Robots and Other Artificial Intelligence Applications – V

Here is the final installment in this series – at least for now. 

I start with three pieces from earlier this month giving the state of ChatGPT and chatbots in particular, which will continue to evolve but has reached a point where we can talk usefully about where it is going.  The first, “The Chatbots Are Here, and the Internet Industry Is in a Tizzy” (Tripp Mickle et al., The New York Times, March 8^th), said that “not since the iPhone has the belief that a new technology could change the industry run so deep,” with the authors forecasting massive shifts for cloud computing, e-commerce, social media, and publishing, affecting “$100 million in cloud spending, $500 billion in digital advertising and $5.4 trillion in e-commerce sales,” although “the volatility of chatbots has made it impossible to predict their impact.”  That spells out the situation now, and only the next several months and beyond will tell the story.

As for the current – or at least three weeks’ ago – technical situation, Cade Metz and Keith Collins told us in the March 14 New York Times that there are “10 Ways GPT-4 Is Impressive but Still Flawed.”  Although “it still makes things up,” its improvements are that “it has learned to be more precise,” “it has improved its accuracy,” “it can describe images with impressive detail,” “it has added serious expertise,” “it can give editors a run for their money,” “it is developing a sense of humor.  Sort of,” “it can reason – up to a point,” “it can ace standardized tests,” but “it is not good at discussing the future” and “it is still hallucinating.”  Expect the next release to be better, sometimes massively, at all of these.  And, if there was ever any doubt, we are getting “Microsoft to bring OpenAI’s chatbot technology to the office” (Dina Bass and Emily Chang, Benefit News, March 16^th) – in Office, where I have seen it proposing more continuation text, in LinkedIn, and elsewhere. 

Two more contributions told us things just behind the scenes of artificial intelligence and chatbot’s stunning recent progress.  “The great AI beef,” from Bloomberg Daily on March 8^th, told us that “in Silicon Valley, there’s a small but powerful group of people who believe (such advancement) could be very, very bad news – and that AI, if not handled correctly, could wipe out humanity within a couple decades.”  However, “there’s also a crowd who thinks our AI future will be amazing – bringing about untold future capabilities, abundance and utopia.”  “AI theorist” Eliezer Yudkowsky and OpenAI CEO Sam Altman, exchanging detailed comments from their stances on the former and latter sides respectively, have had “a somewhat inscrutable, inside-baseball catfight.”  David Wallace-Wells, in “Silicon Valley’s futurists have gone from utopian to dystopian” (The New York Times, March 27^th), recapped the situation between Altman and Yudkowsky and saw the latter scenario winning out among AI developers.  That will mean distortion, ultimately for better or worse, in how the technology progresses.

And how about the philosophers?  Because if you purport to perceive how much AI products are doing the equivalent of thinking, that’s what you are.  In “Can a Machine Know That We Know What It Knows?,” also in the March 27^th New York Times, Oliver Whang assumed that role and concluded, after looking at possibly pertinent academic studies, that one academic had concluded that “machines have theory of mind.”  Others responded with further work putting that deduction in doubt. Moving from empirical tests to resolving such issues, which call back to the millennia-old problem of consciousness, might be impossible.

I end with “Noam Chomsky:  The False Promise of ChatGPT,” with two co-authors in the New York Times on March 8^th.  The long-time linguistics professor named two large concerns about chatbot output.  First, despite being able to integrate masses of information, “such programs are stuck in a prehuman or nonhuman phase of cognitive evolution,” with an “absence of the most critical capacity of any intelligence:  to say not only what is the case, what was the case and what will be the case… but also what is not the case and what could and could not be the case.”  Second was chatbots not being “capable of moral thinking” by “constraining the otherwise limitless creativity… with a set of ethical principles that determines what ought and ought not to be,” the lack of these guidelines making it susceptible to incorporating clearly incorrect input data.  Ultimately, “they either overgenerate (producing both truths and falsehoods, endorsing ethical and unethical decisions alike) or undergenerate (exhibiting noncommitment to any decisions and indifference to consequences).” 

Chomsky’s assessment is superb – with one caveat.  If AI devices produce views that offend us, we should be able to objectively assess them.  We are still in charge, but we must be open-minded.  That will be a real 21^st-century intellectual challenge, and will draw as much controversy as ever.  But we will be better as both leaders and followers when we pursue it.  There is no suppressing artificial intelligence, but as has been true with so many past advancements, it will make an increasingly fine servant but will always – always – be a poor master.