Does he have a point?

https://www.youtube.com/watch?v=YhN4X56E7iM

I think he's right that AI is about to take math to a higher level
beyond what we can comprehend from our current point of view that is
most likely misguided and full of historical misconceptions.

The question should be, what's taking these silly AI researchers so long?

We don't even have AI yet that is able to come up with concepts from
scratch for toy problems like the game of go.

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Op 24/04/2024 om 12:17 schreef FromTheRafters:

sobriquet formulated on Tuesday :

Does he have a point?

https://www.youtube.com/watch?v=YhN4X56E7iM

I think he's right that AI is about to take math to a higher level
beyond what we can comprehend from our current point of view that is
most likely misguided and full of historical misconceptions.

The question should be, what's taking these silly AI researchers so long?

It's complicated.

We don't even have AI yet that is able to come up with concepts from
scratch for toy problems like the game of go.

Yes, but that failure, if it is one, is not because of the inability to complete a supertask in the real world. Let the essential workings of
the task *be* the representation.

I have seen some of his other presentations and he is, IMO, a good
teacher. He manages to teach contemporary mathematics while mostly
keeping his opinions of infinity as a side issue.

I think he is wrong about infinite representations. IMO we construct our notion of number as needed for our mathematics. We had to extend the
naturals to include zero, and had to further extend them to integers to accomodate addition's inverse operation. We then had to extend them to rationals to accomodate multiplication's inverse operation. Then to accomodate exponentiation's inverse, we invented the reals. There is
nothing wrong with having real number representations resemble supertasks.

I think one possible objection is that if we assume irrational numbers
like square root of 2 to exist, that presumes that we can in principle
compute it to arbitrary levels of detail. But from physics we know that
reality seems to be discrete rather than continuous for the most part.
So ultimately it seems that the universe is composed of an extremely
large but finite number of discrete elements, so however optimal we
represent things, we'll run out of elements if we try to spell things
out explicitly to unbounded levels of detail. A computation might
involve computing things to a level of detail beyond what we would be
able to compute given the limitations of the universe.

Just like if you want to store digits of the square root of 2 on a usb
stick. At some point you run out of space and you would need a bigger
usb stick to store more digits. With an infinite number of digits, you
would need a usb stick with infinite storage space and they don't exist
and it's unlikely they can exist. So it seems reasonable to claim that
the square root of 2 can't exist as an explicit number with all of its
digits at our disposal. It just exists as a concept (the diagonal length
of a square with unit side length) just like the concept of pi as the
ratio between circumference to the diameter of a circle, rather than the concept of pi as an infinite string of digits in some number system.

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

On 4/23/2024 7:59 PM, sobriquet wrote:

Does he have a point?

https://youtu.be/YhN4X56E7iM?t=120
|
| I think currently what we've done is
| we've managed to insulate ourselves from
| mathematical reality.
| And for me, that means
| being in touch with computation.
| I don't want mathematics to be
| an exercise in philosophy.

I consider
the verification of the validity of a proof to be
a computation.
I think that makes all or nearly.all of
what Wildberger considers problematic
what Wildberger sees going from strength to strength.

Am I outside the mainstream of thought on that?

My philosophy of mathematics,
boiled down to its essentials:

A claim is one of true or false.

In a finite sequence of claims,
if each claim is not.first.false,
then each claim is not.false.

A claim in a sequence might be externally verified
to be not.false (and thus, not.first.false)
by definitions, by axioms, by theorems
external to that finite sequence of claims.

A claim in a sequence might be internally verified
to be not.first.false.
Famously, Q in ⟨P P⇒Q Q⟩ is an example of this.
Is Q true? What does Q even mean?
Such questions don't matter.
Q is not.first false in ⟨P P⇒Q Q⟩

If Q is in a finite only.not.first.false sequence
then Q is true.
And I still don't know what Q means.

If
we see in front of us
a finite sequences of claims which is
only.not.first.false about each one of infinitely.many
then
each claim in that sequence is
true of each one of infinitely.many,
even the claims that are only internally verified
(the interesting claims, learned from that sequence).

We have not laid hands on each one of infinitely.many,
so I'm guessing Wildberger is troubled by my philosophy,
but we have "laid hands on" _each claim_

My philosophy is that logic about
differentiable manifolds or inaccessible cardinals
is primarily logic about finite sequences of claims.
And _the claims_ are accessible to finite beings.


There are many claims in physics about which
it would be perfectly reasonable to ask
how could we _possibly_ know that?

I think that my philosophy (hands.on claims)
answers that question.

A favorite example of mine is the size of
the cosmos, outside our observable universe
Starting from the assumption that
there is nothing unusual about our speck of dust,
we measure the (observable) curvature of the universe
and extrapolate outside the observable.

That method requires us to reason about
that which we cannot, even in principle, observe.
How can we possibly know that?

My answer is:
by assembling finite only.not.first.false
sequences of claims, which we can observe.

I think he's right that
AI is about to take math to a higher level
beyond what we can comprehend
from our current point of view
that is most likely misguided and
full of historical misconceptions.

I don't know. We'll have to see whether
the offspring of our minds slip free of our history.

Each generation of human children,
offspring of our minds and bodies,
struggle to slip free of their parents' history.
Each generation succeeds some and fails some,
I think.

If we don't build into our AI an ability to slip free
of history at least as well as human children,
then I think the AI mathpocalypse won't come,
but I also think that, if we don't do that,
this whole AI thing going on will be found
to be not worth the effort put into it,
and will eventually be thought a passing fad.

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

FromTheRafters <FTR@nomail.afraid.org> writes:

sobriquet explained :

...

I think one possible objection is that if we assume irrational
numbers like square root of 2 to exist, that presumes that we can in
principle compute it to arbitrary levels of detail. But from physics
we know that reality seems to be discrete rather than continuous for
the most part.

Indeed! I feel that a number's value is in its formula and it matters
not if other representations fall short of philosophical goals.

Funny, and good point, that our mathematical model of reality achieves
better results if it is done with discrete quanta rather than
continuous values. God may not throw dice, but it sure looks like he
plays with blocks.

Yeah, and that 1x1 block, when sawn in 2 diagonally, or even just
suitably marked, gives you a physical sqrt(2).

Phil
--
We are no longer hunters and nomads. No longer awed and frightened, as we have gained some understanding of the world in which we live. As such, we can cast aside childish remnants from the dawn of our civilization.
-- NotSanguine on SoylentNews, after Eugen Weber in /The Western Tradition/

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)