Reckless Regulators - Lux Capital

Why are politicians so frightened by AI?

Regulators want to kill the future of artificial intelligence. That’s my only conclusion after a blistering week of global regulatory attacks.

It started on Monday with the long-awaited publication of President Joe Biden’s executive order on “the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence.” At just under 20,000 words, it lays out a gargantuan, whole-of-government effort that “establishes new standards for AI safety and security, protects Americans’ privacy, advances equity and civil rights, stands up for consumers and workers, promotes innovation and competition, advances American leadership around the world, and more.”

“And more.” Not unlike Ronald Reagan’s scariest nine words in the English language (“I’m from the government, and I’m here to help”), we should always be alarmed when regulators start talking about “and more.” But a lot more is what we got this week.

Across the ever-warming pond in the United Kingdom, Prime Minister Rishi Sunak held an AI Safety Summit at Bletchley Park, the famed research center where Alan Turing and a motley band of early cryptographers and computer scientists broke Nazi Germany’s encryption systems and helped Britain and the Allies win World War II.

The highlight of the summit was the “Bletchley Declaration,” which was signed by 28 countries and the European Union and argued that “the protection of human rights, transparency and explainability, fairness, accountability, regulation, safety, appropriate human oversight, ethics, bias mitigation, privacy and data protection needs to be addressed.” It was heart-warming to see China, Saudi Arabia, Nigeria and the United Arab Emirates sign the declaration — if only they would address “human rights,” “transparency,” “accountability,” and “ethics” in their own activities.

But I digress. Hopping the Channel, the European Union made progress on its comprehensive AI Act, legislation that has been debated for two years but is inching every closer to final passage and implementation. The EU’s regulatory model — increasingly popular globally — would divide AI into four risk categories from minimal to unacceptable, and place increasingly onerous testing and security requirements as the risk increases. Problematically, most applications are considered “high risk”, including potentially recommendation systems and social media algorithms. Unacceptable risks are forbidden, so high risk uses are the most regulated within the system.

Finally, former Google CEO Eric Schmidt and DeepMind co-founder Mustafa Suleyman penned an op-ed in The Financial Times two weeks ago calling for an “IPCC for AI.” The goal, in tandem with the United Nations’s High-level Advisory Body on Artificial Intelligence, is to do for AI what the Intergovernmental Panel on Climate Change has done for warming temperatures, namely, scientifically adjudicate exactly what is happening with artificial intelligence and write consensus reports. The goal is to save us from AI as much as the IPCC has saved us from climate change.

All regulators (so far!) emphasize the positive prospects of AI and argue that they don’t want to proscribe the progress of this technology. From the beginning of Biden’s order, “ Artificial intelligence (AI) holds extraordinary potential for both promise and peril. Responsible AI use has the potential to help solve urgent challenges while making our world more prosperous, productive, innovative, and secure.” That parallels the opening line of the Bletchley Declaration, which says that “Artificial Intelligence (AI) presents enormous global opportunities: it has the potential to transform and enhance human wellbeing, peace and prosperity.”

Yet, all of these regulatory actions have come together in just the past few days. For a novel family of technologies that have barely entered the market (OpenAI’s ChatGPT was launched less than a year ago and Stable Diffusion was launched a little more than a year ago), it’s gobsmacking the crazed and craven alacrity by which politicians are attempting to globally regulate a technology we didn’t even know functioned a few months ago.

Why the intensity? From what I can surmise, it’s driven by an imaginative fear propelled by fictional Hollywood thrillers. Take this story in Time Magazine this week:

The issue of AI was seemingly inescapable for Biden. At Camp David one weekend, he relaxed by watching the Tom Cruise film “Mission: Impossible — Dead Reckoning Part One.” The film's villain is a sentient and rogue AI known as “the Entity” that sinks a submarine and kills its crew in the movie's opening minutes.

“If he hadn’t already been concerned about what could go wrong with AI before that movie, he saw plenty more to worry about,” said [deputy white house chief of staff **bruce reed**], who watched the film with the president.

Mission: Impossible! Emphasis on “impossible”! What’s frustrating with this pervasive fear is that while generative AI technologies offer us new and exciting capabilities, the evolution of AI is synonymous with the evolution of software code and the digitalization of human decision-making. This is the most slippery part of the regulatory regime that was activated this week: there’s not even a good definition of what artificial intelligence is.

Take Biden’s executive order, which says “The term ‘AI model’ means a component of an information system that implements AI technology and uses computational, statistical, or machine-learning techniques to produce outputs from a given set of inputs.” Math, in other words.

