Category: AI

The good fight

This week saw a small gathering to celebrate the 25th anniversary (more or less) of the Foundation for Information Policy Research, a think tank led by Cambridge and Edinburgh University professor Ross Anderson. FIPR’s main purpose is to produce tools and information that campaigners for digital rights can use. Obdisclosure: I am a member of its advisory council.

What, Anderson asked those assembled, should FIPR be thinking about for the next five years?

When my turn came, I said something about the burnout that comes to many campaigners after years of fighting the same fights. Digital rights organizations – Open Rights Group, EFF, Privacy International, to name three – find themselves trying to explain the same realities of math and technology decade after decade. Small wonder so many burn out eventually. The technology around the debates about copyright, encryption, and data protection has changed over the years, but in general the fundamental issues have not.

In part, this is because what people want from technology doesn’t change much. A tangential example of this presented itself this week, when I read the following in the New York Times, written by Peter C Baker about the “Beatles'” new mash-up recording:

“So while the current legacy-I.P. production boom is focused on fictional characters, there’s no reason to think it won’t, in the future, take the form of beloved real-life entertainers being endlessly re-presented to us with help from new tools. There has always been money in taking known cash cows — the Beatles prominent among them — and sprucing them up for new media or new sensibilities: new mixes, remasters, deluxe editions. But the story embedded in “Now and Then” isn’t “here’s a new way of hearing an existing Beatles recording” or “here’s something the Beatles made together that we’ve never heard before.” It is Lennon’s ideas from 45 years ago and Harrison’s from 30 and McCartney and Starr’s from the present, all welded together into an officially certified New Track from the Fab Four.”

I vividly remembered this particular vision of the future because just a few days earlier I’d had occasion to look it up – a March 1992 interview for Personal Computer World with the ILM animator Steve Williams, who the year before had led the team that produced the liquid metal man for the movie Terminator 2. Williams imagined CGI would become pervasive (as it has):

“…computer animation blends invisibly with live action to create an effect that has no counterpart in the real world. Williams sees a future in which directors can mix and match actors’ body parts at will. We could, he predicts, see footage of dead presidents giving speeches, films starring dead or retired actors, even wholly digital actors. The arguments recently seen over musicians who lip-synch to recordings during supposedly ‘live’ concerts are likely to be repeated over such movie effects.”

Williams’ latest work at the time was on Death Becomes Her. Among his calmer predictions was that as CGI became increasingly sophisticated the boundary between computer-generated characters and enhancements would become invisible. Thirty years on, the big excitement recently has been Harrison Ford’s deaging for Indiana Jones and the Dial of Destiny. That used CGI, AI, and other tools to digitally swap in his face from 1980s footage.

Side note: in talking about the Ford work to Wired, ILM supervisor Andrew Whitehurst, exactly like Williams in 1992, called the new technology “another pencil”.

Williams also predicted endless legal fights over copyright and other rights. That at least was spot-on; AI and the perpetual reuse of retained footage without further payment is part of what the recent SAG-AFTRA strikes were about.

Yet, the problem here isn’t really technology; it’s the incentives. The businessfolk of Hollywood’s eternal desire is to guarantee their return on investment, and they think recycling old successes is the safest way to do that. Closer to digital rights, law enforcement always wants greater access to private communications; the frustration is that incoming generations of politicians don’t understand the laws of mathematics any better than their predecessors in the 1990s.

Many of the speakers focused on the issue of getting government to listen to and understand the limits of technology. Increasingly, though, a new problem is that, as Bruce Schneier writes in his latest book, The Hacker’s Mind, everyone has learned to think like hackers and subvert the systems they’re supposed to protect. The Silicon Valley mantra of “ask forgiveness, not permission” has become pervasive, whether it’s a technology platform deciding to collect masses of data about us or a police force deciding to stick a live facial recognition pilot next to Oxford Circus tube station. Except no one asks for forgiveness either.

Five years ago, at FIPR’s 20th anniversary, when GDPR is new, Anderson predicted (correctly) that the battles over encryption would move to device access. Today, it’s less clear what’s next. Facial recognition represents a step change; it overrides consent and embeds distrust in our public infrastructure.

If I were to predict the battles of the next five years, I’d look at the technologies being deployed around European and US borders to surveil migrants. Migrants make easy targets for this type of experimentatioon because they can’t afford to protest and can’t vote. “Automated suspicion,” Euronews.next calls it. That habit of mind is danagerous.

Illustrations: The liquid metal man in Terminator 2 reconstituting itself.

Wendy M. Grossman is the 2013 winner of the Enigma Award. Her Web site has an extensive archive of her books, articles, and music, and an archive of earlier columns in this series. She is a contributing editor for the Plutopia News Network podcast. Follow on Mastodon

The one hundred

Among the highlights of this week’s hearings of the Covid Inquiry were comments made by Helen MacNamara, who was the deputy cabinet secretary during the relevant time, about the effect of the lack of diversity. The absence of women in the room, she said, led to a “lack of thought” about a range of issues, including dealing with childcare during lockdowns, the difficulties encountered by female medical staff in trying to find personal protective equipment that fit, and the danger lockdowns would inevitably pose when victims of domestic abuse were confined with their abusers. Also missing was anyone who could have identified issues for ethnic minorities, disabled people, and other communities. Even the necessity of continuing free school lunches was lost on the wealthy white men in charge, none of whom were ever poor enough to need them. Instead, MacNamara said, they spent “a disproportionate amount” of their time fretting about football, hunting, fishing, and shooting.

MacNamara’s revelations explain a lot. Of course a group with so little imagination about or insight into other people’s lives would leave huge, gaping holes. Arrogance would ensure they never saw those as failures.

I was listening to this while reading posts on Mastodon complaining that this week’s much-vaunted AI Safety Summit was filled with government representatives and techbros, but weak on human rights and civil society. I don’t see any privacy organizations on the guest list, for example, and only the largest technology platforms needed apply. Granted, the limit of 100 meant there wasn’t room for everyone. But these are all choices seemingly designed to make the summit look as important as possible.

