[Fis] AI Armchair Experiment?

[Marcus Abundis]

Greetings all,

>From Pedro:
<My initial reading in the screen, was of open mouth and incredulity (and
some mistrust realizing that every response starts by a kind and flattering
consideration on the question itself). After printing and re-reading, . . .
>

— I am a bit surprised by the apparent acclaim(?) given to LLMs. How does
one not see this is all merely as an extension of Shannon Signal Entropy:
one starts with 'entropic base/binary terms', moves on to 'entropic words',
entropic phrases, entropic passages, etc. all *statistically*
massaged/'trained' by presumably 'intelligent beings'. The only real
advance here is in the statistical (i.e., Shannon) twists and turns used,
and advances in database indexing. There is nothing intelligent here, no
actual science. To put the whole AI issue in a more correct scientific
perspective I offer you this talk from Yann LeCun at the 2025 World
Economic Forum. THIS is where the real future (and heavy lifting) lies with
AI and what I focus on:
Yann LeCun at WEF 2025 <https://urldefense.com/v3/__https://youtu.be/8RJb5N-wqNk?si=zhEfddPSHFVpDOcJ__;!!D9dNQwwGXtA!R2kpXP9-1xwIzHQfHcZhhdluo9c2ozRgJMFXbo8S9kCsB2smC-4Z_xFFsPn_4_kG3dMUN8Wvx348LvYF$ >

Pedro – <Could we organize . . . anyone willing to address one of
those aspects,
send me a working title>
— Really?!?! I have shared with you and the FIS group various steps I have
pursued in addressing these CENTRAL issues – where *you* dismissed early
work as "free wheeling speculation". I have NOW had multiple paper reviews,
with papers on this accepted to NeurIPS 2024 and ICLR 2025 (#1 and #2 AI
conferences), where one reviewer even commented on "congratulations on this
remarkable work" So do you REALLY want to explore this?!?! You don't have
to read more than the abstract
Submission to ICLR 2025 Workshop: Agentic AI in Scientific Discovery
<https://urldefense.com/v3/__https://drive.google.com/file/d/11r5LHrzAV2wAPXAPbEMxO58GuTR48hdP/view?usp=drive_link__;!!D9dNQwwGXtA!R2kpXP9-1xwIzHQfHcZhhdluo9c2ozRgJMFXbo8S9kCsB2smC-4Z_xFFsPn_4_kG3dMUN8Wvx0VIQUK9$ >
Sure, why not . . .

 Marcus
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listas.unizar.es/pipermail/fis/attachments/20250330/3661728d/attachment-0001.html>