Oh, to have worked or studied at SAIL, back in the day. Of course, “in the day” was in the 1970s and early 1980s, when the Stanford Artificial Intelligence Laboratory was in its prime – long before AI became shorthand for “any technology that’s 20 years in the future – and always will be.”
When SAIL was at the cusp of computer science, as detailed in a recent essay in the New York Times, many researchers truly believed that artificial intelligence was solvable. Creating computers that would think – not merely follow algorithms – was a hard problem, to be sure. But not an impossible dream. The idea that you’d have a computer that would learn how to play chess, and be able to converse with humans well enough to pass the Turing Test, seemed to be a matter of just a few more transistors, a little more memory, choosing the right programming language and writing some better algorithms.
Hard, but solvable.
“Optimism as Artificial Intelligence Pioneers Reunite,” written by John Markoff, talks about those glory days when John McCarthy and his team thought that a thinking machine would only take a decade to build. However, as the essay describes, by the mid-1980s, AI seemed farther from creating artificial intelligence than ever before.
My own involvement in artificial intelligence comes from about that time. I studied AI a little, but from 1990-1992 served as editor of AI Expert Magazine. (The publication is now gone, alas, but Google found my old writer’s guidelines squirreled away at Carnegie-Mellon Univ. Talk about a flashback!)
Editing AI Expert was a dream job – getting to hang out with the best and brightest in the AI community, and working with many brilliant computer science researchers as authors, attending conferences, and learning from a blue-ribbon advisory board. Ahh, nirvana.
What’s interesting is that many so-called “technologies” that we covered in AI Expert never became part of machine intelligence or artificial intelligence – but instead became part of the mainstream. Expert systems, for example, are a core part of many search engines and data mining systems. Object-oriented programming evolved out of AI. Virtual reality was part of AI. Natural language processing. Object databases. Vision processing and image recognition. Those are all just software development today.
Of course, not everything we covered in AI Expert hit the jackpot. While neural networks, genetic algorithms and fuzzy logic still live, they’re also quite esoteric.
Older computer scientists (heavens, do I now fall into that category?) remember AI’s history and accomplishments. We know that although we don’t have true artificial intelligence systems, those investments and research in SAIL and other similar projects paid off handsomely, and appear in products and technologies we use every day. I hope that younger generations of developers appreciate where we started – and how far we’ve come.