The AI Paradox

Why Both Sides Are Wrong

Everywhere, we’re noticing things are changing. Some embrace it, some fear it. But whatever your perception, AI is here, and its existence cannot be denied. Discussions about it resemble where people stand on coriander: people either love it or hate it. No in-between. Many are overestimating what these models are, and others underestimating what they can do.

So, what is it then? Well, under the hood, it’s just a lot of math doing large-scale pattern recognition to make accurate next-token predictions, using cleverly wired circuits to make the most out of its limited context window. Many people use this argument to downplay the significance of the technology. But as with most advancements, looking back, it seems so simple. We could describe the internet as just electrons travelling through wires to deliver 0’s and 1’s. The truth is, both the internet and large-scale pattern recognition, even in their early stages, have already automated a large portion of specialised work. As a result, we hear AI company founders predicting that AI will eventually do all our work. But why are they saying this?

Well, who wouldn’t, after investing well over a trillion dollars into the technology? That investment needs returns. The fastest path to returns? Replacing human labor. It’s their economic incentive.

On the other side, there are people saying it’s just a bunch of calculators knitted together, and it won’t affect them. But that also makes sense if you’ve just spent years of your life specialising in something that’s now being automated.

The truth, as always, lies somewhere in between. It’s important to keep the conversation going. That is, however, hard when the stakes are high. So let’s address some of these concerns. By opening up about these fears, we allow ourselves to venture into what the actual future might look like.

There’s something paradoxical about what they’re claiming. If we take the economic incentive of investors and assume most (or all) workers get replaced by AI, two things can happen, and they’re both paradoxical:

1. Workers → earn income → buy products → create demand. Remove the workers, and demand disappears. AI can’t be the customer. So the investors don’t get their returns.
2. Demand doesn’t disappear, which creates a huge gap in wealth distribution. Now, governments have to step in and redistribute the wealth away from these companies. Again, out of the pockets of the investors.

This is how it is, and how it has always been. So what’s the other option? Like previous automation revolutions, it will generate new work and new jobs. Historically, that’s the likely outcome. And ironically, it runs directly against the industry’s claims. They can, however, both be right, if we change “most work” to “most CURRENT work”.

A recent Harvard Business Review study found that AI doesn’t reduce work, it intensifies it. One of the things they observed: people expanding into new tasks they’d never done before. We see this ourselves. With AI accelerating delivery, the lines between engineering and product are blurring. Engineers are picking up customer-facing work and Product Managers are opening pull requests.

Embracing these changes and adapting along with them opens up a tremendous horizon of possibilities. Of course, we won’t venture across it without facing some resistance and challenges along the way. Will there be disruption? Yes. Will there be layoffs? Absolutely. Will there be re-hires? For sure.

When everything is possible, applying focus has never been more important. Inefficiency is everywhere. Why? A tale as old as time…

. . .

*Next up. Part 2: So where does all this inefficiency come from?*