Ben Medlock (pictured), co-founder of SwiftKey, which gave Stephen Hawking (1942–2018) better speech software in 2014, offers a perspective on the limitations of artificial intelligence that you may not have heard before. SwiftKey (since sold to Microsoft) is a mobile app that uses predictive technology to adapt to the way that users type. While Hawking went on to warn that AI would take over the world, creating “the worst event in our history,” Medlock said, not so fast: Let’s look at what AI can’t do and why.
First, he says, back in the 1950s, researchers tried to develop artificial intelligence that could think like people, using logical symbols alone. The trouble was, people don’t think that way and in any event the real world doesn’t work that way. Many computer jokes later, programmers started using statistics to deal with huge masses of data instead. Statistics gathered on many thousands of images can form a pattern of an elephant or an umbrella. A powerful computer can also make connections, for example, between umbrellas and rain. This is machine learning and it has produced computers that can wallop through seas of data or calculations in order to beat humans at Jeopardy and Go. But there are limitations, says Medlock, some of which spring from a wrong understanding of evolution:
The traditional view of evolution suggests that our cellular complexity evolved from early eukaryotes via random genetic mutation and selection. But in 2005 the biologist James Shapiro at the University of Chicago outlined a radical new narrative. He argued that eukaryotic cells work ‘intelligently’ to adapt a host organism to its environment by manipulating their own DNA in response to environmental stimuli. Recent microbiological findings lend weight to this idea. For example, mammals’ immune systems have the tendency to duplicate sequences of DNA in order to generate effective antibodies to attack disease, and we now know that at least 43 per cent of the human genome is made up of DNA that can be moved from one location to another, through a process of natural ‘genetic engineering’.
Now, it’s a bit of a leap to go from smart, self-organising cells to the brainy sort of intelligence that concerns us here. But the point is that long before we were conscious, thinking beings, our cells were reading data from the environment and working together to mould us into robust, self-sustaining agents.Ben Medlock, “The body is the missing link for truly intelligent machines” at Aeon (March 14, 2017)
The main thing to see here is that humans are not mere smart robots, even at the biological level. Our systems are constantly redesigning themselves in ways that we don’t even know about unless we study cell biology. And the cell biologists are still learning. Dramatic new discoveries in their field are frequent.
One outcome is that most humans know a great deal about the world that no one needs to explain to us. Artificial intelligence does not know them:
But for an AI algorithm, the process begins from scratch each time. There is an active and important line of research, known as ‘inductive transfer’, focused on using prior machine-learned knowledge to inform new solutions. However, as things stand, it’s questionable whether this approach will be able to capture anything like the richness of our own bodily models.Ben Medlock, “The body is the missing link for truly intelligent machines” at Aeon(March 14, 2017)
One problem might be that machines, not being alive, don’t want anything. Even bacteria want something. Thus, a slime mold of bacteria seeking bread crumbs can solve the Traveling Salesman problem which befuddled computers for years. Bacteria are not smarter than the computers but they all have an inner incentive not to starve. The fact that bacteria are purpose-driven may even help us design better antibiotics.
Medlock thinks that the critical problem with computers is that they don’t have bodies so they don’t feel what the life around them is like. Some commentators think it matters that they don’t have minds either. In any case, the belief that sheer computing power is all it takes to produce intelligence is running up against the realities of the natural world—which is full of intelligence that does not depend mainly on computing power.
You may also enjoy:
Evolution and artificial intelligence face the same basic problem Think of the word ladder game, where we transform one word into another by changing only one letter at a time. Without knowledge about the goal and how to get there, it rapidly becomes first difficult and then completely impossible to reach the goal.
Can AI really evolve into superintelligence all by itself? We can’t just turn a big computer over to evolution and go away and hope for great things. Perpetual Innovation Machines tend to wind down because there is no universally good search. Computers are powerful because they have limitations.