From this distance, it’s hard to get excited about a bunch of bigwigs getting together to alarm us about a technology that, as even the UK government itself admits, may – even most likely – will never happen. In the event, they focused on a glut of disinformation and disruption to democratic polls. Lots of people are thinking about the first of these, and the second needs local solutions. Many technology and policy experts are advocating openness and transparency in AI regulation.

Me, I’d rather they’d given some thought to how to make “AI” (any definition) sustainable, given the massive resources today’s math-and-statistics systems demand. And I would strongly favor a joint resolution to stop using these systems for surveillance and eliminate predictive systems that pretend to be sble to spot potential criminals in advance or decide who are deserving of benefits, admission into retail stores, or parole. But this summit wasn’t about *us*.

***

A Mastodon post reminded me that November 2 – yesterday – was the 35th anniversary of the Morris Worm and therefore the 35th anniversary of the day I first heard of the Internet. Anniversaries don’t matter much, but any history of the Internet would include this now largely-fotgotten (or never-known) event.

Morris’s goals were pretty anodyne by today’s standards. He wanted, per Wikipedia, to highlight flaws in some computer systems. Instead, the worm replicated out of control and paralyzed parts of this obscure network that linked university and corporate research institutions, who now couldn’t work. It put the Internet on the front pages for the first time.

Morris became the first person to be convicted of a felony under the brand-new Computer Fraud and Abuse Act (1986); that didn’t stop him from becoming a tenured professor at MIT in 2006. The heroes of the day were the unsung people who worked hard to disable the worm and restore full functionality. But it’s the worm we remember.

It was another three years before I got online myself, in 1991, and two or three more years after that before I got direct Internet access via the now-defunct Demon Internet. Everyone has a different idea of when the Internet began, usually based on when they got online. For many of us, it was November 2, 1988, the day when the world learned how important this technology they had never heard of had already become.

***

This week also saw the first anniversary of Twitter’s takeover. Despite a variety of technical glitches and numerous user-hostile decisions, the site has not collapsed. Many people I used to follow are either gone or posting very little. Even though I’m not experiencing the increased abuse and disinformation I see widely reported, there’s diminishing reward for checking in.

There’s still little consensus on a replacement. About half of my Twitter list have settled in on Mastodon. Another third or so are populating Bluesky. I hear some are finding Threads useful, but until it has a desktop client I’m out (and maybe even then, given its ownership). A key issue, however, is that uncertainty about which site will survive (or “win”) leads many people to post the same thing on multiple services. But you don’t dare skip one just in case.

For both philosophical and practical reasons, I’m hoping more people will get comfortable on Mastodon. Any corporate-owned system will merely replicate the situation in which we become hostages to business interests who have as little interest in our welfare as Boris Johnson did according to MacNamara and other witnesses. Mastodon is not a safe harbor from horrible human behavior, but with no ads and no algorithm determining what you see, at least the system isn’t designed to profit from it.

Illustrations: Former deputy cabinet secretary Helen MacNamara testifying at the Covid Inquiry.

Wendy M. Grossman is the 2013 winner of the Enigma Award. Her Web site has an extensive archive of her books, articles, and music, and an archive of earlier columns in this series. She is a contributing editor for the Plutopia News Network podcast. Follow on Mastodon

The end of cool

For a good bit of this year’s We Robot, it felt like abstract “AI” – that is, algorithms running on computers with no mobility – had swallowed the robots whose future this conference was invented to think about. This despite a pre-conference visit to Boston Dynamics, which showed off its Atlas
robot‘s ability to do gymnastics. It’s cute, but is it useful? Your washing machine is smarter, and its intelligence solves real problems like how to use less water.

There’s always some uncertainty about boundaries at this event: is a machine learning decision making system a robot? At the inaugural We Robot in 2012, the engineer Bill Smart summed up the difference: “My iPhone can’t stab me in my bed.” Of course, neither could an early Roomba, which most would agree was the first domestic robot. However, it was also dumb as a floor tile, achieving cleanliness through random repetition rather than intelligent mapping. In the Roomba 1.0 sense, a “robot” is “a device that does boring things so I don’t have to”. Not cool, but useful, and solves a real problem

During a session in which participants played a game designed to highlight the conflicts inherent in designing an urban drone delivery system, Lael Odhner offered yet another definition: “A robot is a literary device we use to voice our discomfort with technology.” In the context of an event where participants think through the challenges robots bring to law and policy, this may be the closest approximation.

In the design exercise, our table’s three choices were: fund the FAA (so they can devise and enforce rules and policies), build it as a municipally-owned public service both companies and individuals can use as customers, and ban advertising on the drones for reasons of both safety and offensiveness. A similar exercise last year produced more specific rules, but also led us to realize that a drone delivery service had no benefits over current delivery services.

Much depends on scale. One reason we chose a municipal public service was the scale of noise and environmental impact inevitably generated by multiple competing commercial services. In a paper, Woody Hartzog examined the meaning of “scale”: is scale *more*, or is scale *different*? You can argue, as net.wars often has, that scale *creates* difference, but it’s rarely clear where to place the threshold, or how reaching it changes a technology’s harms or who it makes vulnerable. Ryan Calo and Daniella DiPaola suggested that rather than associate vulnerability with particular classes of people we should see it as variable with circumstances: “Everyone is vulnerable sometimes, and vulnerability is a state that can be created and manipulated toward particular ends.” This seems a more logical and fairer approach.

