11 Comments

Yes, we do need to vet abstractions more. Just taking the word of aged polymaths isn't the way to vet them though. History is replete with aged polymaths, respected by their intellectual communities, who got most everything wrong. Pre-20th-century medicine was run by such polymaths, as was pre-17th-century physics.

The lesson we can take from pre-20th-century medicine and pre-17th-century physics is that it is not enough for an academic community to be respected and prestigious, for it to be right. Respected and prestigious communities can get lost in games of ideological fashion. We also see this with the more recent replication crisis. The only cure to "ideological fashion" is a closer connection to something systematically checkable, independent of fashion - experimental data, math proofs. That's how medicine and physics became modern.

In conclusion, when it comes to vetting ideas, normies are bad and ruled by fashion, autistics are good.

Expand full comment

This post is hyper-abstract--the concept of abstraction itself completely omits detail. And it seems there is no end to the sequence of levels of hyper(-to-the-nth-power-)abstraction that one might attain, with abstractions about the concrete, abstractions about these first-level abstractions, abstractions about those second-level abstractions, etc. ad inf. (But returns may already be diminishing at this first level).

Expand full comment

I'd guess that's deliberate - to make an argument in the abstract so as to avoid the reader being distracted by their allegiance to specific examples is generally very good epistemic practice.

Relatedly, mathematics has time and time again revealed there are real and often unexpectedly powerful advantages to investigating ever higher orders of abstraction.

Expand full comment

I have found examples to be very effective validating tool for abstract reasoning. This to the point that I challenge myself to drop in examples in support of my reasoning. By the same token ask for examples when listening to others.

Abstract and concrete their respective parts to play in our sense-making and skilful and conscious toggling makes a lot of difference. I'd venture a thought that this type of toggling is often driven by implicit rather than explicit decisions, alas. A bit more on this topic in the link below

https://medium.com/@piotr-28397/implicit-decisions-743097d6257c

Expand full comment

Which AI risk abstractions are you skeptical of?

Expand full comment

One unvetted (and usually unstated) assumption in AI risk discussions is that advanced AIs will have a strong "will to live". Therein lies much of their danger.

This is an unvetted abstraction. True, every biological intelligence has a strong will to live owing to its evolutionary origin (survival of the fittest). But we have seen no examples of GPT-4 or any other such system "caring" whether it remains switched on, and there is no reason to believe that a will to live holds water as an abstract truth about advanced AIs.

Expand full comment

Hello Robin Hanson, You said:

There is a pyramid of ascending abstractions in our mental organization, if we take time to notice them. And that, we could know when to think at what level of abstraction. (With practice.) We could look with caution to our reliability of inferences, and if they look good, transfer sometimes by analogy.

Or, not considering abstractions directly: The mind is a connecting organ; it works only by connecting, and it can connect in an indefinitely large number of ways. Words are meeting points at which regions of experience come together; a part of the mind’s endless endeavor to order itself.

You note that, One can seek examples, and test which abstract claims apply to them. Unfortunately, that is the one thing that you did not do. I am supposing that you could have tossed out 5 -6 abstract notions, and analyzing in the other direction, let us test them with our view of reality.

You do say: Remember: reality is actually concrete; it is only thoughts that are abstract. (Though yes, thoughts are real.) Thoughts are real, but it is the content of thoughts that we are interested in. A thought can navigate anywhere in the known universe, and beyond, but you and I live on planet earth and in a society that needs our understanding.

In fact, abstractions are comfortable, because there are no people in them, no one to get upset. You might have collective nouns, like humanity, but Mr. Humanity is not going to knock on your door. The most important thing about abstractions, is if you can find someone to pay you a salary for spinning them. (Please ask them if they would pay me a salary for considering them.) No, they can be useful, but only if you vet them, as you say.

You mention contradictions. I take them to be our persistent "problems". What are they, and where do they come from? Are contradictions real or abstract? Are contradictions resolvable, or are they just Koans to spend the rest of your life on? Incidentally, I didn't find any contradictions in this piece. Can contradictions be nested, or hidden in abstractions? If so, how do we get them out of there.

I think that you've hit on something that begs further investigation.

.

Expand full comment

The concrete representation of the word "vet" employed in this essay distracted me to abstraction.

Expand full comment

> Remember: reality is actually concrete; it is only thoughts that are abstract. (Though yes, thoughts are real.)

Ontologically, are thoughts not a part of reality? if not, what are they a part of (or, where is it that they are)?

Expand full comment

Thoughts contain concrete expressions of abstractions.

Expand full comment

They do indeed...similarly, the airspeed velocity of an unladen swallow is roughly 20.1 miles per hour. However, neither of these facts address the specific questions I posed! ;)

Expand full comment