Analytics And The Quest For Artificial Intelligence
Source: Meta S. Brown
APRIL 09: Bonham’s senior specialist Cassandra Hatton discusses the 1942 56-page notebook belonging to Alan Turing at Bonham’s auction house on April 9, 2015 in New York City. Turing was a legendary WWII codebreaker, one of the developers of the modern computer, and a seminal thinker in artificial intelligence. (Photo by Spencer Platt/Getty Images)
Decades of research in artificial intelligence have taught scientists this: people are really smart.
The other day, I watched a friend pull out his phone and ask, “What was Barry White’s wife’s name?” He wasn’t joking around, he really wanted to know. His personal assistant did pretty well, responding with an article that mentioned one of Barry White’s wives. It wasn’t direct, and it wasn’t the way a human would respond, but it was useful.
To provide responses like that, you need colossal data resources and predictive analytics tuned to the task. The technology builds on earlier developments in speech recognition, information retrieval and statistical analysis. The process is called “natural language understanding” and it is arguably the most important area of technical innovation today.
Even before the first commercially available general-purpose computer came off the production line, one of its creators, Alan Turing, asked the question, “Can machines think?” He proposed a little game, in which a machine and a person would each answer questions with the aim of making it hard to guess which of them was responding. (Though, from the start, Turing acknowledged that not everyone would regard success at the game, now known as the “Turing test”, as real thinking.)
Depending on your interpretation of the test, it may have been passed for the first time in 1966 (by the computer program ELIZA), in 2014, at some other time, or never. But it may not matter much, because the public, as well as the scientific community, has developed new expectations for intelligent machines. These expectations have been explored in decades of pop culture: nobody wants the power-mad computer HAL of the film “2001, A Space Odyssey”, but lots of people want the benign helpful computer of “Star Trek”.
Intelligent personal assistants such as Apple’s Siri, Google’s Google Now and Microsoft’s Cortana aim to fill the public desire for a helpful computer that speaks our language. But human language is complex, far more complex than a list of vocabulary and a set of grammatical rules. Real human communication relies on memory to provide context and analytical capability to assess meaning of questions and identify appropriate responses.
Whole networks of digital computers still can’t match the ability of one ordinary mind to understand and answer questions, but the potential rewards for success are so significant that tech giants are willing to make large investments required for developing intelligent personal assistants. To give a sense of the investments required, Apple bought Siri for an estimated $200 million in 2010, and in 2015, acquired UK speech technology startup Vocal IQ for an estimated $50-100 million.
Google, meanwhile, is more conspicuous about building on its analytics roots. Ray Kurzweil, a famed expert in artificial intelligence and Google engineering director, has been spreading the word that Google will be soon enable you to create a chatbot of yourself that can converse with other chatbots. This is a case of life imitating art, as the concept is modeled on a novel written by Kurzweil himself.
If an artificial self that holds conversations doesn’t excite you, wait for a later version. Maybe the upgrade will do your homework.
| }
|