Policy need not wait for consensus on AGI

Chiefs of global companies, including Elon Musk and Masayoshi Son, and artificial intelligence (AI) experts such as Geoffery Hinton and Yoshua Bengio believe the phenomenal growth of generative AI models indicates that machines will soon think and act like humans, a trend we refer to as artificial general intelligence (AGI).

- 

But they are also concerned that researchers are unable to fully understand how these unsupervised algorithms perform tasks like creating new content, be it audio, code, images, text, simulations or videos. Further, these models can plagiarize, be biased, potentially replace thousands of routine jobs, and also pose security and privacy risks. Self-supervised at most, deep neural networks train on humongous amounts of data, but they learn on their own with no human programming or oversight. Speaking at the recent SoftBank World corporate conference, Son underscored that AI is now “self-learning, self-training, and self-inferencing, just like human beings." However, while researchers understand the mathematical operations of each neuron, it is a challenge to comprehend the connection between the maths and the wide range of observed behaviours. This makes it difficult for them to diagnose failure patterns, provide remedies, or even certify a model’s safety.