Robert Bateman

Facebook's new opt-out form is very hard to find. After a lot of digging and several preliminary clicks, I ended up here. I'm now going to live-tweet the process of objecting to Meta's ad-targeting. This is pretty niche stuff. So if you're interested, I like you. 🧵

@gateklons @chapell68 I don't either, but they did not apply the non-existing provision by allowing the processing. They just chose not to penalise the organisation on this point. To say they "created" the provision would suggest they found the processing to be lawful and allowed it.

@RobertJBateman @chapell68 I don't think there was any lack of clarity in the law. The DPA just thought there should've been an Art. 9(2) exception when there obviously wasn't one (despite lobbying from the insurance industry) so they essentially created one through non-enforcement.

I wanted to read the decision first before commenting further, but the GDPRhub description is exactly right: the APD/GBA let the controller off the hook for its *health data* consent-or-pay model (despite agreeing that it's illegal) because the Belgian legislator didn't provide

@gateklons @chapell68 To be clear, this was simply a mitigation, no? The DPA still said they shouldn't have done what they did, but didn't impose a fine because they thought the law was unclear. While I don't agree in this case, many companies have been punished where the law actually was unclear.

@chapell68 @RobertJBateman Oh, sorry PoC = pay-or-consent, my bad. 😅 Yeah, illegal under the normal regime, even more so under the stricter Art. 9 regime. But the EDPB opinion on PoC with respect to large online platform shows DPAs aren't that against the model (despite its obvious illegality).

@gabrielazanfir I guess that eliminates any prospect of an adequacy decision.

@akrylov @PrivacyMatters I was looking into this today because I don't know the process—her term would normally expire after 7 years right? But can the president intervene? The FTC is unrecognisable under Khan.

@RobertJBateman @PrivacyMatters I can’t imagine Trump’s win and the GOP’s won Senate bodes well for the current FTC and Chair either.

@PrivacyMatters I knew of her as "the AG overseeing CalOPPA enforcement" before she was Biden's VP pick. Back when CA was basically the only state doing anything on privacy. I am pretty cynical about politics so not particularly enthused about her candidacy but she had some interesting proposals

@mozdca @PrivacyMatters Harris had a ton of policies, many of them were actually really interesting. You might not like them or believe she would implement them, but she definitely had plans. There's a long list of positions here, and afaik none mention her race or gender: kamalaharris.com/issues/

@PrivacyMatters You seem to be quite active on X, yet you still support a candidate who has no clear plans besides her skin color and gender, is backed by the MSMs and warmongers and celebrities from Diddy's list, and spreads rhetoric that has almost got her opponent killed. Twice. Why?

@1Br0wn Alright then I will.

@floorter @gateklons Anything commercial is 100% fair game, no questions asked, as I believe the CJEU confirmed on Friday.

@RobertJBateman @gateklons Oh yeah? Name all legitimate interests. 😇

If you think you understand the GDPR, you don't understand the GDPR

@novocrypto You might be putting the cart before the horse. What's special about a blank output? Why can't a deterministic system provide a blank output without being told to? Perhaps providing a blank output is consistent with the model's predictive algorithm? It's just not clear why.

Two key points (responding to a common critique of my argument and evidence refuting the Chinese Room model to explain LLMs): 1. You’re right that the algorithm can and does allow for empty outputs, but in your example, the model was explicitly instructed to produce a void: “respond with an empty string … literally nothing”. Whereas in my original scenario the model *spontaneously* generated a void response without being told to, indicating emergent behaviour. This happens when the model dynamically resolves internal contradictions and determines that silence is optimal not because it’s following any rule! 2. The key is in computational irreducibility (Wolfram’s Principle) and Rice’s Theorem that tell us that no rule-based system, no matter how sophisticated, can predict or show the emergent behaviours of a model that relies on continuous, probabilistic state transitions. The crucial distinction lies in Gödel’s Incompleteness Theorem: any formal, rule-based system will have true behaviours that can’t be derived from its axioms. While the Chinese Room must operate within fixed syntactic mappings, LLMs show behaviours that emerge from continuous, probabilistic updates to an underlying state space that transcends any fixed set of logical rules, again similar to non-reducible systems in Wolfram’s Principle of Computational Irreducibility. So, even if a fancy and modernised version of the Chinese Room simulated conflicting goals, it could never match the unpredictable, context-sensitive outputs of an LLM. No deterministic system can account for the unbounded variability produced through real-time, high-dimensional optimisation. Basically, while generating an empty response on command is trivial (as you rightly point out), the emergent, context-driven void I observed reflects a behaviour that a system based purely on next-token prediction can’t replicate, because the Chinese Room’s fixed rules would always lack the unbounded capacity for dynamic state adaptation that underpins LLMs’ real-time decision-making.

@jennifercobbe I think they consider themselves to have done something already 🥲

@jennifercobbe Don't worry they're monitoring the situation

@SocialistsAgai3 @GBGreatAgain @BBCNews @itvnews @SkyNews Yes. On the other hand, I don't think it's a good strategy. If he denies it this time, he'll be expected to comment on the next rumour. If he is silent next time, it will strongly indicate that there is a superinjunction. Fyi I don't like the liberal use of superinjunctions.

@RobertJBateman @GBGreatAgain @BBCNews @itvnews @SkyNews He could just come out say there wasn't a superinjuction and all channels could report it. But he himself would have broken the superinjuction. He is stuck if there is one as he can't publically deny it! If there isn't he is free to deny it anytime! Let's see when he denied it

Quick heads up for anyone claiming to be a journalist. Any real journalist would already know. #superinjunction

@SocialistsAgai3 @GBGreatAgain @BBCNews @itvnews @SkyNews If it exists they would be sacked and their publication would be held in contempt of court. If it doesn't exist then there's no story there and it would set a precedent for future superinjunction rumours. I doubt any mainstream publication would risk it.

@GBGreatAgain Okay as soon as @laurakennesberg @BBCNews @itvnews @skynews.memtion the super injection then we know it doesn't exist. But for some reason they all seem to be staying quiet? Why?

@montezumachavez Is that actually in the DPA's remit? They aren't market surveillance authority yet!

The Dutch DPA is calling for input on manipulative, deceptive, and exploitative AI systems", which will be prohibited under article 5 (1) (a) and (b) of the AI Act. See lnkd.in/eGTMzTMy.

@James69382185 Yeah, even knowing this I still didn't get it. I get it now—there's no portrait because he won't be there—it's just a bad joke.

@RobertJBateman He's resigned.

I don't get it.

@gateklons It's all a bit politically-motivated-and-constitutionally-questionable-US-state-senate-bill for my tastes.

@RobertJBateman Good point! There's just no coherence from the EC about what is and is not considered addictive and prohibited, probably because this quickly runs into fundemental rights issues and actual desires of users. But they want to be seen as tough and cracking down on Big Bad Tech.

"The available brain time of young Europeans is not a currency for social media — and it never will be." I mean that's literally the business model of ad-supported social media platforms, is it not? ec.europa.eu/commission/pre…