So today I'm attending the "AI for Humanity and Society" conference, connected to the billionaire-funded WASP-HS program.

#ai4hs2022 apparently, I'm wondering if anyone else is going to post to it...

So far it's kind of self-congratulatory in a very liberal academic way: "oh we're so good since we're saying we think about power"

First talk is going to be on the EU AI Act thing.

This is so very capitalism-subsuming-its-own-critics.

At least it's going to be interesting to see how @abebab chooses to play it on her keynote after lunch.

Now Martin Ulbricht from the EU Commission. Claiming that AI Hype has gone down some, which.. Maybe?

I still see a lot of AI Hype, personally, but at least a lot of criticism has started being more mainstream, but which kind of criticism?

What's interesting is that Ulbricht mentions that huge advances have been made in terms of picture/language models which.. Uh, kind of?

Also about data.

"AI doesn't need data, data needs AI"

I think that this kind of logic can and should be questioned: The datafication, especially huge data and "static" data is a very state capitalist thing, imo. "Seeing like a State" and all that.

It is absolute a very natural tendency to both _collect_ lots of data and "use" that data (in AI) if your fundamental logic is that of control, growth, and state capitalism. It doesn't have to be.

Now hyping up self-driving cars making for fewer car accidents.

This is not accurate.

See "Road to Nowhere" by Paris Marx

and just BUILD MORE TRAINS

Seems like at least @vdignum and @recklesscoding are also joining in on tooting a bit about #ai4hs2022 , which might be slightly less unhinged posting than mine.

Now talking about the AI Act, that is a combination of risk-based and application-based AI use. There's a few application for which we don't want to use AI - notably "social scoring" - but all the rest still have risks that need to be managed.

And the "prohibitions" are generally very narrow and even social scoring is something that we're kind of moving towards unfortunately.

Now going into discussing the high risk regulation.

Specific areas in which AI use needs to be explicitly allowed (but also with a bunch of exceptions): medical, safety critical systems, law enforcement etc. etc.

Some of the regulations are about "accuracy" which Ulbricht claims doesn't mean much since "not accurate" products won't be on the market anyway.

But "accuracy" is not ONE THING though, and this is exactly the problem! #ai4hs2022

Governance structure explanations (ECHLEG, AI Board and whatnot), timelines. It's In The Works, and orgs that want to have impact on it should probably get involved asap.

But I dunno, this feels to me like one of the cases where regulation/law people are mostly working with and for other people who are into regulation/law and people for whom this has an actual economic impact (and who can pay for access), i.e. large companies.

But at least they're trying.

Now a comment from fkn SVENSKT NÄRINGSLIV.

THE BUSINESS COMMUNITY.

I am not trusting these folks at all.

Also I kind of don't care what The Business Community says, because all they want is less regulations for themselves and more regulation for everyone else.

Which makes sense for them to want, but I'm unsure why they are given so much time at this academic conference?

Well, no I'm not, in fact, unsure, because this is a Wallenberg-funded conference, and Svenskt Näringsliv is a Wallenberg vehicle.

Talking about regulation costs for "SME"s. for GDPR and for the AI Act. Apparently the AI Act is almost twice the cost.

But the benefits? To the rest of us?

Also I'm not sure that the costs are not proportionally higher for large companies (in particular Big Tech companies) than for smaller companies. Use "less AI" and be smarter about what you do, and you'll be able to dodge a lot of the costs, I think.

Svenskt Näringsliv being sceptical of it is the highest compliment I can see for the #AIAct

Short reply from Martin, he's also sceptical to the comments from Carolina Brånby (Svenskt Näringsliv).

Now onto short topics around the "Grand Social Challenges of AI"

First topic:
Strengthening of the System World at the expense of the Living World

Habermas based. System -> Power and Money, Living -> The rest of it.

Very Dialectics.

I'm seeing echoes of metis/techne as well (yes I should 100% read Scott).

Colonisation of the Mind and the Person etc.

Primitive accumulation.

Musk example.

Something about AI-based hive mind stuff.

Also making a joke that a lot of people in academia have already moved over to Mastodon :)

Time for political sociologist to talk about the same topic from a different angle.

Mentioned the Arab Spring. And uh.. I have Feelings apparently.

Tech and access was said to be "decentralised", which is not necessarily true.

AI as a centralising tendency - very important, but can it be better? I think so, but needs conscious effort to break free from entrenched patterns.

AI as extending the reach of markets, systems, capital, bureaucracy, etc.

Also making references to Habermas System/Living worlds, and about the System "colonising" the Living world.

I'm not sure what I think about this use of the word "colonisation" - it feels like it's demeaning and forgetting and obscuring the very real and material colonisation which is still ongoing.

There's a lot of echoes here from thoughts I've had, but expressed in a different way.

I'm still not certain of the lack of material impacts...

Q about "AI art" supporting the entry of "amateurs" into the art world, and the "colonisation" or "entry" of a different "living world" in this way.

One of the responses is not about art but about citizen science etc.

I think this leads back to automation, and that automating technologies can help different groups differently.

Now Q about social media, and moving away from social media, as proposed by someone involved in building it up.

"Voting with your dollars/feet"

I'm not sure "just walking away" is a good answer - you need to have somewhere else, something else to go.

I think what's missing here is organisation and collective action.

Next session/topic: about misconceptions about/from AI in my understanding. #ai4hs2022

OK, so it's about Oversimplification, and how it leads to these misconceptions.

I think there's two different things that might interact here (before listening to what the presenters are saying): (1) Simplifying explanations of algorithms (it's just maths/statistics) and (2) Anthropomophising of AI (the AI 'wants' this, or 'thinks' that)

The first speaker is talking a lot from the perspective of AI for Education.

So I'm guessing this might be more about anthropomorphism then...

The other speaker in this topic seems to be talking more about the "it's just math" side of things.

I'm not really following currently.

Question about capitalism using AI to grasp more power, and asking the discutants to dispel that.

I don't think he can, but he tries - there are indeed counterdata creation going on, as he says, but I think we need a broader academic "revolutionary" or at least politically conscious approach to AI, being loud and clear in what is going on, in the political consequences and intentions, and the actual power grabs that capital is, indeed, doing.

Including this very conference

Q about HCI and AI, and the implications of that, and how to push HCI to avoid black box systems etc. I think?

I probably didn't follow the question properly, and I'm definitely not following the answer.

Next topic: AI: Equalizer of Oppressor?

#ai4hs2022

My answer: Both

First speaker also seem to be sceptical about it.

Critiquing technical determinism, and well, sure, that's true and all, but technologies do have politics. A car is not a train, is not a bus. A lorry isn't a wheelbarrow.

There was apparently an actual class that a lot of the audience took.

I didn't.

Uh, references to Latour, tools as delegated work. (My comment: Automation and capital as dead labour).

Next: technology as communication (speed bumps) and translation.

And here we go with making people legible to the technology (and the state/bureaucracy) again.

Tech in different contexts makes that tech mean and function differently.

Which is also true, but limited imo.

Not all technology is actually available to all the people, and possible to implement in all contexts

#ai4hs2022

Now a person working on politics and healthcare.

First Q: where are the discussion about AI in HC happening?

Second: where is it _not_ happening?

Participation and exclusion of different groups, and where and how various people are able to contribute and not.

(At this conference I'm missing the inclusion of what is Sweden is called "civil society", basically amateurs and political NGOs and associations, often connected to and based in unions)