aktuel
a year ago
The paper confirms that even the biggest and most advanced models can still be misled by adding irrelevant information that can be easily filtered out by humans.
Understanding the Limitations of Mathematical Reasoning in Large Language Models
10 pointsposted a year ago
by iforgotmysocks3 Comments
belter
a year ago
"The question of whether a computer can think is no more interesting than the question of whether a submarine can swim." - Edsger Dijkstra
msoad
a year ago
A calculator can "think" is "AI" and however you want to frame it. Reasoning is a very specific and defined concept. Computers can not reason (per this paper)