Sujet : Weekly AI Top Ten List (Re: Divide, Conquer and Loose [Limitations of AI])
De : janburse (at) *nospam* fastmail.fm (Mild Shock)
Groupes : sci.physics.relativityDate : 11. Feb 2025, 21:48:17
Autres entêtes
Message-ID : <vogd2g$3nk3$2@solani.org>
References : 1
User-Agent : Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:128.0) Gecko/20100101 Firefox/128.0 SeaMonkey/2.53.20
Hi,
1. Russia: Stealing data and blackmailing companies
already for years, now selling the data all over
the globe for pre-training
2. China: Has the most inovative heads, that
can sequence a LLM on a finger nail
3. India: Gifted hackers that can write DeepSeek
via Unix sed, 15 year olds showing machine learning
videos already 10 years ago
4. Europe: Everything invented anyway here
5. USA: Sam Altman and Elon Musk having a silly feud
Bye
Mild Shock schrieb:
Hi,
Now history repeats itself. There are
many pseudo software engineering program
verification papers, that say, just chop
up a problem, and solve each part separately,
and put the things together. Which is utter
nonsense. Try this with a SAT Solver,
the smallest unit is the propositional variable
you cannot solve it independently, only
"try" 0 and 1, eh voila you are back to
NP Complete. Thank god we have Christos
Papadimitriou. According to Wikipedia he is
is the author of the textbook Computational
Complexity, one of the most widely used
textbooks in the field of computational
complexity theory. Might be the cure for
any singularity and AGI dreams:
Chatbot Software Begins to Face Fundamental Limitations
Recently, computer scientist Binghui Peng and
his team proved mathematically that there may
be a hard limit to LLMs’ compositional
task-solving abilities
https://www.quantamagazine.org/chatbot-software-begins-to-face-fundamental-limitations-20250131/ Bye