One of the sections of the order demands that the government “Provide clear guidance to landlords, Federal benefits programs, and federal contractors to keep AI algorithms from being used to exacerbate discrimination.” First, we know from books like Richard Rothstein’s The Color of Law: A Forgotten History of How Our Government Segregated America that humans have done a pretty good job of creating incredibly discriminatory housing systems, no computers or AI algorithms required.

But let’s worry about the human dimension later and just focus on the digital: where exactly is the line between dumb software code and artificial intelligence? In New York City, it’s common practice for landlords to adjudicate rental applicants with what’s dubbed the 40x formula: documented annual income must be 40 times the monthly rent of the apartment. Thus, an entry-level studio at $3,000 per month requires an annual income of $120,000. This formula is built into housing applications as a form of mechanical intelligence.

The irony, of course, is that the use of formulas and very basic “AI models” was pursued by landlords precisely because it warded off human discrimination (and limited lawsuits). Rather than having a biased human rent officer adjudicate an application, every applicant would instead be judged by an objective standard regardless of background. We actively pursued AI as a solution to human problems, and now regulators want to reverse course.

In fact, what’s jarring is just how many negative uses identified across this week’s regulatory actions already exist in law and how many have nothing to do with artificial intelligence. From the Biden order: “These principles and best practices will benefit workers by providing guidance to prevent employers from undercompensating workers, evaluating job applications unfairly, or impinging on workers’ ability to organize.”

What does any of this specifically have to do with artificial intelligence? AI overlords aren’t underpaying employees — human managers and corporate executives are. Unions aren’t busted by Terminator-style Pinkerton agents, but rather by … human Pinkerton agents. If we’re concerned about employee welfare, that goal should itself be the focus of regulatory and legislative action — not this histrionic focus on artificial intelligence.

Even in the national security realm, AI’s supposedly new capabilities are wildly oversold. For instance, there are widespread fears that AI will help to accelerate the development of next-generation bio and chemical weapons, and yet, as I wrote a year and a half ago before all of this hubbub in “AI, dual-use medicine, and bioweapons”:

While there are widespread fears of mad scientists inventing deadly contagions in hidden wetlabs in the caves of Waziristan, the reality is that the world is already familiar with incredibly viral and deadly pathogens. Ebola, as just one choice example, kills roughly half of anyone infected with a relatively high virality rate. As one former presidential advisor on bioweapons explained to me years ago, Mother Nature is quite efficient at producing terrifying bioweapons all on her own, no mad scientists required. Our public health response to a naturally-occurring pandemic and one that is man-made will be exactly the same.

We don’t need better weapons — the best possible weapons are already available today. Humanity was pretty good at finding ways to completely destroy itself well before AI ever arrived.

For such a nascent and unproven technology, it behooves us to tread much slower on regulation. We need to encourage widespread experimentation and openness in the development of the bleeding-edge of AI performance and capabilities. We should be encouraging the distribution of open-source AI models to as many scientists and institutions and users as possible. Everyone should have access to the best AI models humanity has ever crafted.

Why? Look at the incredible human progress made possible by the diffusion of powerful computation over the past few decades. What’s bizarre is that the computer has created trillions of dollars of wealth, improved the lives of billions of people, and allowed scientists and inventors unparalleled power to chart past the frontiers of humanity’s knowledge into the depths of the unknown.

Critically, we did all of that with minimal global regulation despite the dangerous uses of the computer. Imagine if regulators came together in the 1970s at the dawn of the personal computing revolution and the internet and attempted to manage the imaginative “risks” of these new devices? Imagine if we had locked in IBM, Burroughs, DEC, and other large incumbents by preventing new entrants like Apple and Microsoft from transforming and democratizing computation?

Artificial intelligence is yet another, more advanced form of computation. The existential risks from AI are no different than that of supercomputers crunching numbers the past few decades, because AI is just supercomputers crunching numbers. Better numbers, more numbers, but the same computations nonetheless.

Admittedly, there will be bad actors who use AI tools for evil. Some nations will equip their drones with killer AI technology that murders civilians. Terrorists will use AI to scan maps and visual data to identify physical weaknesses and exploit them. Landlords will mold AI to match the discrimination of tenants they are looking for in a building. Employers will use AI to optimize shift schedules and make the lives of employees even more unstable than before.

But here’s the thing: all of that was done before. That’s the permanent promise and peril of artificial intelligence, technology and humanity itself. We are capable of extraordinary leaps of knowledge and wisdom, of developing capabilities for self-actualization and expression that our ancestors can only dream of. But we are also capable of the most heinous and extraordinary acts of evil imaginable. The curse of humanity is how little our technologies ever seem to really influence our own moral decision-making. Instead of regulation, I want AI to flourish as far as it can — “and more.”

