The next era of computing will be the Provably Correct

Y’all.

I have a confession.

I hate the modern incarnation of the computer. It’s worthless. More of a distraction than an asset. It’s nothing like what it’s supposed to be, and it’s actively being wielded against you. Against us.

It sucks up your time as much as humanly possible. Did you see that DT did something? Did you see that UN said stuff? UPDATE WINDOWS DEFENDER. Did you know that we offer a desktop application? CLICK HERE! FRee TeSTosterone.

Ugh.

Sorry. I’ve got one life on God’s green earth and dealing with that version of the computer ain’t it.

Everyone moment I spend not in front of a computer is a win these days (I’m currently writing this on a computer).

Current computing

I don’t think it can last much longer. People will realize eventually that there is no magic bullet. Yes, I’m also talking about AI (and also all computing).

I’m watching folks entertain multi-million dollar deals with no clear service level agreements (SLA), or programmatic access, or really clear deliverable. We’re turning over our expert time for doing real work to become beta testers for computing N.0.0 that no clear offering over the last one. Solely on the promise that it might improve our situation. We do know however that this version uses 200x more power than the deterministic version.

I gotta say: I’m skeptical based on the evidence I currently have. And I’m really actually trying to make it work, you know? I’m not really a hater, although it sure is easy to be one right now.

To what end are we “problem solving”? What problem are we trying to solve?

Has anyone realized that half of computing right now exists (supposedly) to make computing easier?? Is the irony of that lost on everyone??

Are we sure that more computer is the solution? It can literally only be the solution inasmuch as literal power remains relatively “cheap” (it’s actually not, we’re just all breathing the consequences and that’s difficult to include in a balance sheet). Every actual person I talk to has the same answer for what would help us humans more: more time; fewer problems requiring us to split our focus; useful help; and honestly, more pay to help provide some security in these uncertain time.

“We’re 10x more profitable (if you ignore the externalities and consequences).”

The problem isn’t us doing our job as much as it is trying to get rid of the distractions, IT problems, etc. I fail to see how the latest technology is actually helping us with that problem. Before AI, it was software as a service, before software as a service, it was the cloud, before the cloud, it was just “software”, and then buried waaay deep down underneath it all, you could just go learn how a computer works and actually make helpful things for yourself.

I’m struggling these days because everyone is trying to figure out how to make money off of the computer, but I have this deep suspicion that because bits are so darn copyable (the best part of computing) there’s not a real “value” that we have in our scarcity-focused world. You can’t have [this] because you have to pay first.

This is not something our economy quite knows how to deal with yet, I think/suspect.

The Provably Correct

There is a flavor of computing that most folks are not aware of, and I do not expect them to understand. Under the hood, this is all just logic, and logic is just math. Which means that we can use math to interrogate our programs and literally prove that they deliver us the intended results. It’s harder, and not all problems are suitable for that sort of interrogation, but when they are, computing can offer guarantees that we’re not currently used to having.

This isn’t like a “should I run on-prem vs cloud” type decision, but rather a fundamental questioning of what the computer is supposed to deliver to us. Is that computer you paid thousands of dollars for an app store for someone to deliver shitty software with an 80% markup (aside: never trust the numbers they quote you) and an ongoing subscription? Or is it like a hammer you bought at home depot that, when deployed well/correctly, delivers hundreds of thousands of useful blows to the problems you are trying to solve.

The provably correct begins at understanding the problem that we’re trying to solve. Which is harder than you might think.

Given a well-defined problem, we might be able to model it something like a deterministic finite state machine, or a pushdown automaton.

Ok, John, so what?

Well these things we can reason about. Formal language theory gives us a pathway through which we can claim things and then also check them. If you deliver me an implementation that does not meet the claim, then I can prove you wrong.

This applies to nearly all computing. And it forces us to define the problem in mathematical terms. If you’re aiming to apply a computer to a problem that you have, this is not a strictly required pre-requisite, but it’s one that you have the power to demand when the context calls for it.

IMO, anyone claiming to sell softare must be clear about where these boundaries lie and do not lie, otherwise they’re no better than snake oil. I mean this literally.

If they are not, then they are free to deliver you endless bullshit under the contracts you currently hold.

The real ones will be able to deliver this, because some of the most important mathematicians of our time also agree with me on this point (specifically Terence Tao and the idea of this which also kind of… I dunno, vaguely gross, but maybe better than everything else.

If you cannot provably verify your software, then at some point it will turn into rent-seeking. I would bet my whole bottom dollar on that.

Learning/Knowledge as the true primitive

Really what I see more than anything is folks that have enjoyed the era of “computers make my life easier for free” now bumping into the ugly reality that this shit ain’t free. Someone has to write it. Someone has to check it. Someone needs to prove it. If you leave it open-ended, they’ll start charging you increasingly more, forever.

There are people, organizations, physical servers, etc. that make it all tick. Literally TICK. Did you know that the square wave clock is among the most fundamental parts of a computer? GO LEARN IT.

I feel less concerned than others do about “AI” infiltrating everything (although believe you me: it’s a HUGE concern of mine), and more about our people (ALL people) not knowing enough about how computers work to detect when they are getting fleeced by a techno-feudalist; techno-capitalist; a techno-fascist. Complete with racism to boot.

It should actually speak volumes right now, to every individual: who is promoting “AI”? Not the underlying techology (i.e., diffusion models, GANs, architectures, etc.) but the concept of “AI” or “Machine Learning”? Hm.

It’s the same idea as someone who wants to understand trees versus someone who wants to capitalize on “forests” and “forestry” because we “have so much of it.” We don’t, and they do not understand trees, which is why they can feel free to cut them all down.

Don’t let them cut all of our trees down. Some of us still hope to be breathing in the next few decades/centuries.