The Incessant Overhyping of Machine Intelligence is Out of Control
Another day another article in the popular press warning us of our impending obsolescence. This time it is a piece in Mother Jones
The sub lead just beneath the not really clever title suggests we should all fear for our jobs and “sooner than you think!” Gasp..oh God no, holy shit, wtf. The first paragraph claims that by 2040 computers the size of a football are as smart as human beings. Smarter even. Plus they are computers! They never get tired or grouchy, or have to take shit or any of that stupid useless stuff that we pitiful humans have been cursed with.
Set aside for the moment the fact that this article references no publication, cites no expert, or even bothers to try and justify the bold statement about computers and intelligence in 2040. It simply states it as a matter of fact. Of course how could anyone disagree. It’s obvious. Only a dumb human would dare question such a totally non-disputable sentence like that. I am sure the author would respond by saying something like “but Watson, Deep Blue, that computer that won at Go, Google, Microsoft, Silicon Valley, the Staniford Curve, etc.” and on and on. Any person is free to speculate at their leisure but please don’t present pure speculation as matter of fact.
I take issue with the thesis at the heart of the article. Tell me what it “means” for a computer to be smarter than a human. How do you define intelligence? What about human intelligence? How about machine intelligence? If you can not give a precise and generally accepted definition of all of these things how do you propose to know when a machine is smarter than a human? For that matter how do you define knowledge? What does it mean to know a computer is smarter than a human? Does beating a human at a game count? If that is your criteria than surely machines are already smarter than us. Of course even this depends on what games you think should be included in this test of intellect. Chess, Go, Jeopardy? Check, check, and check to the machines.
What about basketball, football, golf, soccer? No computer has beat a human at these games. I suppose you would say intelligence is not required to win at games like this. Even dumb people can excel at athletic games. What about boxing-chess. This is an actual game played by real people. Could a computer win at this?
And so we come to the crux of the issue. What about my job? Will an “intelligent” computer, assuming this can ever be defined, someday take my job. To this I say relax. Take a deep breath. Ask yourself the following. Do I need to use my hands in complex, non-repetitive, and emotional state specific ways to do my job? Do I need to walk around and socialize with others to function at work effectively? Do I need to make decisions based on instinct or “gut feelings?” Do I need to hypothesize from new information, some of which is emotional state specific, to plan my next steps? Is the information I use not quantifiable? Do I ever take visual input, translate it to thought, cogitate a bit on those thoughts by comparing them to previous feelings, than take action by deriving the optimum course of action from those thoughts? Do I modify my actions by taking into consideration the impact they might have on others emotional states? Do I need to interact with the physical environment using parts of my body, limbs or fingers or toes, hands, etc. in a non-repetitive, situation specific way? Am I required to make judgements based on limited, imperfect data? Do I ever need to calm someone down who is upset or agitated? Do I need to navigate a constantly changing physical environment, with an ever changing set of tools, to do tasks of differing importance an intensity, depending on things that happened in the past?
I could go on and on but I hope my point is clear. If you are in a job for which you would answer yes to any of the above than you should not be worried about losing it to a computer. But wait, the techno utopian replies, an intelligent robot could probably do many or all of these things. After all the article specifically mentions robot overlords not computer overlords.
Unfortunately for the author their argument still fails and we should not be afraid of losing our jobs to intelligent robots anytime soon. Not even by 2040 I would suggest. All of the problems related to AI that I discuss in the first part of this post are nowhere close to resolution. Moreover the field of robotics is even further behind. It is just not conceivable at present to imagine how these two pursuits will be merged to create an “intelligent” robot that can even make a first approximation of a human being.
Even monotonous supposedly “low skill” jobs like janitor, maid, and administrative assistant have yet to be taken over by AI, let alone robots. Let us move up the job food chain, does your children’s school have any AI computers or robot teachers? Have you seen any robot police or fireman? What about android lawyers, doctors, or scientists? Robot politicians, EPA administrators, USDA or FDA regulators? Even in the realm of illegal activities have you heard about any AI or robot prostitutes or drug dealers? How about robot hit men or mob bosses?
Finally we come to the military. Certainly it is here where our robot overlords have shown their inevitability. Drones, bomb defusing bots, etc. I only ask you to consider this. Who is piloting and guiding our drone killers? Who holds the remote controls making our bomb diffusing bots go where they need to go? Is it an AI? Is it a robot? It is still us, the pathetic stupid human beings.
I will conclude by stating the obvious. It is foolish and egotistical to suggest that mankind is irreplaceable in every domain. It is without a doubt that certain jobs may be lost as advances in AI and robotics continue to expand. However it is downright wrong and irresponsible to suggest that the majority of us should be afraid for our livelihoods. Please stop hyping this fake threat and report on these issues with the seriousness and skepticism they deserve.
