#19 AGI & digital abyss in a slop world.
We have the winning gazpacho
The debate about whether AGI (or general AI) is (a) just around the corner or (b) “only in your dreams motherfucker” is beyond me, because I don’t have access to truly top-notch firsthand information...
What I do have access to is the ability to observe — uneasily — how many voices, not exactly gullible or sycophantic, are falling these days on the side of those who say it’s coming very soon.
Just to be clear — and to get a sense of how relevant timing might be — during Trump’s presidency.
Besides Ezra Klein in December, the latest to say so is Kevin Roose.
We all know Amara’s Law, which states that people tend to overestimate the impact of a new technology in the short term, but underestimate its real, lasting impact in the long term. Will AGI break this law? Given the current state of things, I’d say “abso-fucking-lutely.”
As Delia said the other day: “We don’t know how important this is, we have no idea.”
You’re reading ZERO PARTY DATA. The newsletter about the crazy crazy world news from a data protection perspective by Jorge García Herrero and Darío López Rincón.
In the spare time this newsletter leaves us, we like to solve complicated issues in personal data protection. If you’ve got one of those, give us a little wave. Or contact us by email at jgh(at)jorgegarciaherrero.com
🗞️News from Data World 🌍
.- They’re going to reform GDPR. I’ve seen things you humans wouldn’t believe. I’ve seen gamma-ray storms at the Tannhäuser Gate. And Axel Voss and Max Schrems shaking hands over the GDPR reform. We’ll see. For now, I’ll stick with grounded, well-argued takes like Markus Sullivan’s on “the disappearance of the RAT.”
.- On Tuesday, you got a special newsletter with a preview of what Jorge will contribute to the public consultation on the AEPD’s strategic plan (Jorge dixit via LinkedIn). Part 2 will drop next Tuesday, but it’s already clear that our favorite point in the plan is: “The Agency’s position on artificial intelligence, biometric and facial recognition systems, data spaces, neurodata, quantum technologies, blockchain, anonymization, and digital identity will be investigated and updated where necessary.”
The AEPD itself hints that it knows it needs to update the guide (in a kind of “we might consider it” tone), but it never hurts to give a little push.
And what about the staggering absence of video games in this point? One of life’s mysteries: focusing on futuristic neurodata and quantum stuff, but not on the sector currently affecting about one-third of the population with real issues. Maybe it’s a matter of scale, maybe the AEPD has already taken a couple of stabs at the topic, maybe the report on addictive patterns hit the mark, maybe one study called it one of the most invasive, or maybe it’s just a sector that joins every innovation party without giving it much thought. Hello, Pokémon Sleep.
In a more formal, serious, and legal tone, there’s been a proposal to include it in the plan. Darío won’t stop until they lock the doors at Jorge Juan when they see him coming.
.- I will be overusing the word essential all over the place this week, but this thread is indeed one: it highlights the striking idea shared by Sam Altman in the much-discussed interview with Chris Anderson at the TED Conference (five minutes from 38:59): ethical safeguards and alignment of AI models can be delegated to the model itself, interpreting (and applying or not) the relevant ethical framework in each case… according to its… statistical distribution among its users (that is, based on what the majority thinks). Of course, this is equivalent to adopting the Dorito credo, albeit expressed in classic BigTech jargon. The democratic automation of AI ethical safeguards is something I did not see coming…
.- Another long and tasty read (and one that, in my opinion, does not fall into ageism): Have today’s university students turned into idiots?
.- Don’t miss this essential, gripping article in the Financial Times: Welcome to slop world: how the hostile internet is driving us crazy, by Jacob Silverman, via Sergio Maldonado.
“The last bits of fellowship and ingenuity on the web are being swept away by a tide of so-called artificial intelligence.”
“The internet is now optimised for metrics that have nothing to do with human enjoyment, or convenience, or the profits of anyone except the platform overseers.”
