Tuesday, September 24, 2024

The Cost of AI

 

 (Image by Joshua Sortino at Unsplash)

I’ve been an IT professional for over 40 years. The idea of artificial intelligence (AI) has been around for longer, and was well-known enough to become popular (and possibly prophetic) entertainment, from Colossus: The Forbin Project to Skynet and the Matrix.

We’ve been using rough forms of AI for years to go beyond simple scripting in order to automate mundane tasks, using machine learning and inference engines to handle common alerts and respond to them without having to wait on a human to react.

So the concept of AI is nothing new, but both the rapid advancements of the past few years and the capturing of the public imagination have kept AI (in particular, generative AI) heavily in the news, and there’s no end in sight.

As an IT guy, AI fascinates me. I think the possibilities for how it could improve our lives are boundless. On the other hand, as a photographer and writer, I agree with those that find it horribly threatening. The advances made in just the last couple of years allow generative AI platforms to “create” visuals that leap right over the “uncanny valley” into photorealistic images that are not easy to differentiate from
an actual photo. 

The huge AI engines have done this by scraping massive amounts of copyrighted material (quite possibly including my own) with no compensation for or even acknowledgement of the actual creators. And those enormous corporations that are doing this are not being held accountable.

That fact does get some airtime, and while I think absolutely nothing will be done about it, at least there is some general understanding that there’s something wrong there. But there is another fact about AI that I don’t see much written about, and I think it’s important. In addition to AI platforms learning from ingesting the work done by writers, artists, researchers, etcetera, every time you use an AI platform, you are teaching it something. 

AI platforms require an inordinate amount of compute power. They also require a ridiculous amount of storage — more data, more better. All of that costs a ton of money (and sucks up a huge amount of very costly energy — something else no one wants to talk about). Put simply, they are expensive as hell.

So how can they incorporate AI for free into the platforms like Windows operating systems that we use every day? How can platforms like ChatGPT or Midjourney be free or very inexpensive?

It’s because they are using you, the consumer of those services, to continue to teach the AI engines. Which will make them better and better, very quickly. And they will continue to supplant, well, you. Writers, artists, editors, reporters… that’s just the start.

We already run IT shops with a small fraction of the personnel that it took in the 1980s when I started out as a mainframe support engineer. We call that progress, and I suppose that it is. And as rapidly as computing platforms continue to grow, we still hire a bunch of people. But I can see the end of that workstream coming soon as AI is becoming capable of performing more and more of those functions.

When the AI platforms are capable of making better corporate financial decisions than the finance guys, do we need anyone below the CFO? When the AI can make better technology decisions, do we need anyone other than the CTO, who can just rubber-stamp the AI’s decisions until they decide they don’t need him/her anymore either?

Publicly traded corporations are required by law to maximize stakeholder profits. That often leads to decisions that are not in the best interest of the people that work for those companies. What happens when, in ten or twenty years from now, when there ARE no workers? When major corporations are just shareholders and a handful of rubber-stamping execs? Honestly, that is not that far-fetched.

I’m not suggesting that you avoid AI. It would be hard to do today and it will be impossible tomorrow. But I do think we need to understand what is happening and what is inevitably coming next. There are smart people working on safeguards, but I do fear that they will be well behind the curve in controlling AI growth.

All I’m suggesting is that you be aware that every time you create a Leonardo image and tweak it until you’re happy with it to use in your ad campaign, and every time you use Google Bard to clean up a report for work, you are providing a free service to a major corporation (just like those artists whose work was stolen) and I’d be willing to bet that those services won’t remain free/cheap for long.

I’ve seen some pretty major inflection points in my six decades here. The democratization of computing with the introduction of the PC, the widespread availability of knowledge (and bullshit) with the advent of Internet and the ability to put ALL of that in your back pocket with the revolution of the smart phone. I believe AI is the fourth, and probably the most impactful.

Labels: , , , ,