This is a news story, published by Quanta Magazine, that relates primarily to Einstein news.
For more Einstein news, you can click here:
more Einstein newsFor more Ai research news, you can click here:
more Ai research newsFor more news from Quanta Magazine, you can click here:
more news from Quanta MagazineOtherweb, Inc is a public benefit corporation, dedicated to improving the quality of news people consume. We are non-partisan, junk-free, and ad-free. We use artificial intelligence (AI) to remove junk from your news feed, and allow you to select the best tech news, business news, entertainment news, and much more. If you like this article about Ai research, you might also like this article about
compositional reasoning tasks. We are dedicated to bringing you the highest-quality news, junk-free and ad-free, about your favorite topics. Please come every day to read the latest sophisticated tasks news, reasoning news, news about Ai research, and other high-quality news about any topic that interests you. We are working hard to create the best news aggregator on the web, and to put you in control of your news feed - whether you choose to read the latest news through our website, our news app, or our daily newsletter - all free!
artificial intelligenceQuanta Magazine
•Technology
Technology
87% Informative
Machine learning models have limited ability to solve Einstein’s puzzle or riddle.
The Allen Institute for AI recently set transformer-based large language models, such as ChatGPT, to work on such tasks.
The results were so powerful that the models seemed, at times, capable of reasoning.
GPT-3 failed when asked to answer bigger versions of the puzzle compared to the ones it was fine-tuned on.
The team observed the same pattern when it came to solving Einstein ’s riddle.
Some compositional problems will always be beyond the ability of transformer-based LLMs, researchers say.
A new technique known as chain-of-thought prompting can give an LLM a newfound ability to solve more varieties of related tasks.
As a result, the model could be trained on 20 -digit numbers and still reliably (with 98% accuracy) add 100 numbers.
But, Ye cautions, their result does not imply that real-world models will actually solve such difficult problems.
VR Score
92
Informative language
93
Neutral language
60
Article tone
informal
Language
English
Language complexity
51
Offensive language
not offensive
Hate speech
not hateful
Attention-grabbing headline
not detected
Known propaganda techniques
not detected
Time-value
long-living
External references
12
Source diversity
7
Affiliate links
no affiliate links