Lux Recommends

Ian Urbina, founder of non-profit The Outlaw Ocean Project, has an enrapturing investigative report on the overseas fishing fleet of China in The New Yorker. “The country now catches more than five billion pounds of seafood a year through distant-water fishing, the biggest portion of it squid. China’s seafood industry, which is estimated to be worth more than thirty-five billion dollars, accounts for a fifth of the international trade, and has helped create fifteen million jobs. … China’s fleet has also expanded the government’s international influence. The country has built scores of ports as part of its Belt and Road Initiative, a global infrastructure program that has, at times, made it the largest financier of development in South America, sub-Saharan Africa, and South Asia.”
Our “Securities” producer Chris Gates enjoyed a recent episode of the War on the Rocks podcast featuring Raj Shah of Shield Capital, who talks about his transition from F-16 pilot to the Pentagon and now into defensetech investing. Shield announced a $186 million inaugural fund last month.
Alexander Sammon wrote a complex yet human investigation into avocado production in Michoacán, Mexico for Harper’s. “In Cherán, however, there was no such violence. Nor were there any avocados. Twelve years ago, the town’s residents prevented corrupt officials and a local cartel from illegally cutting down native forests to make way for the crop. A group of locals took loggers hostage while others incinerated their trucks. Soon, townspeople had kicked out the police and local government, canceled elections, and locked down the whole area. A revolutionary experiment was under way.”
Tess Van Stekelenburg highlights the announcement from Isomorphic Labs about expanding the capabilities of AlphaFold, Google/DeepMind’s protein-folding model. The new updates allow for more precision and better predictions around "ligands (small molecules), proteins, nucleic acids (DNA and RNA), and those containing post-translational modifications (PTMs).”
In the China-watcher community, the article to read was Evan Osnos’s “China’s Age of Malaise” in The New Yorker, a plangent follow up to Osnos's best-selling 2014 book, Age of Ambition: Chasing Fortune, Truth, and Faith in the New China. “The space for pop culture, high culture, and spontaneous interaction has narrowed to a pinhole. Chinese social media, which once was a chaotic hive, has been tamed, as powerful voices are silenced and discussions closed. Pop concerts and other performances have been cancelled for reasons described only as ‘force majeure.’ Even standup comics are forced to submit videos of jokes for advance approval.”
Our scientist-in-residence Sam Arbesman enjoyed Jose M. Gilgado’s piece on “The beauty of finished software.” “Once you get used to the software, once the software works for you, you don’t need to learn anything new; the interface will exactly be the same, and all your files will stay relevant. No migrations, no new payments, no new changes.”
Finally, I have been enjoying Santi Ruiz’s newish newsletter, Statecraft, which explores policies that have actually delivered results. His best so far was on PEPFAR, in a piece entitled, "How to Save Twenty Million Lives.” “The interagency fights were blood on the floor. Hatchet work. It was the president that stopped all that. When you go meet people in the government one-on-one, they are the most dedicated, committed people. Most people don't go into government to fight over budgets, over turf. They go in and are not paid well and work their tails off because they want to make a difference, but the system doesn't give them much chance to make a difference. They get ground down and ultimately start fighting for budgets, protection, and turf. You get about 20% of people who are so jaded by the whole thing. 80% still wanna do something. If you don't offer the 80% something, the 20% win.”

That’s it, folks. Have questions, comments, or ideas? This newsletter is sent from my email, so you can just click reply.

Forcing China’s AI researchers to strive for chip efficiency will ultimately shave America’s lead

In incididunt ad qui nostrud sint ullamco. Irure sint deserunt Lorem id officia dolore non. Anim dolor minim sit dolor et sint aliquip qui est. Ex in tempor laborum laboris dolor laboris ullamco quis. Enim est cupidatat consequat est culpa consequat. Fugiat officia in ea ea laborum sunt Lorem. Anim laborum labore duis ipsum mollit nisi do exercitation. Magna in pariatur anim aute.

In incididunt ad qui nostrud sint ullamco. Irure sint deserunt Lorem id officia dolore non. Anim dolor minim sit dolor et sint aliquip qui est. Ex in tempor laborum laboris dolor laboris ullamco quis. Enim est cupidatat consequat est culpa consequat. Fugiat officia in ea ea laborum sunt Lorem. Anim laborum labore duis ipsum mollit nisi do exercitation. Magna in pariatur anim aute.