.- Like everyone else, I’ve watched Adolescense. Like everyone else, I was deeply struck by it. The series is absolutely amazing, even if it loses a bit of momentum as it goes. It was inevitable: the first episode is so powerful that keeping that level of intensity was impossible. But I came here to say that — trying to move past my “delulu armchair expert” phase on the whole manosphere topic — I started reading Men Who Hate Women by Laura Bates. We’ll come back to this. For now, I found this related piece really interesting.
.- By the way, the latest season of Black Mirror — a cornerstone series for any self-respecting data-nerd TV addict — is quite a step up from the last few. What’s your favorite episode? Mine is Eulogy, just edging out the jaw-dropping Common People.
💀Death by Meme🤣
The cycle of the knife — of apps, of any digital service, and of all the AIs that are popping up like mushrooms these days.
📄Data-heavy documents for coffee-lovers☕️
.- Although the Dun case from the CJEU on algorithm transparency and explainability isn’t exactly new, the truly worthwhile analyses (in this case because they’re both simple and thorough) sometimes take time to emerge. I’m saving this one by Stefano Rossetti.
.- Europol released last week (perfect timing for maximum impact, honestly) a guide on vulnerabilities in biometric systems. It’s not exactly a data-nerd’s dream, but it can definitely help inspire how to approach these kinds of systems — especially when it comes to asking uncomfortable questions to the provider or manufacturer about the safeguards in place against Mission Impossible scenarios: silicone fingerprint molds (with or without the cooperation of the fingerprint’s owner), synthetic digital prints, whether fingerprints are left on any recoverable surface, how the system prevents fingerprints from being swapped around god knows how, whether it opens the door when targeted with a reset or bricking attack, hyper-realistic silicone masks (very Cruise saga, but with a real example in Canada mentioned by Europol), face morphing that merges multiple identities so both can be recognized, or facial deepfakes.
And watch out for how some commercial access control systems detect whether the finger placed is from a living person. It starts with a sensor measuring the body’s natural conductivity (to defeat silicone tricks), and then the next version adds pulse, vein patterns, and other “innovations.”
.- ¿Providers or deployers? : Everything everywhere all at once
🤖 Robots.txt or the AI thing
.- A document from the European Commission is out on generative AI for researchers. It’s not a big deal, but it’s useful to go over the principles. Just one more doc for the ever-growing repository on the hottest topic around.
.- The big big compulsory document that is required reading on the subject is AI Privacy Risks & Mitigations: Large Language Models (LLMs) by Isabel Barberá. Under the banner of the EDPB expert pool, we get a thorough and grounded piece: What is an LLM? How does it work? Data flow and related privacy risks, identification, evaluation and classification of risks, mitigation measures, residual risk assessment, three use cases, and links to methodologies, tools, and guidelines.
🧷 Useful tools 🔧
.- “Build Your Own” O365 Data Protection Impact Assessment for the Public Sector. That’s it: if anyone gives this thing a try, let us know how it goes.
.- The “non-tool”: Microsoft’s Copilot Recall. They’re back at it again with this privacy and confidentiality disaster.
📖Book of the week
Careless people, de Sarah Wynn-Williams. From 404 post:
The reason the book feels so important and cathartic is because, as a memoir, it does something that reported books about Facebook can’t quite do. It follows Wynn-Williams’ interior life as she recounts what drew her to Facebook (the opportunity to influence politics at a global scale beyond what she was able to do at the United Nations), the strategies and acts she made for the company (flying to Myanmar by herself to meet with the junta to get it unblocked there, for example), and her discoveries and ultimate disillusionment with the company as she goes on what often feels like repeated Veep-like quests to get Mark Zuckerberg to take interactions with world leaders seriously, to engineer a “spontaneous” interaction with Xi Jinping, to get him or Sandberg to care about the role Facebook played in getting Trump and other autocrats elected.
And this one from Techcrunch.
🙄 Da-Ta-dum bass
.- Why do AI company logos all look like buttholes?
.- It’s the end of the world: they’ve automated Italians.
.- META is META. You can object to their AI training in a much easier way with our script. Absolutely no sarcasm intended.
Thanks for reading Zero Party Data! Don’t miss next week’s edition!
Leave a comment!