O PTIMISM IS HARD to come by among those trying to build artificial intelligence safely. An abiding fear is that neural networks—the complex structures that underpin most modern AI systems—are largely inscrutable. Understanding how part of a neural network functions, in other words, does not always allow engineers to predict a model’s behaviour.
If models continue to become ever more capable without being easier to interpret, say researchers, the chances of apocalyptic malfunctions will grow too. A string of cyber-security incidents has caused websites to go dark To enhance your experience and ensure our website runs smoothly, we use cookies and similar technologies. Registered in England and Wales.
No. 236383 | Registered office: The Adelphi, 1-11 John Adam Street, London, WC2N 6HT | VAT Reg No: GB 340 436 876
Extract — continue reading at the source.
