Speaking of Artificial Intelligence
The Corner Office
“Skynet may take my body, but it won’t take my spirit” William Wallace
The word artificial has multiple meanings; among them are fake, specious, and simulated. When speaking of artificial intelligence all three of these meanings may be applied. Two synonyms for specious seem, at first glance, to be at odds with one another; deceptive and plausible. At present AI is developing rapidly. There are some that say AI is being developed too rapidly and as such are looking for a pause button to slow implementation down while proposed legislation sets up guard rails on future development. If you look closely at the Venn of the words deceptive and plausible you can see why guard rails are needed. In a recent interview between Scott Pelly and Sundar Pichai, the following was discussed: “In 2023, we learned that a machine taught itself how to speak to humans like a peer. Which is to say, with creativity, truth, error and lies (emphasis applied). This technology, known as a chatbot, is only one of the recent breakthroughs in AI – machines that can teach themselves superhuman skills. Sundar Pichai further said, “…that Google did not fully understand how its AI technology produced certain responses. ‘There is an aspect of this which we call, all of us in the field call it as a ‘black box’. You know, you don’t fully understand. And you can’t quite tell why it said this, or why it got wrong.’” Is it wise to produce such a powerful tool that has a non-deterministic output or is able to combine deception and plausibility in such a manner as to make it indistinguishable from truth? In the medical device field one major issue with the FDA and OEMs alike is ensuring that AI algorithms are not biased toward various patient groups and that machine learning does not create unequal outcomes relative to the treatment of disease. Question – if we don’t, in fact, fully understand how AI produces certain responses, how can we be assured that the output from an AI/ML algorithm is not deceptively plausible? Sleep well dear readers.
Until Next Month,