‍

Right now, pathbreaking AI foundation models follow an inverse Moore’s law (sometimes quipped “Eroom’s Law”). Each new generation is becoming more and more expensive to train as researchers exponentially increase the number of parameters used and overall model complexity. Sam Altman of OpenAI said that the cost of training GPT-4 was over $100 million, and some AI computational specialists believe that the first $1 billion model is currently or will shortly be developed.

As semiconductor chips rise in complexity, costs come down because transistors are packed more densely on silicon, cutting the cost per transistor during fabrication as well as lowering operational costs for energy and heat dissipation. That miracle of performance is the inverse with AI today. To increase the complexity (and therefore hopefully quality) of an AI model, researchers have attempted to pack in more and more parameters, each one of which demands more computation both for training and for usage. A 1 million parameter model can be trained for a few bucks and run on a $15 Raspberry Pi Zero 2 W, but Google’s PaLM with 540 billion parameters requires full-scale data centers to operate and is estimated to have cost millions of dollars to train.

Admittedly, simply having more parameters isn’t a magic recipe for better AI end performance. One recalls Steve Jobs’s marketing of the so-called “Megahertz Myth” to attempt to persuade the public that headline megahertz numbers weren't the right way to judge the performance of a personal computer. Performance in most fields is a complicated problem to judge, and just adding more inputs doesn't necessarily translate into a better output.

And indeed, there is an efficiency curve underway in AI outside of the leading-edge foundation models from OpenAI and Google. Researchers over the past two years have discovered better training techniques (as well as recipes to bundle these techniques together), developed best practices for spending on reinforcement learning from human feedback (RLHF), and curated better training data to improve model quality even while shaving parameter counts. Far from surpassing $1 billion, training new models that are equally performant might well cost only tens or hundreds of thousands of dollars.

This AI performance envelope between dollars invested and quality of model trained is a huge area of debate for the trajectory of the field (and was the most important theme to emanate from our AI Summit). And it’s absolutely vital to understand, since where the efficiency story ends up will determine the sustained market structure of the AI industry.

If foundation models cost billions of dollars to train, all the value and leverage of AI will accrue and centralize to the big tech companies like Microsoft (through OpenAI), Google and others who have the means and teams to lavish. But if the performance envelope reaches a significantly better dollar-to-quality ratio in the future, that means the whole field opens up to startups and novel experiments, while the leverage of the big tech companies would be much reduced.

The U.S. right now is parallelizing both approaches toward AI. Big tech is hurling billions of dollars on the field, while startups are exploring and developing more efficient models given their relatively meagre resources and limited access to Nvidia’s flagship chip, the H100. Talent — on balance — is heading as it typically does to big tech. Why work on efficiency when a big tech behemoth has money to burn on theoretical ideas emanating from university AI labs?

Turn to China now, and the story is in a very different place. The Biden administration has placed a series of export controls on advanced computation and semiconductor fabrication to China, including blocking the export of Nvidia’s H100 and A100 chips. Nvidia responded in March by offering tweaked versions that comply with U.S. export controls (with what it dubs the H800 chip), but the general policy remains in force: prevent China from acquiring the best chips that Earth produces.

Without access to the highest-performance chips, China is limited in the work it can do on the cutting-edge frontiers of AI development. Without more chips (and in the future, the next generations of GPUs), it won’t have the competitive compute power to push the AI field to its limits like American companies. That leaves China with the only other path available, which is to follow the parallel course for improving AI through efficiency.

For those looking to prevent the decline of American economic power, this is an alarming development. Model efficiency is what will ultimately allow foundation models to be preloaded onto our devices and open up the consumer market to cheap and rapid AI interactions. Whoever builds an advantage in model efficiency will open up a range of applications that remain impractical or too expensive for the most complex AI models.

Given U.S. export controls, China is now (by assumption, and yes, it’s a big assumption) putting its entire weight behind building the AI models it can, which are focused on efficiency. Which means that its resources are arrayed for building the platforms to capture end-user applications — the exact opposite goal of American policymakers. It’s a classic result: restricting access to technology forces engineers to be more creative in building their products, the exact intensified creativity that typically leads to the next great startup or scientific breakthrough.

If America was serious about slowing the growth of China’s still-nascent semiconductor market, it really should have taken a page from the Chinese industrial policy handbook and just dumped chips on the market, just as China has done for years from solar panel manufacturing to electronics. Cheaper chips, faster chips, chips so competitive that no domestic manufacturer — even under Beijing direction — could have effectively competed. Instead we are attempting to decouple from the second largest chips market in the world, turning a competitive field where America is the clear leader into a bountiful green field of opportunity for domestic national champions to usurp market share and profits.

