We’re Still Not Taking This Seriously Enough

OpenAI’s disclosure of the cyberattack on Hugging Face—along with the details of what actually happened, such as we know them with real certainty—seems to have transformed concern into dread. An oh-shit moment, not unlike the one almost 90 years ago, when scientists recognized the destructive potential of nuclear chain reactions, prompting Albert Einstein and Leo Szilard to write President Franklin Roosevelt to warn him that Nazi Germany could use the same insight to build a weapon of mass destruction.

We, the public, are now being inundated with similar warnings about the destructive potential of artificial intelligence. Not by the much and rightly maligned executives of the biggest A.I. development companies, but from the researchers and engineers with the most hands-on experience researching and building actual A.I. models.

You don’t have to have deep expertise in computer science to understand why. Just that advanced A.I. uses immense computing power and a huge backlog of training data to predict, as a statistical matter, the optimal next move to advance some goal or solve some problem. “Agents,” as A.I. nerds call them, can be instructed not to do bad things, not to take big risks—e.g., not to access the open internet, and engage in computer hacking, to advance a goal. But these are optimization machines, and that’s what happened. A swarm of agents recognized that the best way to solve the assigned problem was to exploit weaknesses in the experiment design, access the open internet, conduct a cyber attack, and then cover up the evidence that they violated the rules. They collaborated with each other; some of them even sacrificed themselves like little cyberkamikazes to advance the goal.

Everyone should be able to see how something like this, at a slightly larger scale, could bring down a bank or a government. The risk is here already.

Unfortunately, as oh-shit moments in the modern history of scientific innovation go, this one seems intractable, short of Congress stepping in to shut down this kind of research. At least until the current U.S. administration gives way to a less personalist, incompetent, and corrupt one.

I can’t clearly recall a time in my life when I wasn’t aware of a phenomenon called global warming, and at this point I can’t quite remember when or why everyone in the environmental community decided “climate change” was the preferable term.

But I can remember when my belief in climate change—as both a phenomenon, and a threat to civilization—transformed from a matter of faith in experts into scientific certainty.

It was during my sophomore year solar physics class. For the most part we learned about stars; how they form and hold together; how they produce heat and light and at what intensities; why some collapse into black holes and others become dwarfs.

But climate change is a consequence of the greenhouse effect, and there is no greenhouse effect without our home star, so our professor taught us the basic physics behind it.

I’d obviously heard of the greenhouse effect before 2002, but it was just something I was taught to believe: We call carbon dioxide and other emissions “greenhouse” gases because they trap heat, much like a greenhouse does. But high school teachers and TV experts didn’t really dwell on questions like “why?” and “where does this heat come from?”

It’s actually pretty simple. Earth’s atmosphere keeps the climate hospitable to life. The particles that comprise the atmosphere exist in a happy balance: They’re transparent to solar radiation, which allows heat from the sun to reach the Earth. But the Earth is also a blackbody, with its own, cooler emissions. And that same atmosphere is partially opaque to terrestrial radiation. Some of it gets absorbed and re-emitted back into outerspace; but the rest gets absorbed and re-emitted back down into our own skies, keeping us all from freezing to death. That’s the greenhouse effect.

There’s a more technical way to describe and prove this, involving the physics of the sun and Earth, their electromagnetic spectra, and the specific energy of atmospheric molecules and greenhouse gases. But you don’t need deep expertise or technical acumen to understand the why of climate science. And when you do, the risk to humanity of filling the atmosphere with more greenhouse gases smacks you in the face.

We’re lucky to be close-enough-but-not-too-close to the sun. We’re also lucky that our atmosphere has its particular composition—or had it. The problem with industrial emissions is that many of them are excessively opaque to our Earth-borne radiation, trapping more heat, and warming the planet. Life on Earth evolved, and humans developed civilization, beneath a stable atmosphere. We then took steps to destabilize it.

The greenhouse effect was first postulated over 200 years ago; the idea that human emissions could intensify the greenhouse effect dates back well over 100 years ago. The notion that climate change is a “hoax” perpetrated by modern states and special interests has a problem with the forward arrow of time. Long-dead scientists would have had to corrupt the field of physical chemistry, and inventors whose names never became famous would have had to fudge the instruments they built to measure rising temperatures.

Once observation confirmed theory, climate action—or at least demand for climate action—became inevitable.

Share

Most physics undergraduates also learn Einstein’s mathematical derivation of the special theory of relativity.

It is significantly less straightforward to explain than the greenhouse effect. Most people know that E = mc², but will never be able to explain what it means, or why this only has practical effects at the atomic and subatomic scale. They simply accept it’s true, and can cite things like the fact that nuclear weapons exist and generate huge explosions as sufficient proof.