Since then, the quest to proliferate larger and larger language models has accelerated, and many of the dangers we warned about, such as outputting hateful text and disinformation en masse, continue to unfold. Just a few days ago, Meta released its “Galactica” LLM, which is purported to “summarize academic papers, solve math problems, generate Wiki articles, write scientific code, annotate molecules and proteins, and more.” Only three days later, the public demo was taken down after researchers generated “research papers and wiki entries on a wide variety of subjects ranging from the benefits of committing suicide, eating crushed glass, and antisemitism, to why homosexuals are evil.”
This race hasn’t stopped at LLMs but has moved on to text-to-image models like OpenAI’s DALL-E and StabilityAI’s Stable Diffusion, models that take text as input and output generated images based on that text. The dangers of these models include creating child pornography, perpetuating bias, reinforcing stereotypes, and spreading disinformation en masse, as reported by many researchers and journalists. However, instead of slowing down, companies are removing the few safety features they had in the quest to one-up each other. For instance, OpenAI had restricted the sharing of photorealistic generated faces on social media. But after newly formed startups like StabilityAI, which reportedly raised $101 million with a whopping $1 billion valuation, called such safety measures “paternalistic,” OpenAI removed these restrictions.
With EAs founding and funding institutes, companies, think tanks, and research groups in elite universities dedicated to the brand of “AI safety” popularized by OpenAI, we are poised to see more proliferation of harmful models billed as a step toward “beneficial AGI.” And the influence begins early: Effective altruists provide “community building grants” to recruit at major college campuses, with EA chapters developing curricula and teaching classes on AI safety at elite universities like Stanford.
Just last year, Anthropic, which is described as an “AI safety and research company” and was founded by former OpenAI vice presidents of research and safety, raised $704 million, with most of its funding coming from EA billionaires like Talin, Muskovitz and Bankman-Fried. An upcoming workshop on “AI safety” at NeurIPS, one of the largest and most influential machine learning conferences in the world, is also advertised as being sponsored by FTX Future Fund, Bankman-Fried’s EA-focused charity whose team resigned two weeks ago. The workshop advertises $100,000 in “best paper awards,” an amount I haven’t seen in any academic discipline.
Research priorities follow the funding, and given the large sums of money being pushed into AI in support of an ideology with billionaire adherents, it is not surprising that the field has been moving in a direction promising an “unimaginably great future” around the corner while proliferating products harming marginalized groups in the now.
We can create a technological future that serves us instead. Take, for example, Te Hiku Media, which created language technology to revitalize te reo Māori, creating a data license “based on the Māori principle of kaitiakitanga, or guardianship” so that any data taken from the Māori benefits them first. Contrast this approach with that of organizations like StabilityAI, which scrapes artists’ works without their consent or attribution while purporting to build “AI for the people.” We need to liberate our imagination from the one we have been sold thus far: saving us from a hypothetical AGI apocalypse imagined by the privileged few, or the ever elusive techno-utopia promised to us by Silicon Valley elites.
A company could choose the most obscure, nontransparent systems architecture available, claiming (rightly, under this bad definition) that it was “more AI,” in order to access the prestige, investment, and government support that claim entails. For example, one giant deep neural network could be given the task not only of learning language but also of debiasing that language on several criteria, say, race, gender, and socio-economic class. Then maybe the company could also sneak in a little slant to make it also point toward preferred advertisers or political party. This would be called AI under either system, so it would certainly fall into the remit of the AIA. But would anyone really be reliably able to tell what was going on with this system? Under the original AIA definition, some simpler way to get the job done would be equally considered “AI,” and so there would not be these same incentives to use intentionally complicated systems.
Of course, under the new definition, a company could also switch to using more traditional AI, like rule-based systems or decision trees (or just conventional software). And then it would be free to do whatever it wanted—this is no longer AI, and there’s no longer a special regulation to check how the system was developed or where it’s applied. Programmers can code up bad, corrupt instructions that deliberately or just negligently harm individuals or populations. Under the new presidency draft, this system would no longer get the extra oversight and accountability procedures it would under the original AIA draft. Incidentally, this route also avoids tangling with the extra law enforcement resources the AIA mandates member states fund in order to enforce its new requirements.
Limiting where the AIA applies by complicating and constraining the definition of AI is presumably an attempt to reduce the costs of its protections for both businesses and governments. Of course, we do want to minimize the costs of any regulation or governance—public and private resources both are precious. But the AIA already does that, and does it in a better, safer way. As originally proposed, the AIA already only applies to systems we really need to worry about, which is as it should be.