There were of course other goals outside of economic growth for restricting China’s access to chips. America is deeply concerned about the country’s AI integration into its military, and it wants to slow the evolution of its autonomous weaponry and intelligence gathering. Export controls do that, but they are likely to come at an extremely exorbitant long-term cost: the loss of leadership in the most important technological development so far this decade. It’s not a trade off I would have built trade policy on.

The life and death of air conditioning

Across six years of working at TechCrunch, no article triggered an avalanche of readership or inbox vitriol quite like Air conditioning is one of the greatest inventions of the 20th Century. It’s also killing the 21st. It was an interview with Eric Dean Wilson, the author of After Cooling, about the complex feedback loops between global climate disruption and the increasing need for air conditioning to sustain life on Earth. The article was read by millions and millions of people, and hundreds of people wrote in with hot air about the importance of their cold air.

A few weeks ago, Bloomberg had a great piece about the rising toll of air conditioning under the similarly scary headline, “A Billion New Air Conditioners Will Save Lives But Cook the Planet.” The numbers, as always, are staggering:

Demand for air conditioners is surging in markets where both incomes and temperatures are rising, populous places like India, China, Indonesia and the Philippines. By one estimate, the world will add 1 billion ACs before the end of the decade. The market is projected to
before 2040. That’s good for measures of public health and economic productivity; it’s unquestionably bad for the climate, and a global agreement to phase out the most harmful coolants could keep the appliances out of reach of many of the people who need them most.

This is a classic feedback loop, where the increasing temperatures of the planet, particularly in South Asia, lead to increased demand for climate resilience tools like air conditioning and climate-adapted housing, leading to further climate change ad infinitum.

And there doesn’t appear to be a way out. Scientists have determined that there is a zone of comfort for humans where we can function at peak performance. Much of the industrialized world happens to sit in these temperate zones, which is one argument for their long-term economic success compared to the tropics. But as climate change intensifies, the world’s temperate zones are expected to narrow dramatically, and some countries like India may have next to no temperate zones left to protect its people. Air conditioning is a necessity (with some even calling it a human right), but greener coolant alternatives are still in their infancy. In other words, cooling innovation needs to heat up in a big way.

Lux on the Farm

Josh Wolfe speaking at Nvidia Auditorium at Stanford. Photo by Chris Gates.

Josh Wolfe gave a talk at Stanford this week as part of the school’s long-running Entrepreneurial Thought Leaders series, talking all things Lux, defense tech and scientific innovation. The
.

Lux Recommends

As Henry Kissinger turns 100, Grace Isford recommends “Henry Kissinger explains how to avoid world war three.” “In his view, the fate of humanity depends on whether America and China can get along. He believes the rapid progress of AI, in particular, leaves them only five-to-ten years to find a way.”
Our scientist-in-residence Sam Arbesman recommends Blindsight by Peter Watts, a first contact, hard science fiction novel that made quite a splash when it was published back in 2006.
Mohammed bin Rashid Al Maktoum, and just how far he has been willing to go to keep his daughter tranquilized and imprisoned. “When the yacht was located, off the Goa coast, Sheikh Mohammed spoke with the Indian Prime Minister, Narendra Modi, and agreed to extradite a Dubai-based arms dealer in exchange for his daughter’s capture. The Indian government deployed boats, helicopters, and a team of armed commandos to storm Nostromo and carry Latifa away.”
In the happier news department of science, Ina Deljkic recommends “’Quite science fiction’: Brain-spine interface system helps man with paralysis walk again.”
Sam recommends Ada Palmer’s article for Microsoft’s AI Anthology, “We are an information revolution species.” “If we pour a precious new elixir into a leaky cup and it leaks, we need to fix the cup, not fear the elixir.”
I love complex international security stories, and few areas are as complex or wild as the international trade in exotic animals. Tad Friend, who generally covers Silicon Valley for The New Yorker, has a great story about an NGO focused on infiltrating and exposing the networks that allow the trade to continue in “Earth League International Hunts the Hunters.” "At times, rhino horn has been worth more than gold—so South African rhinos are often killed with Czech-made rifles sold by Portuguese arms dealers to poachers from Mozambique, who send the horns by courier to Qatar or Vietnam, or have them bundled with elephant ivory in Maputo or Mombasa or Lagos or Luanda and delivered to China via Malaysia or Hong Kong.”
Finally, Sam recommends Brick Technology’s video of “Making A Billion-Year Lego Clock.”

That’s it, folks. Have questions, comments, or ideas? This newsletter is sent from my email, so you can just click reply.

‍

October 28, 2023

Existential Engineer

<- Previous

continue

reading

November 11, 2023

Powerful Productivity

Next ->