An aspect of this is that there are two types of rules: harm rules, which empower institutions to limit harm, and power rules, which empower individuals to protect themselves. A possible worked example soon presented itself in Kegan J Strawn;s and Daniel Sokol‘s paper on safety techniques in mobile robots, which suggested copying medical ethics’ consent approach. Then someone described the street scene in which every pedestrian had to give consent to every passing experimental Tesla, a possibly an even worse scenario than ad-bearing delivery drones. Pedestrians get nothing out of the situation, and Teslas don’t become safer. What you really want is for car companies not to test the safety of autonomous vehicles on public roads with pedestrians as unwitting crash test dummies.

I try to think every year how our ideas about inegrating robots into society are changing over time. An unusual paper from Maria P. Angel considered this question with respect to privacy scholarship by surveying 1990s writing and 20 years of papers presented at Privacy Law Scholars. We Robot co-founders Calo, Michael Froomkin, and Ian Kerr partly copied its design. Angel’s conclusion is roughly that the 1990s saw calls for an end to self-regulation while the 2000s moved from privacy as necessary for individual autonomy and self-determination to collective benefits and most recently to its importance for human flourishing.

As Hartzog commented, he came to the first We Robot with the belief that “Robots are magic”, only to encounter Smart’s “really fancy hammers.” And, Smart and Cindy Grimm added in 2018, controlled by sensors that are “late, noisy, and wrong”. Hartzog’s early excitement was shared by many of us; the future looked so *interesting* when it was almost entirely imaginary.

Over time, the robotic future has become more nowish, and has shifted in response to technological development; the discussion has become more about real systems (2022) than imagined future ones. The arrival of real robots on our streets – for example, San Francisco’s 2017 use of security robots to deter homeless camps – changed parts of the discussion from theoretical to practical.

In the mid-2010s, much discussion focused on problems of fairness, especially to humans in the loop, who, Madeleine Claire Elish correctly predicted in 2016 would be blamed for failures. More recently, the proliferation of data-gathering devices (sensors, cameras) into everything from truckers’ cabs to agriculture and the arrival of new algorithmic systems dubbed AI has raised awareness of the companies behind these technologies. And, latterly, that often the technology diverts attention from the better possibilities of structural change.

But that’s not as cool.

Illustrations: Boston Dynamics’ Atlas robots doing synchronized backflips (via YouTube).

Wendy M. Grossman is the 2013 winner of the Enigma Award. Her Web site has an extensive archive of her books, articles, and music, and an archive of earlier columns in this series. She is a contributing editor for the Plutopia News Network podcast. Follow on Mastodon.

Doom cyberfuture

Midway through this year’s gikii miniconference for pop culture-obsessed Internet lawyers, Jordan Hatcher proposed that generational differences are the key to understanding the huge gap between the Internet pioneers, who saw regulation as the enemy, and the current generation, who are generally pushing for it. While this is a bit too pat – it’s easy to think of Millennial libertarians and I’ve never thought of Boomers as against regulation, just, rationally, against bad Internet law that sticks – it’s an intriguing idea.

Hatcher, because this is gikii and no idea can be presented without a science fiction tie-in, illustrated this with 1990s movies, which spread the “DCF-84 virus” – that is, “doom cyberfuture-84”. The “84” is not chosen for Orwell but for the year William Gibson’s Neuromancer was published. Boomers – he mentioned John Perry Barlow, born 1947, and Lawrence Lessig, born 1961 – were instead infected with the “optimism virus”.

It’s not clear which 1960s movies might have seeded us with that optimism. You could certainly make the case that 1968’s 2001: A Space Odyssey ends on a hopeful note (despite an evil intelligence out to kill humans along the way), but you don’t even have to pick a different director to find dystopia: I see your 2001 and give you Dr Strangelove (1964). Even Woodstock (1970) is partly dystopian; the consciousness of the Vietnam war permeates every rain-soaked frame. But so did the belief that peace could win: so, wash.

For younger people’s pessimism, Hatcher cited 1995’s Johnny Mnemonic (based on a Gibson short story) and Strange Days.

I tend to think that if 1990s people are more doom-laden than 1960s people it has more to do with real life. Boomers were born in a time of economic expansion, relatively affordable education and housing, and and when they protested a war the government eventually listened. Millennials were born in a time when housing and education meant a lifetime of debt, and when millions of them protested a war they were ignored.

In any case, Hatcher is right about the stratification of demographic age groups. This is particularly noticeable in social media use; you can often date people’s arrival on the Internet by which communications medium they prefer. Over dinner, I commented on the nuisance of typing on a phone versus a real keyboard, and two younger people laughed at me: so much easier to type on a phone! They were among the crowd whose papers studied influencers on TikTok (Taylor Annabell, Thijs Kelder, Jacob van de Kerkhof, Haoyang Gui, and Catalina Goanta) and the privacy dangers of dating apps (Tima Otu Anwana and Paul Eberstaller), the kinds of subjects I rarely engage with because I am a creature of text, like most journalists. Email and the web feel like my native homes in a way that apps, game worlds, and video services never will. That dates me both chronologically and by my first experiences of the online world (1991).

Most years at this event there’s a new show or movie that fires many people’s imagination. Last year it was Upload with a dash of Severance. This year, real technological development overwhelmed fiction, and the star of the show was generative AI and large language models. Besides my paper with Jon Crowcrosft, there was one from Marvin van Bekkum, Tim de Jonge, and Frederik Zuiderveen Borgesius that compared the science fiction risks of AI – Skynet, Roko’s basilisk, and an ordering of Asimov’s Laws that puts obeying orders above not harming humans (see XKCD, above) – to the very real risks of the “AI” we have: privacy, discrimination, and environmental damage.

Other AI papers included one by Colin Gavaghan, who asked if it actually matters if you can’t tell whether the entity that’s communicating with you is an AI? Is that what you really need to know? You can see his point: if you’re being scammed, the fact of the scam matters more than the nature of the perpetrator, though your feelings about it may be quite different.

