04/02/2017 Click - Short Edition


04/02/2017

Similar Content

Browse content similar to 04/02/2017. Check below for episodes and series from the same categories and more!

Transcript


LineFromTo

but now on BBC News it's time for Click.

:00:00.:00:32.

For decades, scientists all around the world have been trying to create

:00:33.:00:35.

a machine that can match our intelligence.

:00:36.:00:37.

And nowadays artificially intelligent algorithms can perform

:00:38.:00:39.

For a long time scientists have been the use in games like chess,

:00:40.:00:45.

drafts and go as a benchmark for testing AI.

:00:46.:00:48.

And that's because all these games have a certain amount

:00:49.:00:51.

But this week the AI community has been celebrating a big win

:00:52.:00:59.

after a poker playing algorithm called Liberatus defeated four top

:01:00.:01:04.

human players in a 20 day match of heads up no limit Texas

:01:05.:01:09.

I've been using poker as a benchmark for 12 years.

:01:10.:01:14.

Now the best AI has surpassed the level of the best humans

:01:15.:01:17.

in the strategic reasoning under imperfect information.

:01:18.:01:22.

But even at this big win is only a little step towards creating

:01:23.:01:25.

Intelligence one capable of sophisticated thought

:01:26.:01:38.

across a wide spectrum of areas, and solving problems just

:01:39.:01:41.

It's an incredible time. And it's very hard to forecast.

:01:42.:01:45.

and has really incredible possibilities.

:01:46.:01:50.

But it's, I think it's impossible to forecast accurately.

:01:51.:01:57.

Speech has been another big challenge for AI researchers.

:01:58.:01:59.

Personal assistants and chat bots are becoming more sophisticated,

:02:00.:02:02.

but they so far can't fool us into thinking that they're human.

:02:03.:02:07.

But what if you thought you were talking to another person?

:02:08.:02:11.

Konichiwa. Do you speak English?

:02:12.:02:13.

Would that make you more likely to trust it?

:02:14.:02:19.

Well, two researchers at the London School of economics

:02:20.:02:25.

came up with an experiment to see if we would communicate better

:02:26.:02:28.

with AI if its messages were delivered to us by a human.

:02:29.:02:33.

They call this computer human hybrid the Echoborg.

:02:34.:02:38.

And to explore the concept, Jane Copestick found herself

:02:39.:02:40.

The Echoborg was inspired by research from Stanley Milgram.

:02:41.:02:50.

He is the Professor behind the controversial experiments

:02:51.:02:53.

on obedience in the 1960s, to see if people would deliver

:02:54.:02:56.

electric shocks to others if instructed to buy

:02:57.:02:59.

Milgram also studied body perception, to determine if we hold

:03:00.:03:04.

preformed opinions of other people based on their looks.

:03:05.:03:08.

By using hidden earpieces, people could speak someone else's

:03:09.:03:10.

The Echoborg has updated this research for the 21st century,

:03:11.:03:17.

to see if people will react better to artificial intelligence.

:03:18.:03:20.

Such as the messages from an online chat bot.

:03:21.:03:23.

If they are being delivered by a human.

:03:24.:03:26.

I'm in the first stages of testing this out by becoming

:03:27.:03:29.

I'm starting my speech shadowing practice.

:03:30.:03:33.

The first step in becoming a fully fledged Echoborg.

:03:34.:03:35.

The professors have told me this process will take at least eight

:03:36.:03:38.

I'm starting my first practice with JK Rowling's Harvard

:03:39.:03:43.

Members of the Harvard Corporation and the board of overseers...

:03:44.:03:49.

By shadowing speech, I should be able to quickly repeat

:03:50.:03:52.

back the messages from a chat bot so people won't realise

:03:53.:03:55.

It may seem something paradox, but there's horses in the cow...

:03:56.:04:02.

I did something and scuttled somewhere.

:04:03.:04:05.

Now, to put it to the test, I'm meeting creator

:04:06.:04:08.

Professor Alex Gillespie at the London School of Economics.

:04:09.:04:11.

And Kevin Corti, who called in on Skype.

:04:12.:04:21.

That goes around your neck. First step.

:04:22.:04:25.

Kevin is using a chat bot called Rose, which is not preprogrammed.

:04:26.:04:28.

The most noticeable problem in becoming a convincing AI

:04:29.:04:30.

are the delays while Rose thinks of a response to the question.

:04:31.:04:33.

Where do you come from? I was born in San Francisco.

:04:34.:04:39.

Really? In United States.

:04:40.:04:41.

When were you born? Couple of years ago.

:04:42.:04:46.

Are you a chat bot? I'm a human.

:04:47.:04:57.

I thought for a moment you might be a chat bot.

:04:58.:05:00.

Republic of Ireland and Croatia and France.

:05:01.:05:02.

My brother's from Ireland. Really?

:05:03.:05:05.

