The Xprize has issued a challenge to the world, develop an artificial intelligence capable of delivering a TED Talk.1 Peter Diamandis’s foundation, known by the tagline “Ideas Worth Spreading”, wants input on the selection of the AI’s topic, the required amount of preparation, how long the AI will drone, and how someone might cheat to win.2 Here’s their sample concept.3
During the TED Conference, the TED Audience chooses one of these subjects (or the subject is randomly chosen) and then the competing A.I. is given 30 minutes to prepare a compelling 3 min TED Talk.
The Team could decide how their A.I. would present on stage – would it be a physical robot that walks out to present? Or a disembodied voice?
After the talk, the audience would vote with their applause and, if appropriate, with a standing ovation.
Next, the A.I. would need to answer two questions from Chris Anderson, the host of the conference, and then a panel of experts would also add their votes.
Why let the computer ‘talk’ for only three minutes when humans present for 15-20 minutes? If we’re talking about artificial intelligence, with all the science fictional trappings behind the topic, an AI could wax poetic on something as mundane as quinoa, framing it as the second coming. I’d say it’s a tacit acknowledgement this is a PR stunt (Oh noes! I’m sooooo disillusioned!), and true, hard AI remain a Banksian pipe dream.
As for answering questions from Chris Anderson, that’s simple to accomplish with IBM’s Watson being a primitive proof-of-concept for interaction. Despite winning Jeopardy!, Watson was stumped by the Final Jeopardy answer, “Its largest airport is named for a World War II hero; its second largest for a World War II battle.” responding with the question, “What is Toronto?” Watson apologists at IBM say their computer was confounded by the question not being prefaced by “What U.S. city…” despite the category being U.S. Cities.4 Well, okay. But did Watson learn from its mistake without outside intervention from its programmers?
QED, Watson is neither intelligent nor self-aware.
What’s currently considered to be artificial intelligences are complex scripts iterating over a set of data, then chunking it out. Garbage in, garbage out. There is no independent thought, and any perceived spark of creativity is mere aleatory5. The challenge of taking an idea, presenting something completely revolutionary, using only a computer without any human intercession, leaves me skeptical.
Developing a machine that can cogitate a non-linear topic and apprehend new ideas (worth spreading), is exponentially more difficult than programming a computer to ape being clever by randomly assembling a presentation based on Wikipedia articles and their citations.6
Until then, I won’t believe a true AI presented at TED until P.Z. Myers, Richard Dawkins, and their Guerilla Skeptic minions force TED to censor the machine’s talk like Rupert Sheldrake and Graham Hancock.