Artificial intelligence, as Summer Wraps Up

What has been written about AI since late July?

First, we had “4 careers where workers will have to change jobs by 2030 due to AI and shifts in how we shop, according to a McKinsey study” (Jacob Zinkula, Business Insider, July 28^th).  The areas are “office support, customer service and sales, food services, and production work (e.g. manufacturing).”  The emphasis here is on “lower-wage jobs,” with “clerks, retail salespersons, administrative assistants, and cashiers” each expected to lose more than 600,000 positions, net, in the next seven years.

Kevin Roose spotlighted one apparent area of early adoption in “Aided by A.I. Language Models, Google’s Robots Are Getting Smart,” in The New York Times on the same date.  He started by describing an automaton responding to “pick up the extinct animal” by doing so with a dinosaur model instead of a lion or a whale, a seeming merger between AI and robotics, and meaning that much more along that line would also be possible.  Additionally, we have, per a Google scientist, such devices discovering “how to speak robot” by guessing “how a robot’s arm should move to pick up a ball or throw an empty soda can into the recycling bin.”  When machines do these things consistently correctly, which they do not yet, they will be especially valuable.  The next day the Times published Ben Ryder Howe’s “The Robots We Were Afraid of Are Already Here,” which was disappointing, as there seemed little new here among these automata’s industrial capabilities.

A subject we would all like to succeed at is “How to invest in AI” (Kim Clark, Kiplinger, July 29^th).  Since February, people expecting to be hugely important have long since pushed up some stocks and have done much more trading as developments and even plans have materialized, so we’re way up from any ground floor.  It’s reminiscent of buying automotive stocks in the 1920s, when knowing the industry had great potential did not mean we knew who the winners and losers would be.  Given that, though, there are industry leaders looking less risky than others – ones Clark named were chip designers Nvidia, Broadcom, and Taiwan Semiconductor, along with chip software maker Synopsis.  All are risky, but huge-potential investments always are. 

Controversy stepped into an area in progress for years, as “This tech is the ‘sad reality’ of restaurant industry’s future, business owner says robot works 12 hours a day” (Hannah Ray Lambert, Fox News, August 13^th).  When an Estacada, Oregon eatery, in response to not finding enough servers, deployed a Plato automaton to do their work, the owner got “customer pushback.”  Strange, but may prove to be common. 

Expecting the technology to be stronger, not weaker, than people thought months ago, was Arantza Pena Popo, in “AI is going to eliminate way more jobs than anyone realizes” (Insider, August 14^th). The author said that “permanent mass employment can safely be ruled out,” but hundreds of millions of people may not be able to work their current jobs in 35 years or less.  Beyond that, it’s just a great mass of unknowns.

A question on the minds of most in this field is “The U.S. Regulates Cars, Radio and TV.  When Will It Regulate A.I.?” (Ian Prasad Philbrick, The New York Times, August 24^th).  Per Philbrick’s determination, it “probably won’t happen soon,” as while television was regulated within five years of “invention or patenting,” radio took 20, telephones took 30, and railroads and automobiles were not delimited until 60 and 70 years later.  While “regulation often happens gradually as a technology improves or an industry grows,” “sometimes it happens only after tragedy.”  Accordingly, it may, or may not, take a while.

Related to my post suggesting similar things four days earlier was “The A.I. Revolution Is Coming, But Not as Fast as Some People Think,” by Steve Lohr in the New York Times on August 29^th.  Reservations and reasons for going slower named here are “risks of leaking confidential data, questions about how the data is used and about the accuracy of the A.I.-generated answers,” and it cited another McKinsey study suggesting “mainstream adoption” would take somewhere between 8 and 27 years.  By then, other problems, such as copyright infringement, may still be significant.  So don’t expect anything big for a while – but let’s still stay tuned.