That artificial intelligence (AI) poses a potential threat to people is not news. Tech gurus and scientists, ethicists and futurists, sounded the alarm years ago. It turns out they were both right and wrong: right in their assessment about the dangers but wrong in their timing. The rapid advances being made in AI make it a concern much sooner than they anticipated.
On May 18, DeepMind, a London-based firm owned by Google, said it was on the verge of creating artificial intelligence that is as intelligent as humans.
Dr. Nando de Freitas, a research scientist at the company and machine learning professor at Oxford, said, “the game is over.” De Freitas was talking about solving the most difficult challenges in developing a machine that is able to “learn” any task that requires intelligence - and do so without being trained. The hardest part of this task has been completed, he said.
Until now, only humans were able to do this. With DeepMind having reached this milestone it’s very likely that it will accelerate additional advances in this technology. In theory, this development may even prove to be a giant leap toward propelling society into a new era. According to de Freitas, scientists are already developing programs with this learning capability and have made amazing progress.
Like A Human Brain
In mid-May, DeepMind unveiled a new AI program called Gato, which the company claims can do more than 600 different tasks. One of Gato’s new features is that it utilizes a single neural network – a computing system that works like nerve cells in the human brain. Gato can “chat” with humans, write captions for photos, stack blocks using a robotic arm, and even play on a 1980s Atari game machine. Of course, going forward its capabilities certainly will improve dramatically.
DeepMind also has developed a program to play Go, a game more complicated than chess. The program beat Lee Sodul, the South Korean world champion, in a five-game match - and that was back in 2016.
The Daily Mail Online reports that an opinion piece appearing in “The Next Web” claimed humans will never be able to develop a machine with these learning skills. De Freitas rejected this idea, responding that such an outcome was not only inevitable but on the way.
“It’s all about scale now,” he wrote on Twitter, “making models bigger, safer, faster, able to innovate.... Solving these will create an AI capable of rivaling human intelligence.”
Will this technology really be as helpful as it sounds? Not necessarily.
Long before these advances, AI was already generating concerns. During his lifetime, British astrophysicist Stephen Hawking said that AI could spell the end of the human race. Elon Musk said that with AI we are summoning the devil.
And more recently, Dr. Stuart Armstrong at Oxford University’s Future of Humanity Institute said that AI will eventually make humans “redundant” and wipe us out. He believes machines will ultimately work at speeds now inconceivable and will skip communicating with humans. They will control the economy, financial markets, transportation, healthcare, and more, he warned.
Of course, they will also be used by the military and, considering their capabilities, clearly have the potential to cause havoc - either because they weren’t programmed carefully enough or because they were intentionally programmed to create problems. Armstrong gave this example: A simple instruction to “prevent human suffering” could easily be misinterpreted by a super-computer as meaning “kill all humans.”
In 2016, DeepMind researchers acknowledged the need for a “big red button” that could prevent a machine from completing “a harmful sequence of actions.” However, as has been said about guns, that button would be only as good or as bad as the person using it.
According to CNBC, “In 2020, DeepMind said it had solved a 50-year-old problem in biology, known as the ‘protein folding problem’ – understanding how a protein’s amino acid sequence dictates its 3D structure... DeepMind claimed to have solved this with 92 percent accuracy by analyzing ‘a neural network with 170,000 known protein sequences and their different structures.’”
That’s not all. A very recent study reported on Vice showed that humans can learn things from artificial intelligence systems and pass these lessons to other humans in ways that could influence our culture. In the game of Go, for example, DeepMind›s AI system made moves that no humans ever would. Since then, however, its strategy has been analyzed by Go players and those moves frequently copied.
One researcher said he believes AI-developed industrial and scientific applications will produce “noticeable leaps” this year. Gary Marcus, another researcher/entrepreneur, said this year or next we will see “the first medicine in which AI played a substantial role in the discovery process.” And Catherine Breslin, a machine learning scientist and consultant, noted another advance: She told CNBC that AI being developed now will combine vision, speech, and language capability rather than treat them as separate tasks.
IBM, Google, Meta (Facebook), Amazon, Apple, and Microsoft are among the high-tech giants that have invested in combination tens of billions of dollars in this technology by developing their own products, hiring the best researchers to develop new ones, and/or by purchasing start-ups.
To be sure, despite all of the advances, AI still can’t match or threaten humans - but that day could be on the way. Meanwhile, it’s a good bet that AI is much further developed than the public has been told, the real facts hidden so as not to alarm people, invite government oversight and regulation, or to tip off the competition about what companies are working on.
DeepMind says that it’s “on a mission to push the boundaries of AI, and develop programs that can learn to solve any complex problem without needing to be taught how.” The company believes this technology will become one of the most important and widely-beneficial scientific advances ever made.
Along the way, AI will create many new jobs, save industry a great deal of money by adding new efficiencies, make life better and some lucky investors very wealthy individuals.
Meanwhile, the risks it poses will continue to increase. Russian President Vladimir Putin is today one of the most hated people today, but few can disagree with an observation he made in 2017. Putin predicted that whichever country leads the way in AI research will dominate the world.
“Artificial intelligence is the future, not only for Russia, but for all humankind,” he said. “It comes with colossal opportunities, but also threats that are difficult to predict. Whoever becomes the leader in this sphere will become the ruler of the world,” he said
One hopes that the good guys will be the leader. The alternative is unthinkable.
Sources: cnbc.com; dailymail.co.uk; forbes.com; newscientist.com; techrepublic.com; theverge.com; verdict.co.uk; vice.com