AI, AI, OH!!!

AI is not going away, so we need to find ways -- both public and private -- to maximize its net benefits

Photo by Andrea De Santis on Unsplash

This morning while reading the latest music news, I came across the following quote in Blabbermouth from the always delightfully controversial Sebastian Bach:

So I will always give you real stupidity instead of artificial intelligence… I mean, just the word ‘artificial’ — that should turn off most people. Artificial — no thanks. How about real? I’m not into artificial anything, really.

I have to say the man speaks the truth: he often does give us real stupidity! :-)

But I also share his opposition to AI-generated music. The first time I encountered it was when I listened to a cover of a Sisters of Mercy song on Youtube, and only then realized there were no actual musicians involved. There was no monetary cost to me of watching the video, but I still felt ripped off for incurring the time costs of playing it.

To explain it using even more economic jargon, my opportunity costs outweighed the benefits of playing the video, and since this hidden information was revealed to me only after playing it, I fell victim to an adverse selection problem (more specifically, a Lemons Problem).

AI-generated music bothers me so much because there is zero musical ability being demonstrated; just programming capabilities. It is even worse than using electronics in actual music, because at least the programmer has to have some talent to create music with those electronics. But with AI, a programmer just tells their computer to scan existing music from real musicians, and then copy it in certain ways.

AI music “creators” are hardly different from ticket scalpers: they offer no value-added to music because they rely entirely on what real musicians have already done, and therefore profit from those efforts. As argued by Greg Kennelty for Metal Injection:

Let’s get this out of the way right up top – A.I. “music” isn’t music. It’s slop given to the uncreatives of the world by corporations whose only real goal is to make money. And if that comes at the expense of destroying art, then whatever – the shareholders are happy folks and all’s right in the world of the worst people you know.

In a recent interview with El Jevilongo, Lacuna Coil vocalist Cristina Scabbia had a similar viewpoint…

Then there is the plagiarism aspect, where AI is used to write papers based on what is already out there. Similarly, “scam journalists” have been detected (too late) in the past, to the embarrassment of media outlets who were fooled into publishing articles “written” by these scammers (credit to Michael Barclay for directing me and his other readers to that article).

I experienced something similar when I hosted a practicum student who I asked to do research for me: based on the robotic tone and vagueness of what they submitted, as well as how quickly they submitted to me, it was clear to me they used AI to write it; I just could not prove it. It was very frustrating, but fortunately I felt much better when a subsequent practicum student proved to be a godsend to me! Now she clearly put in a tonne of effort and gave me high-value output!

Of course, not all musicians are against AI, so to be fair I will quote Clown from Slipknot, who is definitely not speaking to the choir on this issue:

In a new interview with gaming magazine The Escapist, the co-founder and percussionist for nu metal’s nine-man wrecking crew calls AI “a professor in my pocket who only wants to do what I ask it”. “I’m employing AI 190 percent,” he states.

He also says, “I’ve been using AI my whole life,” viewing it as the latest in the series of tools which have been set up to help musicians over the years. He then claims to have “thousands and thousands” of poems that he’s written since he was young, adding that, through AI, he can transform them into something else without compromising his vision.

“Here are my words,” he says. “Don’t change them. Don’t alter them. But show me some different ways to sing it.”

He continues: “What’s the difference between me pulling out my pocket producer… or me trying to get a famous producer that might not even work with me and could potentially cost me $150,000… who will only give me one or two ways – I’m not mentioning any names!”

Personally, I think an important distinction between his “pocket professor” and a human producer is the human can bring an original perspective to the music, while AI will only spit back what already exists online in one form or another, which again raises plagiarism concerns. Of course, a human producer can also plagiarize and there needs to be guardrails against it, but they are not limited to that form of writing.

However, this post is not just about music, because the Blabbermouth article also got me to thinking about AI from economic and policy perspectives. On that note, there are two particular people who write often about AI from very different viewpoints. Given those differences, I often agree with both of them despite the fact only one of them is very much a critic of it:

1. James Pethokoukis: a columnist and economist at the American Enterprise Institute, who operates the newsletter Faster, Please! He is very much on the “pro” side of AI because he focuses on it from an economic perspective which views AI as a potentially pro-growth technology, similar to technologies that emerged from historical Industrial Revolutions.

