AI Playing Video Games, Learning Warehouse Jobs and Building other AI
Artificial Intelligence (AI) is advancing faster and farther than most people realize. The things that The New York Times found AI doing in recent months are absolutely frightening.
The bothersome activities that AI are now capable of include:
AI are now Playing Video Games
Teaching themselves how to play video games. An AI at Elon Musk’s company OpenAI has figured out to how to play an old boat-racing video game called Coast Runners, The Times reported. That is pretty scary because some of the activities in video games are not very healthy.
Would Doom teach an AI how to gun people down, and can Call of it Duty or Destiny teach it how to fight a war? Some organizations are using virtual worlds based on Grand Theft Auto to educate AI, The Times reported. Is AI learning how to carjack vehicles and kill people?
Pretty scary, perhaps we should be worried about what else is out there playing on World of Tanks. Maybe it’s the AI that will be driving the next generation of tanks.
AI Learning how to do Warehouse Jobs on their Own
Those who doubt that technological employment is a real threat, need to read a New York Times article by Cade Metz.
Researchers at the University of California at Berkley have developed a machine that can learn how to recognize objects and load books, Metz reported. Metz mistakenly calls the machine a “robot” but it is AI because it can learn. The difference between a robot and AI is that can learn; robots can only perform a preprogrammed task.
The UC Berkley AI can recognize items it has never seen before and sorts through hundreds of objects. The goal at the Berkley lab is to develop an AI-controlled robot that can perform all the jobs in an Amazon (NASDAQ: AMZN) fulfillment center, so say goodbye to all those thousands of high-paying e-commerce positions.
“This can extend to tasks of assembly and more complex operations,” Juan Aparicio, the head of advanced manufacturing at Siemens (OTC: SMAWF) said. Siemens is financing the Berkley research in an effort to create next-generation factory and warehouse AI.
The Berkely AI features a neural network, a special 3D camera that enables it to recognize objects or patterns, and two ABB (NYSE: ABB) robot arms for grabbing stuff. A neural network is an advanced algorithm loosely based on the network of neurons in the human brain that enables AI to perform complex tasks and learn.
Amazon is also funding the research at Berkley, probably in an attempt to reduce labor costs. Amazon owns Kiva Robotics; the company has deployed thousands of robots in its fulfillment centers.
AI that can build Artificial Intelligence
The most frightening AI of all is being developed by a Google project called AutoML (machine learning). The goal of AutoML is to create an AI that can design and build other AIs, The Times reported.
Alphabet (NASDAQ: GOOG) is trying to build AI that can reproduce. The thinking behind AutoML is to reduce the need for the highly-paid engineers and computer scientists who to design and build AIs.
Such experts are rare; there are only around 10,000 of them in the world, which makes the creation of new AI difficult. That limits future development, but Alphabet (NASDAQ: GOOGL) hopes to change that by developing AI which writes algorithms.
That means computer scientists and engineers might be facing technological unemployment in a few years. Such a development will sound like poetic justice to the working stiffs whose manual labor jobs are threatened by AI. Nobody it seems is safe from technological unemployment, not even Google engineers.
Everybody will soon be building AI even the Bad Guys
Another result of that will be the make creation of new AI widespread. That would allow almost anybody, anywhere to create and deploy AI. It would probably include criminals and terrorists, so even the bad guys’ jobs are threatened by technological unemployment.
This is not a dystopian fantasy last year researchers at White Ops uncovered a Russian criminal cartel called Ad Fraud Komanda or AFK13 that had deployed 570,000 algorithm powered meth bots, Forbes reported. The meth bots used algorithms and possibly machine learning to mimic the behavior of human web surfers.
That enabled AFK13 to defraud advertisers of up to $5 million a day by inflating websites’ click rates. The Ad Fraud Komanda made that money by selling advertising based on fraudulent click rates.
It looks as if the age of AI has begun and it is truly frightening. One wonders what AI will do next and if this genie can be put back in the bottle.
