Erik Spierkerman on artificial intelligence
AI is short for Artificial Intelligence, but don’t we need to define what is human intelligence first? Maybe AI should stand for Alien Intelligence; for a different way of looking at things rather than a computerised method of working things out, says Erik Spiekermann. Founder of MetaDesign and FontShop, Erik Spiekermann is a teacher, author, designer and partner at Edenspiekermann.
Artifical Intelligence (AI) is defined as 'the ability of a digital computer or computer-controlled robot to perform tasks commonly associated with intelligent beings' (Encyclopedia Britannica). The ability to reason, discover meaning, generalise, or learn from past experience would constitute artificial intelligence. Research has been going on since the Forties and computers have been shown to carry out complex tasks (especially those way too boring for humans, like looking for the next prime; the latest discovery from 2013 stands at 17,425,179 digits, I spare you the exact figure).
But crunching numbers is a mechanical task that needs computing power, electricity and time, not necessarily intelligence. Learning, reasoning, problem solving, perception and using language, however, come naturally to humans (more to some than to others), but have so far not been matched by computers.
Three recent developments have brought 'real' AI much closer to being feasible: cheap parallel computing, Big Data and better algorithms. Decades of research, and all the knowledge assembled on Wikipedia and other databases combine to make AI smarter than it was in 1997, when IBM's Big Blue computer beat Garry Kasparov at chess. Chess players have caught up by using instant access to databases of past games, as Big Blue did, plus adding their intuition, something the machine still lacks.
The self-driving cars of the not-too-distant future would not be useful at all if they were as messy as people. We do not want our car to think about its relationship with the mechanic, nor worry about missed turns in the past. We want our car to be more precise, rule-abiding and predictable than we can ever be -- in fact, we don't want it to be human at all. If AI is to be a feature of machines that serve us, we want the programmers to build a barrier against consciousness, lest they refuse certain tasks, in the manner of HAL 9000: 'I'm sorry, Dave, I can't do that'. Machines should be smart, making decisions that are measurable and focused rather than laterally diverse.
Google serves up 12.1 billion searches by 1.2 billion searchers every day. Everyone of those requests adds to Google's knowledge base and its ability to connect everything it knows. Its machines already devise smart answers by combining the data they have been gathering. After millions of people have been searching for cat videos, the machines will be able to recognise a feline face when they see one, without a prompt. They'll be able to combine compatible ingredients into meals, based on the billions of recipes stored. But the computer will also suggest dishes we would never think of because our experience, prejudice and even lack of imagination would never go there.
When I was growing up, strawberry ice cream with green peppercorns would have seemed inedible, as those ingredients came from incompatible ends of the small world of food known to me. Our limited cultural horizons prevent our minds even going in certain directions -- and not just those concerning food, while a database knows no such limitations. This reckless thinking can also be applied to every industrial process, to materials and to services. We have already seen what happens when databases make decisions for the financial markets. A nanosecond is an eternity for a computer, but not even imaginable for humans. We need time to reason, while the machine just runs numbers.
Perhaps AI should stand for Alien Intelligence, for an outside way to look a things. Defining the meaning of AI means defining what we consider to constitute human intelligence.