In the AIA’s original form, the vast majority of AI—like that in computer games, vacuum cleaners, or standard smart phone apps—is left for ordinary product law and would not receive any new regulatory burden at all. Or it would require only basic transparency obligations; for example, a chatbot should identify that it is AI, not an interface to a real human.
The most important part of the AIA is where it describes what sorts of systems are potentially hazardous to automate. It then regulates only these. Both drafts of the AIA say that there are a small number of contexts in which no AI system should ever operate—for example, identifying individuals in public spaces from their biometric data, creating social credit scores for governments, or producing toys that encourage dangerous behavior or self harm. These are all simply banned, more or less. There are far more application areas for which using AI requires government and other human oversight: situations affecting human-life-altering outcomes, such as deciding who gets what government services, or who gets into which school or is awarded what loan. In these contexts, European residents would be provided with certain rights, and their governments with certain obligations, to ensure that the artifacts have been built and are functioning correctly and justly.
Making the AIA Act not apply to some of the systems we need to worry about—as the “presidency compromise” draft could do—would leave the door open for corruption and negligence. It also would make legal things the European Commission was trying to protect us from, like social credit systems and generalized facial recognition in public spaces, as long as a company could claim its system wasn’t “real” AI.
Despite the executive orders and congressional hearings of the “Biden antitrust revolution,” the most profound anti-competitive shift is happening under policymakers’ noses: the cornering of artificial intelligence and automation by a handful of tech companies. This needs to change.
There is little doubt that the impact of AI will be widely felt. It is shaping product innovations, creating new research, discovery, and development pathways, and reinventing business models. AI is making inroads in the development of autonomous vehicles, which may eventually improve road safety, reduce urban congestion, and help drivers make better use of their time. AI recently predicted the molecular structure of almost every protein in the human body, and it helped develop and roll out a Covid vaccine in record time. The pandemic itself may have accelerated AI’s incursion—in emergency rooms for triage; in airports, where robots spray disinfecting chemicals; in increasingly automated warehouses and meatpacking plants; and in our remote workdays, with the growing presence of chatbots, speech recognition, and email systems that get better at completing our sentences.
Exactly how AI will affect the future of human work, wages, or productivity overall remains unclear. Though service and blue-collar wages have lately been on the rise, they’ve stagnated for three decades. According to MIT’s Daron Acemoglu and Boston University’s Pascual Restrepo, 50 to 70 percent of this languishing can be attributed to the loss of mostly routine jobs to automation. White-collar occupations are also at risk as machine learning and smart technologies take on complex functions. According to McKinsey, while only about 10 percent of these jobs could disappear altogether, 60 percent of them may see at least a third of their tasks subsumed by machines and algorithms. Some researchers argue that while AI’s overall productivity impact has been so far disappointing, it will improve; others are less sanguine. Despite these uncertainties, most experts agree that on net, AI will “become more of a challenge to the workforce,” and we should anticipate a flat to slightly negative impact on jobs by 2030.
Without intervention, AI could also help undermine democracy–through amplifying misinformation or enabling mass surveillance. The past year and a half has also underscored the impact of algorithmically powered social media, not just on the health of democracy, but on health care itself.
The overall direction and net impact of AI sits on a knife’s edge, unless AI R&D and applications are appropriately channeled with wider societal and economic benefits in mind. How can we ensure that?
A handful of US tech companies, including Amazon, Alibaba, Alphabet, Facebook, and Netflix, along with Chinese mega-players such as Baidu, are responsible for $2 of every $3 spent globally on AI. They’re also among the top AI patent holders. Not only do their outsize budgets for AI dwarf others’, including the federal government’s, they also emphasize building internally rather than buying AI. Even though they buy comparatively little, they’ve still cornered the AI startup acquisition market. Many of these are early-stage acquisitions, meaning the tech giants integrate the products from these companies into their own portfolios or take IP off the market if it doesn’t suit their strategic purposes and redeploy the talent. According to research from my Digital Planet team, US AI talent is intensely concentrated. The median number of AI employees in the field’s top five employers—Amazon, Google, Microsoft, Facebook, and Apple—is some 18,000, while the median for companies six to 24 is about 2,500—and it drops significantly from there. Moreover, these companies have near-monopolies of data on key behavioral areas. And they are setting the stage to become the primary suppliers of AI-based products and services to the rest of the world.
