Previous: SciShow Quiz Show: Reid Reimers
Next: Kids and Sugar: The Sweet-and-Lowdown



View count:280,865
Last sync:2023-01-21 20:45
Hank takes you to the next frontier of innovation: the XPrize for Artificial Intelligence, talking about how true AI can be measured, and what the future might look like.

Like SciShow? Want to help support us, and also get things to put on your walls, cover your torso and hold your liquids? Check out our awesome products over at DFTBA Records:

Or help support us by subscribing to our page on Subbable:
Looking for SciShow elsewhere on the internet?

Thanks Tank Tumblr:


No one is happier than I am, that we live in a time when our phones can give us directions and our cars can tell us where we can find the nearest glazed doughnut covered with bacon sprinkles.

But I think we all agree artificial intelligence wise, we could be doing a lot better. I mean C-3PO was fluent in over 6 million forms of communication, and we dreamed him up, like, forty years ago!

Thankfully we have the spirit of private enterprise and friendly competition, to make us challenge ourselves. And it seems our next challenge is creating machines that are every bit as articulate, intelligent, and compelling as we are.

I'm is Hank Green. This is SciShow news, and if I were Siri, I would be a little bit nervous.


Last week the XPrize Foundation, that ambitious non-profit that brought us private space flight and tricorders, announced a new competition: a cash reward for whoever creates artificial intelligence so advanced that it could deliver its own TED talk, without human assistance, and earn a standing ovation from the audience.

TED talks, as you probably are aware, are around 18 minute long speeches on big ideas. Technology, design, innovation etc.

So it seems like a natural platform for a computer smart enough to lecture you on your topic of choice. But is it possible?

Well when TED's official creator Chris Anderson and XPrize visionary, Peter Diamandis announced this new challenge, they said that this territory is so new that they need the public’s help in laying down the rules.

Like 'how long should the AI speech be?' 'How long will it have to prepare it?' 'How will the topic be selected?' 'Should the entrants be actual robots, or just programs?'

These kinds of things are important in determining what sort of intelligence we're looking for. Because any computer could present a pre-written speech. A more intelligent computer could use data to assemble an argument on the fly. But would even that be enough?

That's where the audience response comes in. The XPrize hasn't actually decided whether an actual standing ovation should be the goal, but the point is that success should be determined, by what is known in AI circles, as the Turing Test.

In 1950, computer scientist Alan Turing proposed that true artificial intelligence could be measured by whether a person using a program could even tell if it's a computer, or if the user could mistake it for another human.

Since Turing first proposed his test, lots of machines have actually passed it. Beginning with Eliza, an interactive program that conversed with users, by charading as various characters, the most famous being a psychotherapist, in 1966.

But a computer that has been programmed to act like a person, doesn’t prove intelligence.

Lots of clever programmers have designed their machines to misspell words or use slang to impersonate humans. It's like, "Not a Cylon here."

Instead our most intelligent computers, are specialists. Designed to do one task, and do it well. But not as a human would.

IBM's Deep Blue, for example beat world chess champion, Gary Kasparov in 1997, but it couldn't, like, identify the person it was playing against.

Google and Facebook use statistical learning programs, to tailor ads and finish your search queries. But they can't judge how much of a hipster you are.

More than likely, the TED-talking AI will be a specialist as well. Designed to search for and assemble data points in response to a topic.

It probably won't be able to make analogies, or tell anecdotes in its speech. Although I, for one, would love to hear it reminisce about the old eight bit days.

Still there are plenty of attempts, as you might imagine, to make computers think like humans.

Google's new algorithm, Deep Mind, is based on real human neurological systems.

And Elon Musk and Mark Zuckerberg have invested 40 million dollars in Vicarious, a company that wants to translate the brain's neocortex, the parts that sees, controls the body, understands language and does math, into computer code.

XPrize and TED say they don't expect their new AI challenge to be won in the next 20 years or so. But every year, TED may host two entrants to compete for it.

They say it's 'a way to track the growth of artificial intelligence and to generally contribute to powerful dissemination of great ideas.'

You can disseminate your great ideas about the AI XPrize at, and your great ideas about SciShow News on Facebook and Twitter, and in the comments below.

Thanks as always for watching, especially to our super intelligent Subbable subscribers.

And if you want to keep getting smarter with us, you can go to, and subscribe.

(End Slate)