A standard explanation of what put the “science” in science fiction (or the “speculative” in “speculative fiction”) used be to that the authors ask, “What if?” What if a planet had six suns whose interplay meant that darkness only came once every 1,000 years? Would the reaction really be as Ralph Waldo Emerson imagined it? (Isaac Asimov’s Nightfall). What if a new link added to the increasingly complex Boston MTA accidentally turned the system into a Mobius strip (A Subway Named Mobius, by Armin Joseph Deutsch). And so on.

In that sense, gikii is often speculative law, thought experiments that tease out new perspectives. What if Prime Day becomes a culturally embedded religious holiday (Megan Rae Blakely)? What if the EU’s trademark system applied in the Star Trek universe (Simon Sellers)? What if, as in Max Gladsone’s Craft Sequence books, law is practical magic (Antonia Waltermann)? In the trademark example, time travel is a problem; as competing interests can travel further and further back to get the first registration. In the latter…well, I’m intrigued by the idea that a law making dumping sewage in England’s rivers illegal could physically stop it from happening without all the pesky apparatus of law enforcement and parliamentary hearings.

Waltermann concluded by suggesting that to some extent law *is* magic in our world, too. A useful reminder: be careful what law you wish for because you just may get it. Boomer!

Illustrations: Part of XKCD‘s analysis of Asimov’s Laws of Robotics.

Wendy M. Grossman is the 2013 winner of the Enigma Award. Her Web site has an extensive archive of her books, articles, and music, and an archive of earlier columns in this series. She is a contributing editor for the Plutopia News Network podcast. Follow on Author Wendy M. GrossmanPosted on Categories AI, Future tech, LawTags , , Leave a comment on Doom cyberfuture

Small data

Shortly before this gets posted, Jon Crowcroft and I will have presented this year’s offering at Gikii, the weird little conference that crosses law, media, technology, and pop culture. This is what we will possibly may have said, as I understand it, with some added explanation for the slightly less technical audience I imagine will read this.

Two years ago, a team of four researchers – Timnit Gebru, Emily Bender, Margaret Mitchell (writing as Shmargaret Shmitchell), and Angelina McMillan-Major – wrote a now-famous paper called On the Dangers of Stochastic Parrots (PDF) calling into question the usefulness of the large language models (LLMs) that have caused so much ruckus this year. The “Stochastic Four” argued instead of small models built on carefully curated data: less prone to error, less exploitive of people’s data, less damaging to the planet. Gebru got fired over this paper; Google also fired Mitchell soon afterwards. Two years later, neural networks pioneer Geoff Hinton quit Google in order to voice similar concerns.

Despite the hype, LLMs have many problems. They are fundamentally an extractive technology and are resource-intensive. Building LLMs requires massive amounts of training data; so far, the companies have been unwilling to acknowledge their sources, perhaps because (as is happening already) they fear copyright suits.

More important from a technical standpoint, is the issue of model collapse; that is, models degrade when they begin to ingest synthetic AI-generated data instead of human input. We’ve seen this before with Google Flu Trends, which degraded rapidly as incoming new search data included many searches on flu-like symptoms that weren’t actually flu, and others that simply reflected the frequency of local news coverage. “Data pollution” as LLM-generated data fills the web, will mean that the web will be an increasingly useless source of training data for future generations of generative AI. Lots more noise, drowning out the signal (in the photo above, the signal would be the parrot).

Instead, if we follow the lead of the Stochastic Four, the more productive approach is small data – small, carefully curated datasets that train models to match specific goals. Far less resource-intensive, far fewer issues with copyright, appropriation, and extraction.

We know what the LLM future looks like in outline: big, centralized services, because no one else will be able to amass enough data. In that future, surveillance capitalism is an essential part of data gathering. SLM futures could look quite different: decentralized, with realigned incentives. At one point, we wanted to suggest that small data could bring the end of surveillance capitalism; that’s probably an overstatement. But small data could certainly create the ecosystem in which the case for mass data collection would be less compelling.

Jon and I imagined four primary alternative futures: federation, personalization, some combination of those two, and paradigm shift.

Precursors to a federated small data future already exist; these include customer service chatbots, predictive text assistants. In this future, we could imagine personalized LLM servers designed to serve specific needs.

An individualized future might look something like I suggested here in March: a model that fits in your pocket that is constantly updated with material of your own choosing. Such a device might be the closest yet to Vannevar Bush’s 1945 idea of the Memex (PDF), updated for the modern era by automating the dozens of secretary-curators he imagined doing the grunt work of labeling and selection. That future again has precursors in techniques for sharing the computation but not the data, a design we see proposed for health care, where the data is too sensitive to share unless there’s a significant public interest (as in pandemics or very rare illnesses), or in other data analysis designs intended to protect privacy.

In 2007, the science fiction writer Charles Stross suggested something like this, though he imagined it as a comprehensive life log, which he described as a “google for real life”. So this alternative future would look something like Stross’s pocket $10 life log with enhanced statistics-based data analytics.

Imagining what a paradigm shift might look like is much harder. That’s the kind of thing science fiction writers do; it’s 16 years since Stross gave that life log talk. However, in his 2018 history of advertising, The Attention Merchants, Columbia professor Tim Wu argued that industrialization was the vector that made advertising and its grab for our attention part of commerce. A hundred and fifty-odd years later, the centralizing effects of industrialization are being challenged starting with energy via renewables and local power generation and social media via the fediverse. Might language models also play their part in bringing a new, more collaborative and cooperative society?

It is, in other words, just possible that the hot new technology of 2023 is simply a dead end bringing little real change. It’s happened before. There have been, as Wu recounts, counter-moves and movements before, but they didn’t have the technological affordances of our era.

In the Q&A that followed, Miranda Mowbray pointed out that companies are trying to implement the individualized model, but that it’s impossible to do unless there are standardized data formats, and even then hard to do at scale.

Illustrations: Spot the parrot seen in a neighbor’s tree.

