I believe AI is going to kill us all if it is not eradicated and I have to spend a lot of time explaining that no, I don't mean it's going to become superintelligent https://fediscience.org/@ct_bergstrom/113028760435643985
-
If you want we can do some kinda terminology parsing where we separate OpenAI-style genAI (the technology that is dangerous because it produces bad quality results but people use it anyway) from traditional AI (relatively uncontroversial pre-2019 technologies like NLP, or machine learning in apt situations like OCR) from hypothetical sci-fi AI (a sentient being existing as executed software) but like, I think we're seeing that marketing creates language and "AI" means whatever OpenAI wants it to
-
@mcc It *really* ires me that a specific technology has claimed the broadly descriptive term for itself.
-
@michaelgemar It's frustrating, but also it's hard to get too mad because "AI" has always been a misnomer, we just keep changing what thing it inaccurately describes
-
@mcc @michaelgemar Alleged Intelligence
-
UPDATE: Again this is what I mean when I say AI is going to "kill us all". Okay, I admit "us all" is hyperbole because logically some of us may survive. But it is not hyperbole to say that *this* system is going to kill people: https://gizmodo.com/googles-ai-will-help-decide-whether-unemployed-workers-get-benefits-2000496215
This demonstrates well the true underlying purpose of AI: Giving people false confidence. Both before and after this tech the system made arbitrary decisions, but now the decisions will feel better because they're approved by Advanced Technology.
-
@mcc the thing that *really* gives me a sinking feeling in the pit of my stomach is that this might actually be a good thing. the phrase "backlog of claims" is euphemistic, and one needs to restate this internally as "large group of desperate people being made more desperate by neoliberal technocratic insistence upon massively inefficiently means-testing that illegitimately denies people resources that have been made their right by law"
-
@mcc it would probably be *better* if the "AI" here were a coin flip rather than a convolution matrix, so that certain people would be illegitimately and unaccountably denied their benefits *quickly* so that at least *some* proportion of the population receive their benefits promptly and fairly and there is no "claims backlog", just a light smattering of stochastic state crime against its citizens rather than the status quo, pervasive and near-universal theft of resources by bureaucracy
-
@glyph @mcc What will happen, as others have pointed out, is that if the "AI" is trained to duplicate the decisions that are currently being made, any unconscious biases will become explicit. The AI will wind up with explicit weights that penalize people for having a name associated with the discriminated-against group, and that happens because it's trained to match the human decision makers as closely as possible. This kind of thing happens even when the developers try to prevent it, because they'll take out obvious, explicit bias but leave in the correlations because they can't find them all.
-
@not2b @glyph That's the beauty of "AI", though. It can't be inspected, because we hid the decisions inside this box made of numbers. You can't prove we denied people benefits based on a Hispanic name, because instead of putting a regexp that an auditor could inspect in a piece of written source code we embedded it in an inauditable bunch of numbers. You can't prove we plagiarized your novel, because we scrambled the evidence we plagiarized into this big noisy matrix. Etc
-
@mcc @not2b @glyph this may have been mentioned sorry, but it'll be interesting to see how they square these "AI" with GDPR's right to understand how an automated decision was made.
Maybe they just won't be used, or will somehow blag that "thing goes in, decision comes out" is sufficient explanation.
For those lucky enough to be protected by such laws, of course.
-
-
-
-
-
-
@[email protected] @[email protected] @[email protected] @[email protected] @[email protected] a fun fact about gdpr is that you can just move your user data across continents and there's not really much teeth the law has on that? it's not against it (I'm sure not giving proper notice will get you in shit, but that's not my ass that's our ex administrator)
-
@ireneista @glyph @mcc @tehstu @not2b Sweden exempted the privately run website which publishes their entire population database including every resident’s address and which cars they own (incl. license plate)
Which, to be fair, there is a complaint about going on to the EU Data Protection Board right now, but like, there are many holes in the GDPR especially around government data
-
@puppygirlhornypost2 @tehstu @not2b @glyph @ireneista I thought the EU was introducing data residency laws for precisely this reason
-
@mcc @puppygirlhornypost2 @tehstu @not2b @glyph @ireneista you can only move an EU citizens’ personal data to a third country with an adequacy decision
The European Commission has so far recognised Andorra, Argentina, Canada (commercial organisations), Faroe Islands, Guernsey, Israel, Isle of Man, Japan, Jersey, New Zealand, Republic of Korea, Switzerland , the United Kingdom under the GDPR and the LED, the United States (commercial organisations participating in the EU-US Data Privacy Framework) and Uruguay as providing adequate protection.
Note that every few years the ECJ overturns the recognition of the US and then the parliament & US government ink a very slightly tweaked but obviously noncompliant deal and the process repeats because sometimes people make a new law and then decide they don’t actually like the consequences of it :drgn_scream_angry:
-
@erincandescent @tehstu @glyph @mcc @not2b @puppygirlhornypost2 yeah, we have no specific knowledge of any particular violations but in our privacy work we see things that make it very obvious that violation is constant and ongoing