What if Tony Stark is a better model for using AI than Sam Altman?

It’s amazing how many professional conversations I have these days that turn to AI. When people ask how I use it, how I think it should be used or how I see it being used, I usually offer up the example of Tony Stark and JARVIS in the Iron Man and Avengers movies from the Marvel Cinematic Universe.

I know. Geek alert. But bear with me.

In these movies, J.A.R.V.I.S. — Just A Rather Very Intelligent System — is Stark’s machine intelligence. Always on. Vast computational capability. Information, analysis, modelling, systems and execution.

Importantly though, Tony remains Tony.

JARVIS doesn’t replace Stark’s curiosity, imagination, judgement, ambition or personality. He enhances Stark’s ability to use them and what he can get out of them in the time available.

They’re not competing intelligences. They’re complementary intelligences. And that matters.

JARVIS is useful precisely because he isn’t another Tony Stark. Human intellect and machine thinking bring different capabilities and tolerances. They understand where each other is useful and, importantly, where each other is an irritant.

JARVIS doesn’t simply obey. He questions, challenges and occasionally disagrees. There is friction in the relationship. But it is useful friction.

Stark and JARVIS aren’t two people doing broadly the same job. They’re doing different things, in different ways, towards the same outcome.

Fantastical as the example is, this increasingly describes how I work with machine thinking. Not asking it to think instead of me, but introducing another kind of thinking into my own process.

I question it. It questions me. I introduce context. It finds connections. I challenge its conclusions. It challenges mine. Each iteration changes what happens next — and all this happens incredibly quickly.

This isn’t artificial intelligence replacing human intelligence. It’s symbiotic, harmonious enhancement.

This matters because much of the AI conversation is still framed around substitution. What jobs can AI do? What tasks can it automate? When will it become better than us?

There are legitimate questions around all of those things. But another question interests me more:

What becomes possible when human and machine intelligence work together positively?

The machine informs, challenges, calculates, synthesises and generates. The human brings intent, judgement, values, taste and accountability.

Of course, machines can increasingly simulate aspects of judgement and taste.

But should they?

Just because we can hand something over doesn’t mean we should. If the value of combining human and machine intelligence comes from their differences, making one behave increasingly like the other may be missing the point.

JARVIS doesn’t need to become Tony Stark. And Tony Stark doesn’t need to become JARVIS.

Using AI badly risks making us intellectually lazier. Using it well should make us more capable thinkers and allow us to do more with what is distinctly human.

I’ve always been a futurist and, in this future we are all creating, I’d like to see much less effort spent trying to make machines into artificial humans and much more spent understanding how their differences can enhance our capabilities.

Perhaps the most useful question about AI isn’t what it can do without us.

It’s what we can do together.

--
From the original on Substack.