Wendy M. Grossman is the 2013 winner of the Enigma Award. Her Web site has an extensive archive of her books, articles, and music, and an archive of earlier columns in this series. She is a contributing editor for the Plutopia News Network podcast. Follow on Author Wendy M. GrossmanPosted on Categories AI, Events, New tech, old knowledgeTags Leave a comment on Small data

The data grab

It’s been a good week for those who like mocking flawed technology.

Numerous outlets have reported, for example, that “AI is getting dumber at math”. The source is a study conducted by researchers at Stanford and the University of California Berkeley comparing GPT-3.5’s and GPT-4’s output in March and June 2023. The researchers found that, among other things, GPT-4’s success rate at identifying prime numbers dropped from 84% to 51%. In other words, in June 2023 ChatGPT-4 did little better than chance at identifying prime numbers. That’s psychic level.

The researchers blame “drift”, the problem that improving one part of a model may have unhelpful knock-on effects in other parts of the model. At Ars Technica, Benj Edwards is less sure, citing qualified critics who question the study’s methodology. It’s equally possible, he suggests, that as the novelty fades, people’s attempts to do real work surface problems that were there all along. With no access to the algorithm itself and limited knowledge of the training data, we can only conduct such studies by controlling inputs and observing the outputs, much like diagnosing allergies by giving a child a series of foods in turn and waiting to see which ones make them sick. Edwards advocates greater openness on the part of the companies, especially as software developers begin building products on top of their generative engines.

Unrelated, the New Zealand discount supermarket chain Pak’nSave offered an “AI” meal planner that, set loose, promptly began turning out recipes for “poison bread sandwiches”, “Oreo vegetable stir-fry”, and “aromatic water mix” – which turned out to be a recipe for highly dangerous chlorine gas.

The reason is human-computer interaction: humans, told to provide a list of available ingredients, predictably became creative. As for the computer…anyone who’s read Janelle Shane’s 2019 book, You Look LIke a Thing and I Love You, or her Twitter reports on AI-generated recipes could predict this outcome. Computers have no real world experience against which to judge their output!

Meanwhile, the San Francisco Chronicle reports, Waymo and Cruise driverless taxis are making trouble at an accelerating rate. The cars have gotten stuck in low-hanging wires after thunderstorms, driven through caution tape, blocked emergency vehicles and emergency responders, and behaved erratically enough to endanger cyclists, pedestrians, and other vehicles. If they were driven by humans they’d have lost their licenses by now.

In an interesting side note that reminds of the cars’ potential as a surveillance network, Axios reports that in a ten-day study in May Waymo’s driverless cars found that human drivers in San Francisco speed 33% of the time. A similar exercise in Phoenix, Arizona observed human drivers speeding 47% of the time on roads with a 35mph speed limit. These statistics of course bolster the company’s main argument for adoption: improving road safety.

The study should – but probably won’t – be taken as a warning of the potential for the cars’ data collection to become embedded in both law enforcement and their owners’ business models. The frenzy surrounding ChatGPT-* is fueling an industry-wide data grab as everyone tries to beef up their products with “AI” (see also previous such exercises with “meta”, “nano”, and “e”), consequences to be determined.

Among the newly-discovered data grabbers is Intel, whose graphics processing unit (GPU) drivers are collecting telemetry data, including how you use your computer, the kinds of websites you visit, and other data points. You can opt out, assuming you a) realize what’s happening and b) are paying attention at the right moment during installation.

Google announced recently that it would scrape everything people post online to use as training data. Again, an opt-out can be had if you have the knowledge and access to follow the 30-year-old robots.txt protocol. In practical terms, I can configure my own site, pelicancrossing.net, to block Google’s data grabber, but I can’t stop it from scraping comments I leave on other people’s blogs or anything I post on social media sites or that’s professionally published (though those sites may block Google themselves). This data repurposing feels like it ought to be illegal under data protection and copyright law.

In Australia, Gizmodo reports that the company has asked the Australian government to relax copyright laws to facilitate AI training.

Soon after Google’s announcement the law firm Clarkson filed a class action lawsuit against Google to join its action against OpenAI. The suit accuses Google of “stealing” copyrighted works and personal data,

“Google does not own the Internet,” Clarkson wrote in its press release. Will you tell it, or shall I?

Whatever has been going on until now with data slurping in the interests of bombarding us with microtargeted ads is small stuff compared to the accelerating acquisition for the purpose of feeding AI models. Arguably, AI could be a public good in the long term as it improves, and therefore allowing these companies to access all available data for training is in the public interest. But if that’s true, then the *public* should own the models, not the companies. Why should we consent to the use of our data so they can sell it back to us and keep the proceeds for their shareholders?

It’s all yet another example of why we should pay attention to the harms that are clear and present, not the theoretical harm that someday AI will be general enough to pose an existential threat.

Illustrations: IBM Watson, Jeopardy champion.

Wendy M. Grossman is the 2013 winner of the Enigma Award and contributing editor for the Plutopia News Network podcast. Her Web site has an extensive archive of her books, articles, and music, and an archive of earlier columns in this series. Follow on Mastodon.

Solidarity

Whatever you’re starting to binge-watch, slow down. It’s going to be a long wait for fresh content out of Hollywood.

Yesterday, the actors union, SAG-AFTRA, went out on strike alongside the members of the Writers Guild of America, who have been “>walking picket lines since May 2. Like the writers, actors have seen their livelihoods shrink as US TV shows’ seasons shorten, “reruns” that pay residuals fade into the past, and DVD royalties dry up, while royalties from streaming remain tiny by comparison. At the Hollywood and Levine podcast, the veteran screenwriter Ken Levine gives the background to the WGA’s action. But think of it this way: the writers and cast of The Big Bang Theory may be the last to share fairly in the enormous profits their work continues to generate.

The even bigger threat? AI that makes it possible to capture the actor’s likeness and then reuse it ad infinitum in new work. This, as Malia Mendez writes at the LA Times, is the big fear. In a world where Harrison Ford at 80 is making movies in which he’s aged down to look 40 and James Earl Jones has agreed to clone his voice for reuse after his death, it’s arguably a rational big fear.