A magical place full of rain and crazy people.

:05:06.:05:08.

She says. Well...

:05:09.:05:12.

That was... Interesting.

:05:13.:05:14.

What you notices, they tend to be quite disconnected.

:05:15.:05:22.

It takes each sentence as a stand-alone sentence.

:05:23.:05:24.

Some of them will speak like they are artificial

:05:25.:05:27.

intelligence, and some of them will pretend not to be.

:05:28.:05:30.

But although last time I spoke to which it said

:05:31.:05:34.

Our final test for the Echoborg was to bring it on stage in front

:05:35.:05:40.

of an audience of 700 people at the BBC Future

:05:41.:05:43.

What a lot of humans find difficult...

:05:44.:05:47.

Canberra. OK, good.

:05:48.:05:54.

How do I know you are human, how do you know I'm human?

:05:55.:06:21.

In fact, some of the audience members were fooled.

:06:22.:06:33.

One thought it was a real conversation with a human,

:06:34.:06:36.

Some people thought you didn't want to talk

:06:37.:06:43.

That you were trying to avoid the question,

:06:44.:06:50.

they really thought you were trying to avoid the questions.

:06:51.:06:53.

Someone even said, had it been a man would it have

:06:54.:06:56.

Without becoming fully fledged Echoborgs, we are already giving

:06:57.:07:00.

a voice to artificial intelligence everyday.

:07:01.:07:05.

Through the algorithms guiding our news consumption,

:07:06.:07:06.

to our shopping habits and online searches.

:07:07.:07:09.

We're bringing AI to life more and more.

:07:10.:07:11.

Projects like the Echoborg let us reflect on what this means

:07:12.:07:14.

for our AI future and perhaps even what it means to be human.

:07:15.:07:18.

Hello and welcome to The Week in Tech.

:07:19.:07:35.

Playtime was never like this in my day.

:07:36.:07:38.

I've been taking a look at some of the latest toys hoping to light

:07:39.:07:42.

up the faces of children and grown-ups.

:07:43.:07:47.

And, inevitably, a few of them could be found at London's toy fair.

:07:48.:07:53.

This looks like a drone in a cage and that's because it is.

:07:54.:07:57.

It's also a proof of concept for a toy that's going to be

:07:58.:08:07.

Its inventor here is wearing this glove, which means you can control

:08:08.:08:12.

It all looks pretty simple, but I know you've been studying

:08:13.:08:18.

robotics for 15 years, so there's quite a bit more to this

:08:19.:08:21.

Once the science of gestures has been codified, and that's what we've

:08:22.:08:29.

been able to do, as you can imagine, we can bring all sorts of robotic

:08:30.:08:33.

The brain itself is in the glove, in the secretary and electronics.

:08:34.:08:40.

And the algorithms embedded in the glove.

:08:41.:08:44.

The drone is merely a conduit for the gestures being recorded

:08:45.:08:47.

There was also a clear trend towards giving kids a deeper level

:08:48.:08:53.

of control when it comes to toy gadgets.

:08:54.:08:58.

This is a robot that aims to help kids learn to code.

:08:59.:09:01.

They can operate it manually through the App, or setup sequences

:09:02.:09:04.

of the functions they'd like it to carry out.

:09:05.:09:07.

It's looks pretty raw when you've got all these leads and buttons,

:09:08.:09:10.

so it really is giving kids a chance to develop something.

:09:11.:09:13.

I also recently got my hand on a drone that kids can programme,

:09:14.:09:17.

spending time tweaking code at a computer or using drag

:09:18.:09:20.

I had a play around with some of the drone's functions.

:09:21.:09:28.

So maybe that shows who the real kid is.

:09:29.:09:31.

First of all I press W, which should get the drone

:09:32.:09:34.

This is a spot of that well-known activity, drone bowling.

:09:35.:09:39.

Is, the skittles are down here on the floor.

:09:40.:09:57.

It's not just about flying, though, you may want to do a bit of driving.

:09:58.:10:04.

To do that, you swap the wings for wheels.

:10:05.:10:17.

Last year we learned quite how much of an appetite

:10:18.:10:20.

there was for augmented reality in gaming.

:10:21.:10:23.

And give the big kids a chance for some play, too.

:10:24.:10:33.

This gaming robot, much like virtual avatars,

:10:34.:10:49.

It's also customisable and upgradable, with the ability

:10:50.:10:55.

to add wheels or even take on another robot in the room.

:10:56.:11:06.

Or if you want to get yourself moving, how

:11:07.:11:09.

This prototype has limited functionality, but still managed

:11:10.:11:22.

Not that it fought too hard when I decided I'd had enough.

:11:23.:11:42.

The full length of version is ready for you on iPlayer. Thank you for

:11:43.:11:56.

watching and see you soon..

:11:57.:11:59.

Download Subtitles

SRT

ASS