2023-07-21 17:00:00
American researchers are interested in the catastrophic forgetting of artificial intelligences, which forget information as they learn new tasks. This phenomenon could become particularly problematic in a society that is increasingly dependent on AI.
You may also be interested
[EN VIDÉO] Machine learning, automatic learning Machine learning translates to “automatic learning”, an artificial intelligence technology that allows…
Artificial intelligences are often described as black boxes, because specialists do not really know what is going on there. Neural networks are trained on large amounts of data and have to perform certain tasks, but the resulting algorithms are too complex to analyze. However, all AIs face the same problem: catastrophic forgetting.
Researchers at Ohio State University have studied this phenomenon in an attempt to find solutions. As the AI learns new tasks, it forgets previously learned information. Catastrophic forgetting is particularly problematic for lifelong learning, and can become dangerous, for example if self-driving cars forget basic safety rules.
Learn distinct tasks early
Researchers have found that AI memory bears similarities to human memory. Humans may struggle to remember details of two similar scenarios, and more easily remember two very separate events. Similarly, artificial neural networks retain information better when the successive tasks are very different.
In particular, the researchers found that by learning a variety of tasks early, AI can more easily learn new information and similar tasks later. Researchers hope that analyzing the similarities between machines and the human brain will create AIs that can learn and adapt like us. They will present the results of their research at the International Conference on Machine Learning (ICML) taking place next week.
1690016764
#Catastrophic #forgetting #Achilles #heel #artificial #intelligence