A Philosophy Of AI That Works

Almost no one asks philosophies to be functional- it's like them not having any measurable impact is a feature. I found that weird, I am a pragmatist by instinct. I always felt your thinking is as good as what it produces.

And I can tell that the philosophy that is being to AI in 2026 does not work.

Let me take a little detour to my favorite unpopular idea- consider that there is a type of intelligence that is beyond your ordinary thinking. Everyone at least occasionally experiences types of experience that are unlike ordinary thinking.

You could, of course, insist that there is nothing special about these situations, but I feel there is much more explanatory power in acknowledging that there is a difference in quality.

• Mathematic intuition
• Fiction writers observing, not making up, characters in imagination
• Taking a walk, then having a solution in your head fully formed that previously eluded you
• Staring at a difficult problem without analyzing it and starting to understand it
• Requiring a previously unknown concept with very many constraints, and arriving at just the thing by free association
• Arriving at a superior engineering solution while looking for aesthetics
• Picking something up for no apparent reason and noticing and fixing expensive oversight
• Meeting the right person at the right time in the unlikeliest of ways

I feel that the idea of a higher intelligence was thrown out of public discourse for good reason- it was the domain of religion, and religious philosophy has had a tendency to develop in ways far removed from practical needs, because everything important was already settled whatever the book said. So, in a way, completely removing the idea of a higher power was a win- now, we had to do the work.

On the other hand- the experiences I noted are near universal, it just became habitual to shrug them off. But just pick one and really look at it and you will see how remarkable it really is. How can that be? And, more to the point- how can I make good use of this?

I find the best way to make use of this idea is cycles of intense thinking and distraction. If I clearly set up a problem space in my mind and just hold it- which is quite uncomfortable- and then go move around or do something unrelated for a while- I find I can, more often than not, cause these large chunks of solutions appear in my mind.

This is not 'thinking' in the conventional sense, I'm not logically working fro assumptions and deriving conclusions. It's working an blocks of thought, words or units.

Now- that's a very interesting philosophical view of technical work, but what does it have to do with AI? Well, I had to take a few paragraphs of spelling out what I'm talking about because this is just not that much of a normal part of discourse. It's all over the place in the more 'right side of the brain' persuits- but not among people who consider themselves 'logical' or 'rational'. Being logical or rational seems to mostly imply a relatively linear thought process that is considered to be certainly generated by the physical brain. Which is odd, because it is clear we know almost nothing about the brain or about thinking. So whether you agree or not that the type of thinking I am speaking of is qualitatively different- we can surely agree that there is that pattern.

So. What I'm getting at- AI doesn't have a higher quality of thinking.

AI can take a bunch of stuff and calculate an approximation based on interpolation, but no matter how far you take it, no matter how much data you feed it- it's never quite on point. AI thinking is not sharp, ever.

That doesn't mean it isn't useful. Most programmers have never written a sharp line of code in their life- and arguably you shouldn't, becuase it's supposed to work. If you want to send someone a particularly eventful cat video, does it matter if that cat is artistically just right for the type of vision you had in mind, like a serious work of art? Of course not. That's why we call it slop. Sloppiness is inherent.

Take any observation of any living creature as a reference point. You get all sorts of imperfections and various approximations of an ideal form. You might even get drowsiness or fatigue or illness or even death, but you never get 'uncannily off the mark'. Good AI videos have so much human input, you can only consider the generated data a raw material. The spirit is the author's.

And yet in spite of this tremendous difference of a living being's thinking and a computer's trained calculations, there is the serious search of higher thinking in the machine. How could so many so intelligent people go down this rabbit hole? It's a red herring, a dead end, a pitfall to walk into. Why?

My answer is that it's a direct effect of denying the idea of the hunch. If you insist it doesn't exist, you unlearn recognizing it- and you can tell the difference or don't know to pay attention to the most significant. And so you go looking for the divine in the data. Whoops.

Let me suggest an alternative. It's a very simple human model, and yet it maps extremely neatly to what we've been discussing. It comes from a traditional worldview, and yet it's remarkable how well it maps. Judge it pragmatically- does it help you structure your thinking in a way that lets you fully make use of AI without going nuts? Great! Must be true. Here it goes.

There is a higher level of human thinking- it's mostly hunches, and it requires distracting or stopping analytical thinking for a while to receive it. You can recognize it by receiving more information that you can readily understand. You might have to write down the concept and really consider it to actually understand it, and it might be understandable in a new way when you return to it.

Then there's selective thinking- chosing what to do next, and doing it. Agency and choice.

And finally, there is memory and organization of memory- the human database.

So from this perspective, let me offer two observations. Judge them on their merit. How reasonable are these projections?

1. AI is stuck on level 3 thinking. It's a fancy, interconnected database. That is useful, but it's not more than that. It's 2026 postgres.

2. AI philosophy is stuck on level 2 and 3 thinking, with the hope the machine will bring about level 1, which will never happen.

So the conclusion is- use your humanity. An AI has no inspiration, and an AI has no agency. If whoever is in charge of your employment is underestimating the importance of that, if anything, it might be a nudge to change that.