FREE NEWSLETTER
As retirees, we spend a lot of time thinking about risks that may never happen. We diversify our investments, buy insurance and keep emergency reserves. We don’t need to know that something bad will happen before protecting ourselves against it.
I’ve been wondering whether we should think about AI the same way?
Some of the people building the world’s most powerful AI systems are loudly warning about the risks, even as their companies spend billions racing to make them more capable. I don’t blame them—if one company slows down, another may get ahead. But that’s also why I’m uncomfortable leaving the safeguards entirely to them.
So, indulging my fondness for Isaac Asimov, here’s my admittedly grand proposal: a 28th Amendment!
I’d establish three principles while we humans are still unquestionably in charge. 1) A machine isn’t a person and has no constitutional rights. 2) An AI can’t have the final say over taking a human life, starting a war or sending someone to prison. 3) And humans must always retain the ability to override or shut down an AI.
Congress could handle the details. Borrowing again from Asimov, I’d require advanced AI to have an independent safety filter it can’t disable. I’d keep it from having unmonitored access to banks, power grids, weapons and other critical systems. And I’d require an off switch if it’s hacked, corrupted or starts operating outside its boundaries.
I don’t know whether the worst predictions about AI will come true. I hope they don’t. But that seems beside the point. We don’t diversify because we know which investment will fail. We do it because we don’t know. Perhaps AI deserves the same humility.
I think most folks think AI is or will become a sentient bogeyman that will rule over us. I think of AI in terms of bringing quantities of compute power to a problem. We have been quietly raising the bar on how much that is and lowering the bar on what problems can be addressed. The compute power is sifting through large data sets in a timely fashion using logic that can grow based on what was learned previously. So, the real issues are how reliable is the data and what logic choices have been used to provide an answer. More transparency in these two areas will lessen our fears going forward. We cannot stop the rise of compute power and we cannot stop more and more applications to solving our issues.
Great idea, only I expect the machines might not much care about a Constitutional amendment, or would simply find some clever legal strategy to circumvent such restrictions. Ultimately, it isn’t the machines I worry about so much, as it is the people who are creating and running the machines.
Let’s face it, we need to have some Government regulation. This will sound funny, but I asked AI the question, and they gave very reasonable answers including that Government needs to take the lead. I have confidence, we will have problems for sure, but also we will proceed into the future as we always have.
I think the situation needs addressing, but I think you have prescribed the wrong tool.
Congress can regulate AI through ordinary legislation. Legislation is vastly easier to change as technology evolves. Constitutional provisions are intentionally difficult to change. And we have very little idea what “AI” will mean 30, 50 or 100 years from now.
Great Mark G with your suggestions. Any Government regulation is better than nothing. USG definitely needs to play a leadership role. Other countries may not even know where to start since all the major AI companies are operating from our land here.
As a simple software engineer continuing in the Tech Industry for four decades, I wrote my view on this topic a week ago in three pages which can be read or downloaded as PDF from the following URL.
https://lnkd.in/p/eyhTxmbs
Cheers.
Your idea might be sound but just by mentioning Congress you’ve illustrated that your thinking is too narrow. AI is not just a American risk it’s a global risk and any one country having a killswitch may not be enough. Plus in a world of highly vested interests and politicians in the pockets of corporations etc the question of vesting authority to exercise a killswitch in an individual etc is itself problematic.
But then multinational consensus is the work of many years to achieve and mankind may not survive that long if rogue AI really gets going.
I’m pretty fatalistic about the whole thing. I’m pretty sure the tech bros already will be hightailing to their bunkers in New Zealand before the rest of the world gets a whiff that the chain reaction has started.
In the meantime it’s the perfect engine for capitalism to eat the ordinary working public and create a global non-working class.
Mark, I think you may be focused on the wrong threats. In my view, the most immediate danger is that AI could lower the technical barriers for small radical groups to develop highly transmissible biological agents. That is a dangerous Pandora’s box—one that could plausibly lead to an extinction-level event.
Mark – You’ve adapted the Three Laws of Robotics! I like it.
I’m as much concerned about how the Large Language Models are being trained on stolen literature (books, magazines, periodicals, etc.), as well as unmoderated/edited/proofed content such as X, Instagram, and Substack. The day is coming when an AI response will be based on a ridiculous Substack post such as 1 + 1 = 1 or worse.
The original three laws were fine, and should be adopted.
Well put, Mark. It’s crucial for the AI industry and Congress to work together to formulate a plan, and enact legislation and regulations to implement it. And they need to start right now.
But the danger exists beyond our borders. If China is not part of the response, we’re still in trouble. Xi Jinping will be here in a couple of weeks to meet with Trump. Let’s hope this topic is at the top of their agenda.
We can’t say we haven’t been warned; and we can’t afford to be too late on this one.
Free Newsletter
Arrives weekly.