Lunch at the Labs

Various web links I found to be of interest recently.

• Play Deprivation Is A Major Cause of the Teen Mental Health Crisis

  Allowing more unsupervised free play is among the most powerful and least expensive ways to bring down rates of mental illness -- Jon Haidt and Peter Gray

  Two Very Well-Established and Disturbing Facts.

  1. Children's freedom to play and explore has declined greatly over the last half-century.
  2. Children's mental health has declined greatly over the last half-century.

  Too many people are focusing on drugs and therapy, as if something is wrong with the kids that needs correction, and not enough of us are thinking about prevention. Prevention would involve bringing normal childhood back to children. Children are designed to play and explore and thereby becoming increasingly independent as they grow older. Their instincts tell them that something is seriously wrong if they don't have such independence.
  ...
  Research shows that people of all ages who have a strong internal locus of control (internal LOC), that is, a strong sense of being able to solve their own problems and take charge of their own lives, are much less likely to suffer from anxiety and depression than those with a weaker internal LOC. Obviously, however, to develop a strong internal LOC a person needs considerable experience of actually being in control, which is not possible if you are continuously being monitored and controlled by others.

• Aging brains benefit from regular exercise, good night's sleep, study finds

  A new study found that people who were more active, but slept less than six hours on average, had faster cognitive (mental) decline. After 10 years, their cognitive function was equivalent to that of their more inactive peers.
  ...
  One limitation of the study is that participants self-reported on their sleep duration and physical activity.

• Hearing aids may cut risk of cognitive decline by nearly half

  A large study showed that older adults with a higher risk of dementia may be able to reduce their risk of cognitive decline by almost 50 percent by using hearing aids.

  Whether hearing treatment reduces the risk of developing dementia in the long term is still unknown.

• Diagnostic medical errors are a huge problem. Will A.I. come to the rescue?

  A new study estimates ~800,000 Americans are permanently disabled or die each year from diagnostic medical errors. Concurrently, multiple reports look at ways this may be reduced, including artificial intelligence. In this edition of Ground Truths, I'll review the scope of the problem and prospects for improvement that are desperately needed. Eric Topol

• The Psychological Depths of Rock-Paper-Scissors

  The reason rock-paper-scissors is not a purely arbitrary game, and the reason that an excellent player will win more often than chance would predict, is that human psychology is not random, and some behaviors are - not necessarily predictable, but likely to occur more often than chance would dictate.
  ...
  One heuristic of experienced "layers is "Losers lead with Rock." This is demonstrably true; naive players will lead with Rock more often than one-third of the time. Your hand begins in the form of a rock, and it is easiest to keep it that way. The name of the game begins with "Rock," and if you are mentally sorting through the options, it is the first one that will occur to you. And the word "rock" itself has connotations of strength and immovability.

• Uri Geller is Not A Magician

  The New York Times continues Geller's rehabilitation tour, surreally romanticizing his decades of harmful deceptions.

  Geller's enduring presence and rehabilitation isn't a testament to his genius, although he may well be a genius (or at the very least, a highly socially intelligent performer with no scruples!). Instead, his ability to remain in the headlines and to charm new generations of unduly credulous reporters is a disturbing reminder of our collective susceptibility to deception. Geller's stubborn refusal or inability to morally develop, to step out of his deceitful persona, is not laudable. It's tragic. As likable as I found the guy, I don't think his reputation should be salvaged, it should serve as a cautionary tale.

Some recent items related to Artificial Intelligence (AI) and Large Language Models (LLM)

• Why transformative artificial intelligence is really, really hard to achieve

  We think AI can be "transformative" in the same way the internet was, raising productivity and changing habits. But many daunting hurdles lie on the way to the accelerating growth rates predicted by some.

  1. The transformational potential of AI is constrained by its hardest problems
  2. Despite rapid progress in some AI subfields, major technical hurdles remain
  3. Even if technical AI progress continues, social and economic hurdles may limit its impact

  Moravec's paradox and Steven Pinker's 1994 observation remain relevant: "The main lesson of thirty-five years of AI research is that the hard problems are easy and the easy problems are hard." The hardest "easy" problems, like tying one's shoelaces, remain. Do breakthroughs in robotics easily follow those in generative modeling? That OpenAI disbanded its robotics team is not a strong signal.
  ...
  It seems highly unlikely to us that growth could greatly accelerate without progress in manipulating the physical world. Many current economic bottlenecks, from housing and healthcare to manufacturing and transportation all have a sizable physical-world component.

• Why AI is Harder Than We Think

  Since its beginning in the 1950s, the field of artificial intelligence has cycled several times between periods of optimistic predictions and massive investment ("AI spring") and periods of disappointment, loss of confidence, and reduced funding ("AI winter"). Even with today's seemingly fast pace of AI breakthroughs, the development of long-promised technologies such as self-driving cars, housekeeping robots, and conversational companions has turned out to be much harder than many people expected. One reason for these repeating cycles is our limited understanding of the nature and complexity of intelligence itself. In this paper I describe four fallacies in common assumptions made by AI researchers, which can lead to overconfident predictions about the field. I conclude by discussing the open questions spurred by these fallacies, including the age-old challenge of imbuing machines with humanlike common sense.

• AI and the automation of work

  ChatGPT and generative AI will change how we work, but how different is this to all the other waves of automation of the last 200 years? What does it mean for employment? Disruption? Coal consumption?

  As an analyst, though, I tend to prefer Hume's empiricism over Descartes - I can only analyse what we can know. We don't have AGI, and without that, we have only another wave of automation, and we don't seem to have any a priori reason why this must be more or less painful than all the others.

• Large language models, explained with a minimum of math and jargon

  Want to really understand how large language models work? Here's a gentle primer.