From Newsgroup: sci.math

Hi,

AI Laptops are technically evolving so much,
the operating systems cannot evolve fast enough.
Its like the cars in Bali geting bigger and bigger,

but the streets are still narrow or bad. Quite
a stunner was the RTX sparc announcement May 2026
that might put 1 PETA OPS AI Inferencing, or will

it be machine learning, into your hands for $2000 ?
There are a lot of things going on, like ARM SME2
unlocking even Apple Silicon for everybody, and

Nvidia countering Apple Silicon with its own unified
memory convential memory. What will be the future? I
guess we might see Laptops having high bandwidth memory:

TSMCrCOs CoWoS (Chip-on-Wafer-on-Substrate) is
an advanced 2.5D packaging technology that
vertically and horizontally integrates multiple
distinct silicon dies
https://www.nomadsemi.com/p/tsmcs-cowos-capacity

But did you know that BYD started as a battery
company and has now a semincoductor branch, already
operating a 4nm? BYD electronics spinning off a laptop?

MidAmerican Energy Holdings, a subsidiary of Warren
Buffett's Berkshire Hathaway Inc, invested about
US$230 million for a 9.89% share of BYD
at HK$8 per share.
https://www.nomadsemi.com/p/byd-semiconductor-deep-dive

Bye

Mild Shock schrieb:

Hi,

One challenge is that the example has
not only anti-unification, like the c*d\=d*c,
the Mc Cune style model finder implementation

makes also heavy use of backtracking. So
I even donrCOt know whether ISAMORE papers
with saturation would help:

Finding Reusable Instructions via E-Graph Anti-Unification
Youwei Xiao et al. - ASPLOS rCO26
https://github.com/pku-liang/ISAMORE

But I guess ISAMORE also delivers
rewriting challenges. But their subject
matter in the above paper is LLVM/GEM5 (*),

so maybe so maybe I rather buy a new Mac
Neo (only $500) and enjoy local AI.

Bye

(*)
https://en.wikipedia.org/wiki/Basic_Linear_Algebra_Subprograms#Level_3

Mild Shock schrieb:

Hi,

Now toying around with a model finder,
for some rewritng problems. Not bad, I
didn't see that before on my computer:

?- time(mccune([e*X=X,X^(-1)*X=e,X*(Y*Z)=(X*Y)*Z, c*d\=d*c], 6, _)).
% Zeit 166.140 ms, GC 0.000 ms, Lips 18616 k
true

?- mccune([e*X=X,X^(-1)*X=e,X*(Y*Z)=(X*Y)*Z, c*d\=d*c], 6, R).
R = [d-2, c-1, 5*2-3, 5*5-4, 4*3-2, 3*5-2, 3*2-5, 4*4-5, 2*4-3,
2*3-4, 3*4-1, 3*1-4, 5*3-1, 4*1-3, 1*5-3, 1*3-5, 2*1-5, 2*5-1,
5*1-2, 4*2-1, 1*4-2, 1*2-4, 5*0-5, 4*0-4, 4*5-0, 5^ -1-4, 5*4-0,
4^ -1-5, 3*0-3, 3*3-0, 3^ -1-3, 2*0-2, 2*2-0, 2^ -1-2, 1*0-1,
1*1-0, 1^ -1-1, 0^ -1-0, 0*5-5, 0*4-4, 0*3-3, 0*2-2,
0*1-1, 0*0-0, e-0]

But still not as good as Mc Cune, wo reported
0.010 seconds , so 10 ms (*). I guess I need to
explore more heuristics. But will it include

union find? The payback of union find in such
search examples is so low. If |D| = n is the
domain size, then although |D x D| has size n^2,

the diagonal E_D, namely E_D = { (x,y) | x e D,
y e D, x = y } has only size n. So the probability
of a pair in certain problem domains, is around 1/n,

getting smaller and smaller with larger and larger n.

Bye

(*)
https://www.mcs.anl.gov/research/projects/AR/mace4/July-2005/doc/mace4.pdf >>

Mild Shock schrieb:

Hi,

This is a nice example how the young generation
doesn't understand systems that were developed
30 years ago:

Example 1:
No clue that SWI-Prolog does have JIT indexes,
which explain mysterious differene between
cold and warm run.
https://arxiv.org/pdf/2511.13600

Example 2:
No clue that subsumption could make a difference,
claims left recursion is fastest, with referene
to [TL10] and a bogus node argument.
https://arxiv.org/abs/2504.21291

Maybe the Prolog eduction group should organize
a summer school about tabling?

Sad news: It will not help to keep up with
the AI boom. Whats behind the AI boom and
computing transitive closures?

Bye

Mild Shock schrieb:

Hi,

Ok, thats noice:

"On a GPU cluster, the ratio of high computing
power to communication bandwidth makes scaling
breadth-first search (BFS) on a scale-free
graph extremely challenging."

Scalable Breadth-First Search on a GPU Cluster
Yuechao Pan, Roger Pearce, John D. Owens
2018 - Lawrence Livermore National Laboratory
https://arxiv.org/abs/1803.03922

Sniff, still passed the deadline Q1 2026 for
my GigaLIPS Prolog. There were so many other
improvements , more low hanging fruits to cover.

Bye

P.S.: Wanna rent some GPUs?

Cloud GPU Index
Compare 2,542 GPU prices across 57 cloud providers
https://getdeploying.com/gpus

Mild Shock schrieb:

Well its called "artificial" intelligence.
Like "atificial" flowers made from plastic
try to share the look and feel of natural flowers,

articial intelligence tries to share the
look and feel of natural intelligence.
AI is not really genetic engineering of

bioforms, right? However, there is a research
frontier sometimes called biocomputing or
synthetic biology AI.

Kunstblumen
https://www.xxxlutz.ch/kunstblumen-C22C8C1

The convergence of AI and synthetic biology: the looming deluge
https://www.nature.com/articles/s44385-025-00021-1

x schrieb:

On 3/7/26 03:08, Mild Shock wrote:

Hi,

Resolution of Erd-Yos Problem #728
We provide a writeup of a resolution of Erd-Yos
Problem #728; this is the first Erd-Yos problem
(a problem proposed by Paul Erd-Yos which has
been collected in the Erd-Yos Problems website [3])
regarded as fully resolved autonomously by an AI
system. The system in question is a combination of
GPT-5.2 Pro by OpenAI and Aristotle by Harmonic,

I guess there is something called 'logic' and then
there is something called 'semantics'.

Does it matter if words have meaning?-a If the
answer is no - then of course you can generate
the look and feel of meaning through 'AI'.

--- Synchronet 3.22a-Linux NewsLink 1.2