Jumbled-up sentences show that AIs still don’t really understand language
By Will Douglas Heaven,
MIT Technology Review
| 01. 12. 2021
Many AIs that appear to understand language and that score better than humans on a common set of comprehension tasks don’t notice when the words in a sentence are jumbled up, which shows that they don’t really understand language at all. The problem lies in the way natural-language processing (NLP) systems are trained; it also points to a way to make them better.
Researchers at Auburn University in Alabama and Adobe Research discovered the flaw when they tried to get an NLP system to generate explanations for its behavior, such as why it claimed different sentences meant the same thing. When they tested their approach, they realized that shuffling words in a sentence made no difference to the explanations. “This is a general problem to all NLP models,” says Anh Nguyen at Auburn University, who led the work.
The team looked at several state-of-the-art NLP systems based on BERT (a language model developed by Google that underpins many of the latest systems, including GPT-3). All of these systems score better than humans on GLUE (General Language Understanding Evaluation)...
Related Articles
By Elizabeth Dwoskin and Zoeann Murphy, The Washington Post | 10.01.2025
MEXICO CITY — When she walked into an IVF clinic in June, Alin Quintana knew it would be the last time she would try to conceive a child. She had prepared herself spiritually and mentally for the visit: She had traveled to a nearby...
By Rob Stein, NPR | 09.30.2025
Scientists have created human eggs containing genes from adult skin cells, a step that someday could help women who are infertile or gay couples have babies with their own genes but would also raise difficult ethical, social and legal issues...
By Jessica Mouzo, El País | 10.03.2025
DNA is the molecule of life: this double-helix structure, present in every cell in the body and organized into fragments called genes, stores the instructions for making organisms function. It is a highly precise biological machine, but sometimes it breaks...
GeneWatch UK has prepared a briefing on the genetic modification of nature for the International Union for Conservation of Nature (IUCN) Congress in October 2025
The upcoming Congress claims to be “where the world comes together to set priorities and drive conservation and sustainable development action.” A major concern for those on the outside is that the Congress may advance plans to develop and encourage the use of synthetic biology in nature conservation. This could at first glance sound like...