Friday, May 17, 2019

"I Am My Brain"

   Richard Restak, M.D.. The Brain has A Mind of its Own (1991) A collection of essays originally written for Newsday and other public forums. As such, they are short, often lack nuance, and too often express far more certainty than the evidence warranted. Overall, a good refresher on neuro-biology, documentation on the state of neurology in the 1980s and ‘90s, and an interesting reminder of a time when experts weren’t afraid of saying things that these days would be heard as triggers or worse.
     Restak’s stance is that we are our brains, which implies that our Western and especially American insistence that we are somehow exempt from biology. His sometimes testy remarks about the interplay between nature and nurture should correct the widespread notion that we can be anything we want to be. Restak was also a psychiatrist, and his anecdotes about his patients and his helplessness show that there are limits to the best-intentioned attempts to help.
     I learned a few things. His comments on the limits of artificial intelligence (AI) appear at first reading to miss the point, since he wrote before neural networks so spectacularly demonstrated the power of computing. But recent equally spectacular failures of neural-net AI support his contention: human (and animal) brains are self-modifying in response to new experiences that don’t compute. AI can’t do that. Once trained, AI stubbornly resists further training. Eg, offered a mildly altered image of an orange, the AI asserts it’s a power drill. AI cannot do what biological brains do: recognise objects in a context. They can only recognise statistical patterns, so that a handful of differences in one region of the pattern prompts an incorrect recognition.
     It may be possible to build some self-modification into AI, but that won’t eliminate the fundamental limitation: it’s domain-specific. AI can deal only with the data that it’s trained on. Brains can somehow apply what’s been learned in one domain to another domain. Humans are especially good at this: mathematics is the abstraction of patterns that apply to multiple domains, which may explain the “unreasonable efficacy of mathematics”, as Wigner called it.
     Worth reading. ** to ***

No comments:

Four ordinary people: Quartet in Atumn (Barbara Pym)

Barbara Pym. Quartet in Autumn (1977) Pym seems to be a nice lady who tells stories of nice and not so nice people of little consequence. T...