sharing it again for them. We're almost at 2k & anybody from anywhere can add their names.
@olivia Not in education at all, but former Electronic Design Engineer and IT person, but I fully agree!
-
@ErikJonker @olivia "we are in control" Are we? My articles are scrapped, some students send me AI slop to be corrected as their work, searching info on the web has been made very difficult.
The main risks are ecological; loss of autonomy by the normalization of gen AI production levels; wealth and power given to a handfull of fashbros. I don't see how to mitigate those without discouraging use of gen AI.
Our disagreement could be on what the main risks are, but I think the three above are pretty high, or that we xould mitigate them while encouraging using gen AI, but I don't see how.@BrKloeckner @olivia i don't have all the answers but one measure i think is national government and education instutions could steer people to use models that are more responsible, follow our laws/regulations, have a limited use of energy and resources then the topmodels from bigtech. This is already happening, inside dutch government an AI service is developed on it's own infrastructure, using Mistral medium and other models in the future.
-
@BrKloeckner @olivia i don't have all the answers but one measure i think is national government and education instutions could steer people to use models that are more responsible, follow our laws/regulations, have a limited use of energy and resources then the topmodels from bigtech. This is already happening, inside dutch government an AI service is developed on it's own infrastructure, using Mistral medium and other models in the future.
@ErikJonker @olivia Does Mistral Medium really have limited use of energy and ressources? I know Mistral is prone to use and sell others models, so could Mistral continue to work without the big ones scrapping the web endlessly and burning fuel to train their models?
I have had many people say "don't reject AI, there are ethical models", then when asked, all confessed they use the no-ethical ones. The power of OpenAI and Anthropic tools is debatable, but seems unachievable ethically (they say so themselves).
-
@ErikJonker @olivia Does Mistral Medium really have limited use of energy and ressources? I know Mistral is prone to use and sell others models, so could Mistral continue to work without the big ones scrapping the web endlessly and burning fuel to train their models?
I have had many people say "don't reject AI, there are ethical models", then when asked, all confessed they use the no-ethical ones. The power of OpenAI and Anthropic tools is debatable, but seems unachievable ethically (they say so themselves).
@ErikJonker @olivia Also, I am in the minority globally (not on mastodon) on this issue, and I pay dearly for it. My government wants, like EU, to build huge datacenters and universities are all about the AI hype now. An applauded colleague presented how he uses AI for dermatology (spoiler: it has zero chance to work in the end, scientifically it was completely bonkers but he does get funding easily).
-
@ErikJonker @olivia Also, I am in the minority globally (not on mastodon) on this issue, and I pay dearly for it. My government wants, like EU, to build huge datacenters and universities are all about the AI hype now. An applauded colleague presented how he uses AI for dermatology (spoiler: it has zero chance to work in the end, scientifically it was completely bonkers but he does get funding easily).
@ErikJonker @olivia I work near Paris while living in Lyon where my wife works, and this year there were 4 job openings roughly at my level . All where in AI (and beyond my own situation, it makes no sense to hire that many people in any single domain, let alone one where competition by the industry is so fierce).
Just so you understand some harm AI is doing but tends do be under the radar. -
@ErikJonker @olivia Does Mistral Medium really have limited use of energy and ressources? I know Mistral is prone to use and sell others models, so could Mistral continue to work without the big ones scrapping the web endlessly and burning fuel to train their models?
I have had many people say "don't reject AI, there are ethical models", then when asked, all confessed they use the no-ethical ones. The power of OpenAI and Anthropic tools is debatable, but seems unachievable ethically (they say so themselves).
@BrKloeckner @olivia ...following AI for several years now, the good news is that there is steady increase in capabilities but also a lot of work is done on making models cheaper/more efficient in their use of resources. I think there is (or will be) a sweet spot with "good enough" models with added value, with acceptable use of resources, but maybe i am too optimistic.
-
@BrKloeckner @olivia ...following AI for several years now, the good news is that there is steady increase in capabilities but also a lot of work is done on making models cheaper/more efficient in their use of resources. I think there is (or will be) a sweet spot with "good enough" models with added value, with acceptable use of resources, but maybe i am too optimistic.
@ErikJonker @olivia In a capitalist world, I see little reason for optimism, since use of ressource is only ever factored through monetary cost.
Moreover, I have a hard time believing training data is not core to the technology. How could one build something decent (beyond specialized ML trained on specific data, but this is mere statistics) with only the data you can gather legitimately? -
@ErikJonker @olivia In a capitalist world, I see little reason for optimism, since use of ressource is only ever factored through monetary cost.
Moreover, I have a hard time believing training data is not core to the technology. How could one build something decent (beyond specialized ML trained on specific data, but this is mere statistics) with only the data you can gather legitimately?@ErikJonker @olivia To mention another source of problems lived by abstinents like me, the web has been rotten down by AI, and this is not something we can hope to repair easily, so I would like us to stop degrading it.
All the harms done by the big players (especially to workers from the global south) has not be done for the sake of it, they are not *that* evil. How can we condone this harm already done? Can we really hope harm will stop at "good enough"? And if this is our desired outcome, should we not wait until this "ethical models" are "good enough" *before* adopting AI?
-
@ErikJonker @olivia In a capitalist world, I see little reason for optimism, since use of ressource is only ever factored through monetary cost.
Moreover, I have a hard time believing training data is not core to the technology. How could one build something decent (beyond specialized ML trained on specific data, but this is mere statistics) with only the data you can gather legitimately?@BrKloeckner @olivia ...that is a real issue that is difficult to solve, there are models however that are more responsble/decent then others, it is not a binary thing. So yes we will use models now and in the future that have some degree of illegitimate trainingdata. In the Netherlands we have a small model that is only trained on legitimate data by the way, GPT-NL.
-
@BrKloeckner @olivia ...that is a real issue that is difficult to solve, there are models however that are more responsble/decent then others, it is not a binary thing. So yes we will use models now and in the future that have some degree of illegitimate trainingdata. In the Netherlands we have a small model that is only trained on legitimate data by the way, GPT-NL.
@ErikJonker @BrKloeckner if you're interested in what I think about models...
-
@ErikJonker @olivia To mention another source of problems lived by abstinents like me, the web has been rotten down by AI, and this is not something we can hope to repair easily, so I would like us to stop degrading it.
All the harms done by the big players (especially to workers from the global south) has not be done for the sake of it, they are not *that* evil. How can we condone this harm already done? Can we really hope harm will stop at "good enough"? And if this is our desired outcome, should we not wait until this "ethical models" are "good enough" *before* adopting AI?
@BrKloeckner @ErikJonker yes, absolutely, ethical models without stolen or unethically obtained data usually do not work well or even at all.
Also have you seen the work by @dingemansemark on this?
-
J jwcph@helvede.net shared this topic