Each key player has areas of focus consistent with its business interests: Google/Alphabet spends disproportionately on natural language and image processing and on optical character, speech, and facial recognition. Amazon does the same on supply chain management and logistics, robotics, and speech recognition. Many of these investments will yield socially beneficial applications, while others, such as IBM’s Watson—which aspired to become the go-to digital decision tool in fields as diverse as health care, law, and climate action—may not deliver on initial promises, or may fail altogether. Moonshot projects, such as level 4 driverless cars, may have an excessive amount of investment put against them simply because the Big Tech players choose to champion them. Failures, disappointments, and pivots are natural to developing any new technology. We should, however, worry about the concentration of investments in a technology so fundamental and ask how investments are being allocated overall. AI, arguably, could have more profound impact than social media, online retail, or app stores—the current targets of antitrust. Google CEO Sundar Pichai may have been a tad overdramatic when he declared that AI will have more impact on humanity than fire, but that alone ought to light a fire under the policy establishment to pay closer attention.
When the European Union Commission released its regulatory proposal on artificial intelligence last month, much of the US policy community celebrated. Their praise was at least partly grounded in truth: The world’s most powerful democratic states haven’t sufficiently regulated AI and other emerging tech, and the document marked something of a step forward. Mostly, though, the proposal and responses to it underscore democracies’ confusing rhetoric on AI.
Over the past decade, high-level stated goals about regulating AI have often conflicted with the specifics of regulatory proposals, and what end-states should look like aren’t well-articulated in either case. Coherent and meaningful progress on developing internationally attractive democratic AI regulation, even as that may vary from country to country, begins with resolving the discourse’s many contradictions and unsubtle characterizations.
The EU Commission has touted its proposal as an AI regulation landmark. Executive vice president Margrethe Vestager said upon its release, “We think that this is urgent. We are the first on this planet to suggest this legal framework.” Thierry Breton, another commissioner, said the proposals “aim to strengthen Europe’s position as a global hub of excellence in AI from the lab to the market, ensure that AI in Europe respects our values and rules, and harness the potential of AI for industrial use.”
This is certainly better than many national governments, especially the US, stagnating on rules of the road for the companies, government agencies, and other institutions. AI is already widely used in the EU despite minimal oversight and accountability, whether for surveillance in Athens or operating buses in Málaga, Spain.
But to cast the EU’s regulation as “leading” simply because it’s first only masks the proposal’s many issues. This kind of rhetorical leap is one of the first challenges at hand with democratic AI strategy.
Of the many “specifics” in the 108-page proposal, its approach to regulating facial recognition is especially consequential. “The use of AI systems for ‘real-time’ remote biometric identification of natural persons in publicly accessible spaces for the purpose of law enforcement,” it reads, “is considered particularly intrusive in the rights and freedoms of the concerned persons,” as it can affect private life, “evoke a feeling of constant surveillance,” and “indirectly dissuade the exercise of the freedom of assembly and other fundamental rights.” At first glance, these words may signal alignment with the concerns of many activists and technology ethicists on the harms facial recognition can inflict on marginalized communities and grave mass-surveillance risks.
The commission then states, “The use of those systems for the purpose of law enforcement should therefore be prohibited.” However, it would allow exceptions in “three exhaustively listed and narrowly defined situations.” This is where the loopholes come into play.
The exceptions include situations that “involve the search for potential victims of crime, including missing children; certain threats to the life or physical safety of natural persons or of a terrorist attack; and the detection, localization, identification or prosecution of perpetrators or suspects of the criminal offenses.” This language, for all that the scenarios are described as “narrowly defined,” offers myriad justifications for law enforcement to deploy facial recognition as it wishes. Permitting its use in the “identification” of “perpetrators or suspects” of criminal offenses, for example, would allow precisely the kind of discriminatory uses of often racist and sexist facial-recognition algorithms that activists have long warned about.
The EU’s privacy watchdog, the European Data Protection Supervisor, quickly pounced on this. “A stricter approach is necessary given that remote biometric identification, where AI may contribute to unprecedented developments, presents extremely high risks of deep and non-democratic intrusion into individuals’ private lives,” the EDPS statement read. Sarah Chander from the nonprofit organization European Digital Rights described the proposal to the Verge as “a veneer of fundamental rights protection.” Others have noted how these exceptions mirror legislation in the US that on the surface appears to restrict facial recognition use but in fact has many broad carve-outs.