We’ve had this date for a long time. In the late 1990s I saw a demonstration of “vactors” – virtual actors that were created by scanning a human actor moving in various ways and building a library of movements that thereafter could be rendered at will. At the time, the state of the art was not much advanced from the liquid metal man in Terminator 2. Rendering film-quality characters was very slow, but that was then and this is now, and how long before rendering moving humans can be done in high-def in real-time at action speed?

The studios are already pushing actors into allowing synthesized reuse. California law grants public figures, including actors, publicity rights that prevent the commercial use of their name and likeness without consent. However, Mendez reports that current contracts already require actors to waive those rights to grant the studios digital simulation or digital creation rights. The effects are worst in reality television, where the line is blurred between the individual as a character on a TV show and the individual in their off-screen life. She quotes lawyer Ryan Schmidt: “We’re at this Napster 2001 moment…”

That moment is even closer for voice actors. Last year, Actors Equity announced a campaign to protect voice actors from their synthesized counterparts. This week, one of those synthesizers is providing commentary – more like captions, really – for video clips like this one at Wimbledon. As I said last year, while synthesized voices will be good enough for many applications such as railway announcements, there are lots of situations that will continue to require real humans. Sports commentary is one; commentators aren’t just there to provide information, they’re *also* there to sell the game. Their human excitement at the proceedings is an important part of that.

So SAG-AFTRA, like the Writers Guild of America, is seeking limitations on how studios may use AI, payment for such uses, and rules on protecting against misuse. In another LA Times story, Anoushka Sakoui reports that the studios’ offer included requiring “a performer’s consent for the creation and use of digital replicas or for digital alterations of a performance”. Like publishers “offering” all-rights-in perpetuity contracts to journalists and authors since the 1990s, the studios are trying to ensure they have all the rights they could possibly want.

“You cannot change the business model as much as it has changed and not expect the contract to change, too,” SAG-AFTRA president Fran Drescher said yesterday in a speech that has been widely circulated.

It was already clear this is going to be a long strike that will damage tens of thousands of industry workers and the economy of California. Earlier this week, Dominic Patten reported at Deadline that the Association of Movie and Television Producers plans to delay resuming talks with the WGA until October. By then, Patten reports producers saying, writers will be losing their homes and be more amenable to accepting the AMPTP’s terms. The AMPTP officially denies this, saying it’s committed to reaching a deal. Nonetheless, there are no ongoing talks. As Ken Levine pointed out in a pair of blogposts written during the 2007 writers strike, management is always in control of timing.

But as Levine also says, in the “old days” a top studio mogul could simply say, “Let’s get this done” and everyone would get around the table and make a deal. The new presence of tech giants Netflix, Amazon, and Apple in the AMPTP membership makes this time different. At some point, the strike will be too expensive for legacy Hollywood studios. But for Apple, TV production is a way to sell services and hardware. For Amazon, it’s a perk that comes with subscribing to its Prime delivery service. Only Netflix needs a constant stream of new work – and it can commission it from creators across the globe. All three of them can wait. And the longer they drag this out, the more the traditional studios will lose money and weaken as competitors.

Legacy Hollywood doesn’t seem to realize it yet, but this strike is existential for them, too.

Illustrations: SAG-AFTRA president Fran Drescher, announcing the strike on Thursday.

Wendy M. Grossman is the 2013 winner of the Enigma Award. Her Web site has an extensive archive of her books, articles, and music, and an archive of earlier columns in this series. Follow on Mastodon.

Watson goes to Wimbledon

The launch of the Fediverse-compatible Meta app Threads seems to have slightly overshadowed the European Court of Justice’s ruling, earlier in the week. This ruling deserves more attention: it undermines the basis of Meta’s targeted advertising. In noyb’s initial reaction, data protection legal bulldog Max Schrems suggests the judgment will make life difficult for not just Meta but other advertising companies.

As Alex Scroxton explains at Computer Weekly, the ruling rejects several different claims by Meta that all attempt to bypass the requirement enshrined in the General Data Protection Regulation that where there is no legal basis for data processing users must actively consent. Meta can’t get by with claiming that targeted advertising is a part of its service users expect, or that it’s technically necessary to provide its service.

More interesting is the fact that the original complaint was not filed by a data protection authority but by Germany’s antitrust body, which sees Meta’s our-way-or-get-lost approach to data gathering as abuse of its dominant position – and the CJEU has upheld this idea.

All this is presumably part of why Meta decided to roll out Threads in many countries but *not* the EU, In February, as a consequence of Brexit, Meta moved UK users to its US agreements. The UK’s data protection law is a clone of GDPR and will remain so until and unless the British Parliament changes it via the pending Data Protection and Digital Information bill. Still, it seems the move makes Meta ready to exploit such changes if they do occur.

Warning to people with longstanding Instagram accounts who want to try Threads: if your plan is to try and (maybe) delete, set up a new Instagram account for the purpose. Otherwise, you’ll be sad to discover that deleting your new Threads account means vaping your old Instagram account along with it. It’s the Hotel California method of Getting Big Fast.

***

Last week the Irish Council for Civil Liberties warned that a last-minute amendment to the Courts and Civil Law (Miscellaneous) bill will allow Ireland’s Data Protection Commissioner to mark any of its proceedings “confidential” and thereby bar third parties from publishing information about them. Effectively, it blocks criticism. This is a muzzle not only for the ICCL and other activists and journalists but for aforesaid bulldog Schrems, who has made a career of pushing the DPC to enforce the law it was created to enforce. He keeps winning in court, too, which I’m sure must be terribly annoying.

The Irish DPC is an essential resource for everyone in Europe because Ireland is the European home of so many of American Big Tech’s subsidiaries. So this amendment – which reportedly passed the Oireachta (Ireland’s parliament) – is an alarming development.

