The Burkean Case for AI Deceleration
Rapid change can lead to social breakdown.
Disclaimer: This article is something of a devil’s advocate exercise, and does not necessarily reflect my true opinions.
Conservatives are sometimes straw-manned as believing in total stagnation, but this is not true. Ask any intelligent conservative, and they will tell you that they’re fine with change, so long as that change is slow and manageable. To a conservative, society as it stands today works fairly well, and we should be cautious before changing it. Not totally hostile, but cautious. We should favor reform instead of revolution.
This has been a standard lowercase-c conservative argument since the time of Edmund Burke, and to show that it’s valid, we only need to look at the American and Russian Revolutions. The American Revolution was extremely conservative, by revolution standards. We basically just toppled the British monarchy, while leaving the rest of society intact. Afterward, the next generation of Americans gave poor people the right to vote, while leaving the rest of society intact. Then the next generation of Americans freed the slaves, while leaving the rest of society intact. Then the next generation of Americans gave women the right to vote, while leaving the rest of society intact. Then the next generation of Americans ended racial segregation, while leaving the rest of society intact. Then the next generation of Americans gave greater respect to women and gays, while leaving the rest of society intact. Nowadays, though America still has many problems, it’s one of the richest and most egalitarian societies on Earth. This level of progress, which would have seemed unimaginable to an American from the Founding Fathers’ generation, was enabled by our slow and deliberate pace of change. At each step, we were cautious not to move too far too fast. We gave people time to adjust to new social realities, and we even occasionally walked back progressive reforms that had gone too far.
Compare this to the Russian Revolution. In Russia, they tried to move from a czarist monarchy to an egalitarian communist society in the span of a single generation. They totally overthrew all pre-existing social hierarchies, and either imprisoned or killed anyone who stood in the way. The result? Tens of millions of people died unnecessarily, and Russia today is no more free than it was 120 years ago.
The lesson of this history is clear: Incremental change is both more durable and more desirable than massive and sudden change. That same lesson that we apply to political revolutions can also apply to technological revolutions. It’s fine for society to change by adopting new technologies, but when too much technology arrives too quickly, it can cause major disruptions.
In previous eras of world history, this wasn’t too much of a problem. A person born in 1900 would have lived a mostly similar life to their parents born in 1870, who would have lived mostly similar lives to their parents born in 1840. While the technological change across many generations was striking, the change within a single generation was manageable. People learned to adapt.
The same cannot be said today. I am living a fundamentally different life than my parents lived when they were my age, which makes it genuinely difficult for me to relate to my parents sometimes. And it’s not just me. Parents around the world are reporting that it’s getting harder and harder for them to understand their own children, and this is causing significant developmental issues. Technology in the 21st Century is moving so quickly that people don’t have time to adapt before the next technology comes along. Hell, we haven’t fully adjusted to the effects of social media yet, and that technology began more than 20 years ago.
That’s why I’m concerned about the extremely rapid pace of AI progress over the last few years. Whether you judge by state-of-the-art technical achievements, industry adoption, or consumer use, AI is accelerating faster than perhaps any previous technology in human history. I spend a large portion of my free time tracking AI developments, and even I feel like I can’t keep up with how fast things are moving. Can the average person realistically hope to keep up? Not even close.
I see no sign of things slowing down, either. Leading AI industry leaders have made repeated claims that they are on track to achieving artificial general intelligence within just a few years. You can reasonably doubt these claims as just a bunch of industry hype, but I wouldn’t be too skeptical. AI has made genuinely crazy progress over the last 5 years, and I would not be surprised if the next 5 years of progress are even crazier. When Sam Altman tells us that the Singularity is coming, I’m inclined to believe him.
Sam Altman assures us that this Singularity will be manageable, but I don’t trust him. The pace of AI progress already isn’t manageable. It will take many, many years for society to adjust to the AI tools that are already on the market. Adding even more AI tools that are even more capable will turn an already out-of-control situation into total anarchy.
The reason why technologies have historically improved society rather than destroying it is because historically, societies have had time to adjust. We’ve found ways to incorporate the benefits of the technology while mitigating the risks. This process hasn’t always been fast or perfect, but it works. It’s the equivalent of the American Revolution. By contrast, what AI companies are doing to our society today is akin to the Russian Revolution. They are dumping generations worth of technological change onto us all at once, and just hoping it will go well.
Our social antibodies are completely underprepared for the onslaught of AI that we are seeing today, and I fear that it could lead to total social breakdown. Already we are seeing the early stages of this. An increasing number of college students no longer know how to write because they’re outsourcing their writing to chatbots, and an increasing number of young people are abandoning human relationships in favor of AI boyfriends and girlfriends.
(I should note that this concern over social breakdown is in addition to other concerns about AI, including the risk that a misaligned AI system will attempt to wipe out all of humanity.)
AI is not inherently evil, and I am not inherently opposed to AI progress. Maybe, at some distant point in the future, once society has finished adjusting to the technologies of today, we can consider adding more technologies to the pile. But until that day comes, it would be reckless to continue accelerating AI. Society needs time to adjust to radical change, and since the AI industry does not want to give us that time, we must demand it. The conclusion is clear: We must act now to decelerate AI, before it is too late.
As I said above, this argument does not necessarily reflect my true beliefs. I think that AI’s benefits could be so massive that they justify taking otherwise unacceptable risks like social breakdown or even human extinction. Still, it’s worth pondering this question. Advocates of progress (myself included) need to remember that progress does not come without a cost. Creating a new world necessarily means destroying the one that came before, and people are justified in wanting to protect the world they have. We must tread carefully.
