John Martinis

You know, the way you make these GPUs or something.

And we think, you know, when we get that to work, we can scale up very rapidly.

So it's a 10 year time scale, something like that.

There may be.

And there's things we can maybe do, modeling and the like.

We also think what we can do is use the quantum computer and AI together to solve the problems better.

So that's what our theory team is proposing.

I used to work with Google Quantum AI.

That's what they're proposing.

So there's a general feeling of that.

My particular view, though, is that in terms of this control, if you don't build your system cleanly enough and, you know, that the control is clear enough, you're not going to get the great performance out of it.

So I'm a little bit old school here and working on, you know, building it that way.

There's certainly some elements where you can use AI.

you know, in the decoding circuit for the error correction and the like.

But the one thing to mention to you is that, you know, these qubits are naturally very noisy and you can maybe do

Sometimes 100 for bad qubits and maybe a thousand, maybe few thousand operations before they kind of lose their memory.

You know, you can think of it as like dynamic RAM where you have to refresh it.

Well, you have to refresh it with error correction.

And because of that, you're talking about a million qubit quantum computers.

to be general purpose and solve really hard problems.