2. Ted Gioia: a writer and musician who therefore focuses more on music and culture with his newsletter The Honest Broker, and is clearly on the “con” side of AI for those reasons.

So all of that is to say I understand both sides of the debate: I care about jobs lost due to AI and other crimes that can be committed with it (e.g., pornographic deepfakes), but just as technologies arising during Industrial Revolutions did not go away, neither will AI. Even if the Canadian government tried to ban it, it would not be able to keep it from being “imported” from other countries like China.

Therefore, I prefer governments legislate and enforce policies which attempt to minimize the costs of AI while not discouraging — and even incentivizing — the benefits, all the while helping ease the transition between equilibria.

That leads me to what the Canadian government is currently doing to regulate AI, particularly as reflected in the 2025 Canadian Budget document. It looks good overall, so I am interested in seeing how it progresses.

Specifically, on pages 14 and 20 of the document, the government states:

A new Office of Digital Transformation will lead the adoption of AI and other new technologies across government. This, coupled with near-term procurement of made-in-Canada sovereign AI tools for the public service, means not just a more efficient government, but also a generational opportunity for our domestic innovators.

…

Driving growth and innovation: Our government will make it cheaper for Canadian companies to invest through a new productivity super-deduction. We will invest in AI projects to further the technology’s adoption at home, and catalyse private sector investment in Canada’s most innovative startups. The government is also developing a new AI strategy, and will consider whether new AI incentives and supports should be provided. We are creating the right conditions for real, sustainable growth.

It goes into more detail on pages 205 and 214-15.

I appreciate the federal government’s intentions, as well as its desire to work with the private sector on these policies. But I do not think it is necessary for it to be “made-in-Canada”, although there are valid reasons to not rely too much on foreign suppliers. As Andrew Coyne wrote recently for The Globe and Mail in the context of defence:

The rule should be: Source from whomever can supply the best equipment at the lowest price, and – since the hour is late, the situation is dire, and we are starting from a long way back – in the shortest time. If that happens to be a Canadian supplier, fine. But if it is a foreign (non-U.S.) supplier, also fine. The focus should be on getting the biggest bang for the buck, not the nationality of the source.

There can be exceptions to this rule. In a global emergency, international supply chains can be interrupted, just when they are most needed, as we learned during COVID-19. Where sudden scalability is an imperative, it might be prudent to insist on locating production on Canadian soil, even if the supplier is foreign-owned.

And certainly we should want to avoid being too dependent on any one company, or any one country, for our defence needs, even if they do not represent the sort of potential threat that the United States has lately become. The watchword of statesmen in the current international climate is diversification, risk-management, hedging.

But the default should be to open competition, at least among non-U.S. suppliers.

In summary, I am glad the Canadian government is putting work into developing policies which signal its acceptance of AI’s involvement in our economy and other areas of our lives, while also trying to maximize its net-benefit to Canadians.

However, it would be very helpful for the private sector to do its own independent work in creating markets aimed at preventing harms done by AI, such as ways to more easily detect whether work is being done using AI. But for that to happen, we need Canadians to demand these products and services so suppliers can foresee profit opportunities in them. Otherwise, we will have a missing-markets problem as we do with pure public goods and common resources.

On that note, we need to find a way to discourage people from wanting to use AI for bad things like plagiarism and deepfake porn, because without the demand there would be no supply. To be honest, I have no solutions to offer in that regard because from my experience, incentives seem to be aligned with people being lazy and taking advantage of other people for their own gain. So I appeal to people smarter than me for ways to align incentives with the common good.

We also need to find ways to reduce the incentives of the social media companies who not only tolerate AI crimes — such as deepfake porn — but also encourage them, and even develop ways to commit these crimes. The most effective way for this to happen would be for “consumers” of social media platforms to stop using those platforms, but all Canadian governments can easily take steps in the right direction by simply getting off these platforms themselves (cough-X/Twitter-cough)!

Thank you for reading to the end, and I hope to see you here again next time. Also, I ask you to please consider a paid subscription, as it will help me to afford to pay my opportunity costs of writing these articles. Regardless, I am happy you are here!