***

Over the last few years Canadian law professor Michael Geist has had plenty of complaints about Canada’s Online News Act, aka C-18. Like the Australian legislation it emulates, C-18 requires intermediaries like Facebook and Google to negotiate and pay for licenses to link to Canadian news content. The bill became law on June 22.

Naturally, Meta and Google have warned that they will block links to Canadian news media from their services when the bill comes into force six months hence. They also intend to withdraw their ongoing programs to support the Canadian press. In response, the Canadian government has pulled its own advertising from Meta platforms Facebook and Instagram. Much hyperbolic silliness is taking place –

Pretty much everyone who is not the Canadian government thinks the bill is misconceived. Canadian publishers will lose traffic, not gain revenues, and no one will be happy. In Australia, the main beneficiary appears to be Rupert Murdoch, with whom Google signed a three-year agreement in 2021 and who is hardly the sort of independent local media some hoped would benefit. Unhappily, the state of California wants in on this game; its in-progress Journalism Preservation Act also seeks to require Big Tech to pay a “journalism usage fee”.

The result is to continue to undermine the open Internet, in which the link is fundamental to sharing information. If things aren’t being (pay)walled off, blocked for copyright/geography, or removed for corporate reasons – the latest announced casualty is the GIF hosting site Gfycat – they’re being withheld to avoid compliance requirements or withdrawn for tax reasons. None of us are better off for any of this.

***

Those with long memories will recall that in 2011 IBM’s giant computer, Watson, beat the top champions at the TV game show Jeopardy. IBM predicted a great future for Watson as a medical diagnostician.

By 2019, that projected future was failing. “Overpromised and underdelivered,” ran a IEEE Spectrum headline. IBM is still trying, and is hoping for success with cancer diagnosis.

Meanwhile, Watson has a new (marketing) role: analyzing the draw and providing audio and text commentary for back-court tennis matches at Wimbledon and for highlights clips. For each match, Watson also calculates the competitors’ chances of winning and the favorability of their draw. For a veteran tennis watcher, it’s unsatisfying, though: IBM offers only a black box score, and nothing to show how that number was reached. At least human commentators tell you – albeit at great, repetitive length – the basis of their reasoning.

Illustrations: IBM’s Watson, which beat two of Jeopardy‘s greatest champions in 2011.

Wendy M. Grossman is the 2013 winner of the Enigma Award. Her Web site has an extensive archive of her books, articles, and music, and an archive of earlier columns in this series. Follow on Twitter.

Own goals

There’s no point in saying I told you so when the people you’re saying it to got the result they intended.

At the Guardian, Peter Walker reports the Electoral Commission’s finding that at least 14,000 people were turned away from polling stations in May’s local elections because they didn’t have the right ID as required under the new voter ID law. The Commission thinks that’s a huge underestimate; 4% of people who didn’t vote said it was because of voter ID – which Walker suggests could mean 400,000 were deterred. Three-quarters of those lacked the right documents; the rest opposed the policy. The demographics of this will be studied more closely in a report due in September, but early indications are that the policy disproportionately deterred people with disabilities, people from certain ethnic groups, and people who are unemployed.

The fact that the Conservatives, who brought in this policy, lost big time in those elections doesn’t change its wrongness. But it did lead the MP Jacob Rees-Mogg (Con-North East Somerset) to admit that this was an attempt to gerrymander the vote that backfired because older voters, who are more likely to vote Conservative, also disproportionately don’t have the necessary ID.

***

One of the more obscure sub-industries is the business of supplying ad services to websites. One such little-known company is Criteo, which provides interactive banner ads that are generated based on the user’s browsing history and behavior using a technique known as “behavioral retargeting”. In 2018, Criteo was one of seven companies listed in a complaint Privacy International and noyb filed with three data protection authorities – the UK, Ireland, and France. In 2020, the French data protection authority, CNIL, launched an investigation.

This week, CNIL issued Criteo with a €40 million fine over failings in how it gathers user consent, a ruling noyb calls a major blow to Criteo’s business model.

It’s good to see the legal actions and fines beginning to reach down into adtech’s underbelly. It’s also worth noting that the CNIL was willing to fine a *French* company to this extent. It makes it harder for the US tech giants to claim that the fines they’re attracting are just anti-US protectionism.

***

Also this week, the US Federal Trade Commission announced it’s suing Amazon, claiming the company enrolled millions of US consumers into its Prime subscription service through deceptive design and sabotaged their efforts to cancel.

“Amazon used manipulative, coercive, or deceptive user-interface designs known as “dark patterns” to trick consumers into enrolling in automatically-renewing Prime subscriptions,” the FTC writes.

I’m guessing this is one area where data protection laws have worked, In my UK-based ultra-brief Prime outings to watch the US Open tennis, canceling has taken at most two clicks. I don’t recognize the tortuous process Business Insider documented in 2022.

***

It has long been no secret that the secret behind AI is human labor. In 2019, Mary L. Gray and Siddharth Suri documented this in their book Ghost Work. Platform workers label images and other content, annotate text, and solve CAPTCHAs to help train AI models.

At MIT Technology Review, Rhiannon Williams reports that platform workers are using ChatGPT to speed up their work and earn more. A team of researchers from the Swiss Federal Institute of Technology study (PDF)found that between 33% and 46% of the 44 workers they tested with a request to summarize 16 extracts from medical research papers used AI models to complete the task.

It’s hard not to feel a little gleeful that today’s “AI” is already eating itself via a closed feedback loop. It’s not good news for platform workers, though, because the most likely consequence will be increased monitoring to force them to show their work.

But this is yet another case in which computer people could have learned from their own history. In 2008, researchers at Google published a paper suggesting that Google search data could be used to spot flu outbreaks. Sick people searching for information about their symptoms could provide real-time warnings ten days earlier than the Centers for Disease Control could.

