Aren’t they?

This is one of those times I would like to be more expert than I am. I don’t know much about computers, much less Artificial Intelligence. I’m smart enough to know that I’m not smart and more often wrong than right.

Various and sundry super-clever people from the world of Artificial Intelligence are scaring the shit out of everyone. Blithely inventing % chances that AI will wipe out the world in x years. One day it’s a 10% chance with 30 years, the next it’s a 15% chance by next Thursday.

Yikes!

There are some voices in dissent. Ed Zitron is one, and Cal Newport another. No offence to Ed, but I’m going to focus on Cal.

Who is he?

I know him through his books, his newsletter and his podcast. He’s an academic. He got his Doctorate in Computer Science at MIT, and is now a tenured professor at Georgetown University.

He is sceptical of the claims emanating from the AI companies and related parties. He points out that most of the claims sound and feel more like marketing copy than a scientific paper. On his podcast, he releases episodes working through specific claims, and I recommend them.

How can it be that some scientists claim that AI is going to be as transformational as the Industrial Revolution, while others think it heralds the apocalypse, and still others are certain that it’s absolutely neither of those things?

Well...

  1. Commerce. The technology requires massive investment. There’s an obvious incentive to talk up the impact.
  2. Commerce. The technology requires massive investment. There’s an incentive to use regulation to raise the barrier to entry into the market - protecting investment in early adopters.
  3. Language. Commentators talk about Artificial Intelligence, or “AI”. Seldom do they define what that means. So, pay attention here, if the term is not defined, then it is perfectly possible to assign absolutely any function to it.

We’re all lazy. None of us wants to learn a whole new register and vocabulary, but we’re going to have to.

In a gross over-simplification, let me put it like this:

Artificial Intelligence (AI) may absolutely destroy the world.

Large Language Models (LLM) will not.

I think there is a perception that Large Language Models are the first lap of an 800m race that ends in Artificial Intelligence, where I believe it is more accurate to think of Large Language Models as the raising of one foot on the start line of an ultra-marathon. Just as the invention of the hammer might claim to be the pre-cursor of the fighter jet, LLM will contribute to the creation of AI.

I don’t know that. I have listened to people like Cal Newport explain to me how LLMs work and why they are not going to launch any missiles.

I have listened to people from the industry make all sorts of hand-wavy gestures about how AI will do x y and z. Vague claims that crumble under examination.

Newport makes more sense.

Is AI exciting and interesting? Yes. Will there be great gains in efficiency and the like? Sure. Will we all lose our jobs in twenty four months?

No. We won’t.

People like you support my writing. Sign up for my free newsletter. Become a paid member, and get access to all my writing, including anything I publish, and a Member's Slack.