Should AI be regulated? How should AI be regulated? Who should regulate AI? Lots of questions here, and lots of answers. As usual, when technology and politics merge, we tend to get the worst of both worlds, but enterprises have a view here, and so do I, based on decades of both tech and political experience. Let’s see if we can find something useful.
To the first question, I think the answer is clear; yes, we need to regulate AI. The real problem comes in the second and third questions, starting with the fact that regulation is justified by preventing harm in some form, and so we have to know just what harm we’re trying to address.
The obvious harm of AI comes from the very terminology. Artificial intelligence is either a nonsensical fraud or it’s a technology that appears to mimic at least some aspects of human intelligence. We don’t let humans do anything they want, so we can’t let AI do any of those things either.
AI can generate voice, images, videos. If these are realistic enough, they can mimic a real person, and so it would be possible to generate stuff that could be embarrassing or even things that might lead to criminal prosecution of the target. To use AI to mimic a real person leads to risks that have to be addressed through regulation.
AI, of course, can also generate stuff that doesn’t mimic a real person, but instead constructs an artificial person. Here, the problem is that this artificial person could still commit real liable or slander, and potentially even conspire in real criminal activity. The harm this could cause is equal to the harm that a real person could cause in the same activity, and so it must also be reduced or eliminated through regulation.
Suppose our artificial person looked/sounded very real. Could it be “hired”, meaning take the job of a human? Hollywood and the music industry are already worried about this, and even more worried if you modeled the artificial person on a real one. I think regulations to prevent this without permission of the target people is essential, but this raises the broader question of whether AI impact on jobs should be regulated.
Enterprises are mixed on that one, with the split being largely based on age and position. Generally, younger and older workers tend to favor some job-protecting rules, and generally workers in higher-level positions are less concerned than those in more clerical or laborer roles. Senior types are concerned about the potential social/political impact should the use of AI displace a lot of workers, creating employment and social welfare problems overall.
I’m ambivalent on this one, so far. The industrial revolution was populizing, so it generated both jobs and lower costs and raised overall living standards and economic strength. The computer revolution has been a bit less populist in impact; you could argue that it exacerbated the division of wealth. An AI revolution could bend in either direction, and at this point I think it would be hard to say which is more likely. I think it would be logical to assume that some policy of expanding data collection and reporting on the topic would be in order for now, with action determined when we have a better handle on just how AI is impacting overall employment.
Then there’s the issue of data centers. Here, enterprises seem to have a consistent viewpoint; AI data centers should not in any way threaten the utility costs for residential and business properties. Regulation to require them to assure that’s the case should be considered strongly, but many think that it will be difficult to establish consistent rules given the patchwork of utility regulations in place in many countries, including the US.
For tech, there’s another issue, which is the impact of AI on ad sponsorship. The impact comes from two sides; the impact on ad delivery and on personalization and targeting.
AI summaries of search requests are already messing up SEO strategies for advertisers and online sites. Google has an obvious advantage in dealing with search impact; some are already seeing ads spliced onto AI summaries, so Google is selling placements there. Can this make up for the smaller number who click on search results? Google seems to think that for them, at least, it could. For most, it depends on the broader issue of personalization and targeting.
Google already personalizes ads based on what it knows about users. Most tell me that they prefer ads that align with their interests to those that do not, though some worry about how much companies like Google know about them. It’s perfectly plausible to assume that more personalization and targeting will emerge as a result of the combination of AI impact on ad delivery and its ability to personalize more effectively.
I’m like many, but not all, people in that I don’t “see” ads at all. They don’t register on me because I simply don’t look. For web ads, I ignore them. For TV ads, I check my phone. However, if I’m actively looking to buy something I would be likely to pay attention to ads related to my purchase goal. Could AI predict, with accuracy, what I’m looking to buy? Some of my Meta contacts suggest that Meta thinks that social media is a forum where people share information about themselves and their goals far more explicitly than they’d reveal indirectly by search behavior. Could that then help Meta target? Sure. Could it encourage companies like Google and Microsoft to seek access to your emails, texts, and even your web activity through their browsers? Maybe.
The problem with personalization is that you can use personal information for more than ad targeting. You could use it to target malware, deny someone insurance of multiple types, and even to run complex scams. To the extent that these outcomes are real risks, they’d need to be considered for regulation.
AI has the potential to alter our social, economic, and quality of life factors in ways that other technologies have not, which means it poses a potential threat to all of them. However, potential threats aren’t a good way to target real regulations; they’re usually too vague to address effectively, and the actual impact is rarely known well enough to compare to the impact of an attempt to regulate.
The “who?” question is the most problematic, and is related to this. AI is today projected through the Internet. If regulatory policies are set by governments at various levels, they’re almost sure to be inconsistent. Would that inconsistency erode any benefit, given that it might be easy to bypass stringent regulations simply by going to the right sites? Or could regulating AI lead some geographies to fall behind in a giant technology advance? Or both?
We need to sort this out, but I doubt that the political arena is suitable for the task. What is? I wish I knew.