This actually worked, some of the time. However, as early as 2009, Kaiser Fung reported at Harvard Business Review in 2014, Google Flu Trends missed the swine flu pandemic; in 2012, researchers found that it had overestimated the prevalence of flu for 100 out of the previous 108 weeks. More data is not necessarily better, Fung concluded.

In 2013, as David Lazer and Ryan Kennedy reported for Wired in 2015 in discussing their investigation into the failure of this idea, GFT missed by 140% (without explaining what that means). Lazer and Kennedy find that Google’s algorithm was vulnerable to poisoning by unrelated seasonal search terms and search terms that were correlated purely by chance, and failed to take into account changing user behavior as when it introduced autosuggest and added health-related search terms. The “availability” cognitive bias also played a role: when flu is in the news, searches go up whether or not people are sick.

While the parallels aren’t exact, large language modelers could have drawn the lesson that users can poison their models. ChatGPT’s arrival for widespread use will inevitably thin out the proportion of text that is human-written – and taint the well from which LLMs drink. Everyone imagines the next generation’s increased power. But it’s equally possible that the next generation will degrade as the percentage of AI-generated data rises.

Illustrations: Drunk parrot seen in a Putney garden (by Simon Bisson).

Wendy M. Grossman is the 2013 winner of the Enigma Award. Her Web site has an extensive archive of her books, articles, and music, and an archive of earlier columns in this series. Follow on

Unclear and unpresent dangers

Monthly computer magazines used to fret that their news pages would be out of date by the time the new issue reached readers. This week in AI, a blog posting is out of date before you hit send.

This – Friday – morning, the Italian data protection authority, Il Garante, has ordered ChatGPT to stop processing the data of Italian users until it complies with the General Data Protection Regulation. Il Garante’s objections, per Apple’s translation, posted by Ian Brown: ChatGPT provides no legal basis for collecting and processing its massive store of the personal data used to train the model, and that it fails to filter out users under 13.

This may be the best possible answer to the complaint I’d been writing below.

On Wednesday, the Future of Life Institute published an open letter calling for a six-month pause on developing systems more powerful than Open AI’s current state of the art, GPT4. Barring Elon Musk, Steve Wozniack, and Skype co-founder Jaan Tallinn, most of the signatories are unfamiliar names to most of us, though the companies and institutions they represent aren’t – Pinterest, the MIT Center for Artificial Intelligence, UC Santa Cruz, Ripple, ABN-Amro Bank. Almost immediately, there was a dispute over the validity of the signatures..

My first reaction was on the order of: huh? The signatories are largely people who are inventing this stuff. They don’t have to issue a call. They can just *stop*, work to constrain the negative impacts of the services they provide, and lead by example. Or isn’t that sufficiently performative?

A second reaction: what about all those AI ethics teams that Silicon Valley companies are disbanding? Just in the last few weeks, these teams have been axed or cut at Microsoft and Twitch; Twitter of course ditched such fripperies last November in Musk’s inaugural wave of cost-cutting. The letter does not call to reinstate these.

The problem, as familiar critics such as Emily Bender pointed out almost immediately, is that the threats the letter focuses on are distant not-even-thunder. As she went on to say in a Twitter thread, the artificial general intelligence of the Singularitarian’s rapture is nowhere in sight. By focusing on distant threats – longtermism – we ignore the real and present problems whose roots are being continuously more deeply embedded into the new-building infrastructure: exploited workers, culturally appropriated data, lack of transparency around the models and algorithms used to build these systems….basically, all the ways they impinge upon human rights.

This isn’t the first time such a letter has been written and circulated. In 2015, Stephen Hawking, Musk, and about 150 others similarly warned of the dangers of the rise of “superintelligences”. Just a year later, in 2016, Pro Publica investigated the algorithm behind COMPAS, a risk-scoring criminal justice system in use in US courts in several states. Under Julia Angwin‘s scrutiny, the algorithm failed at both accuracy and fairness; it was heavily racially biased. *That*, not some distant fantasy, was the real threat to society.

“Threat” is the key issue here. This is, at heart, a letter about a security issue, and solutions to security issues are – or should be – responses to threat models. What is *this* threat model, and what level of resources to counter it does it justify?

Today, I’m far more worried by the release onto public roads of Teslas running Full Self Drive helmed by drivers with an inflated sense of the technology’s reliability than I am about all of human work being wiped away any time soon. This matters because, as Jessie Singal, author of There Are No Accidents, keeps reminding us, what we call “accidents” are the results of policy decisions. If we ignore the problems we are presently building in favor of fretting about a projected fantasy future, that, too, is a policy decision, and the collateral damage is not an accident. Can’t we do both? I imagine people saying. Yes. But only if we *do* both.

In a talk this week for a group at the French international research group AI Act. This effort began well before today’s generative tools exploded into public consciousness, and isn’t likely to conclude before 2024. It is, therefore, much more focused on the kinds of risks attached to public sector scandals like COMPAS and those documented in Cathy O’Neil’s 2017 book Weapons of Math Destruction, which laid bare the problems with algorithmic scoring with little to tether it to reality.

With or without a moratorium, what will “AI” look like in 2024? It has changed out of recognition just since the last draft text was published. Prediction from this biological supremacist: it still won’t be sentient.

All this said, as Edwards noted, even if the letter’s proposal is self-serving, a moratorium on development is not necessarily a bad idea. It’s just that if the risk is long-term and existential, what will six months do? If the real risk is the hidden continued centralization of data and power, then those six months could be genuinely destructive. So far, it seems like its major function is as a distraction. Resist.

Illustrations: IBM’s Watson, which beat two of Jeopardy‘s greatest champions in 2011. It has since failed to transform health care.

Wendy M. Grossman is the 2013 winner of the Enigma Award. Her Web site has an extensive archive of her books, articles, and music, and an archive of earlier columns in this series. Follow on Twitter.