English. I’m Neil.

And I’m Sam.

Do you like cooking, Sam? There’s a new

recipe I’ve been trying out - it’s for

‘frosted oysters’.

Frosted oysters?! Sounds… unusual. How

do you make it?

Well, take a pound of chicken, then some cubed

pork and half a crushed garlic.

Eh? I thought you said it was for ‘frosted

oysters’, whatever they are.

Yes, that’s right. Now heat it up until

boiling and serve with custard.

Ugh, that sounds disgusting! Who on earth

told you that recipe?

It’s not ‘who’ told me, Sam, but ‘what’.

In fact, that recipe was made by computers

using artificial intelligence, or AI, which

is the topic of today’s programme. In real

life, AI is making huge progress - from car

satnavs to detecting cancer cells. But as

you can see from that revolting recipe, things

don’t always go according to plan.

So, just how intelligent is artificial intelligence?

I mean, it definitely needs some cooking lessons!

Right. AI is not as intelligent as we tend

to think. AI programmes use artificial brain

cells to roughly imitate real brain cell activity,

but they’re still a long way behind human

levels of intelligence. And that’s my quiz

question – in terms of brain cell count,

what level of intelligence is AI currently

working at? Is AI as smart as:

a) a frog, b) an earthworm or

c) a bumblebee

Well, I don’t think any of those are good

cooks either, to be honest. I’ll say c)

a bumblebee, because at least they can

make honey!

Nice guess, Sam. We’ll find out the answer

later. But first let’s find out more about

how AI misunderstandings like the oyster recipe

can happen. Janelle Shane is the author of

‘You Look Like a Thing and I Love You’

in which she tells her amusing

experiences and bizarre experiments with AI.

You Look Like a Thing and I Love You – that’s

a strange title for a book, Neil.

Yes. It’s another example of AI

miscommunication.

The book title is what a AI produced when

asked to write chat-up lines – remarks men

and women make to start up a conversation

with someone they don’t know but find attractive.

Here she is talking to the BBC World Service

programme More or Less:

So ‘Machine learning’ is what most programmers

mean when they say ‘AI’. In the programme

that we’re used to, if you want to have

a computer programme solve a problem you have

to have a human programmer write down exhaustive

step-by-step instructions on how to do everything.

But with ‘machine learning’ you just give

it the goal, and then the programme figures

out via trial and error how it’s going to

solve that problem.

So even though we’re talking about machines

learning for themselves, there still need

to be humans involved at the start of the

journey. This human teaching is done by computer

programmers – people who write, or code,

the computer programmes used by AI.

Right. These programmers write algorithms

– a set of rules or procedures to be followed

in problem-solving exercises. So, for example,

the AI that wrote that oyster recipe read

thousands of other recipes before coming up

with its own version.

In other words, artificial intelligence uses

a process of trial and error – repeating

the same task over and over until finding

the most successful way. Only in the case

of the oyster recipe, there was more ‘error’

than ‘trial’!

Well, according to Janelle Shane, we can learn

a lot about something by seeing how it

goes wrong. Here she is, talking about an

AI which had been told to solve maths problems:

It seemed to be that it was getting scored

on how many wrong answers it got, and it was

supposed to be minimising the number of wrong

answers, and just by a stroke of luck as part

of its trial and error flailing around, one

of the flails it did accidentally deleted

the solutions list, and then it and everybody

else got a perfect score.

So, AIs learn by minimising their errors – reducing

them as much as possible. And sometimes,

these algorithms only discover the right answer

by a stroke of luck – when something unexpected

happens by good luck or chance. It seems to

me that they’re not so intelligent

after all!

Well, let’s settle it once and for all by

answering today’s quiz question.

Remember I asked you how intelligent AI was

in terms of brain cell count and you said,

as intelligent as...

I said c) a bumblebee.

Well, here’s Janelle again with the answer…

If you’re looking at rough computing power,

the algorithms we’re working with are probably

somewhere around the level of an earthworm.

So, the correct answer was b) as clever as

an earthworm! No wonder AIs can’t cook!

Or take a maths test without cheating! In

this programme we’ve been looking at artificial

intelligence, or AI, and seeing how programmers

– that’s people who write instructions

for computers to follow create algorithms

– sets of rules used in problem-solving.

AI learns through trial and error – repeating

the same activity again and again until discovering

the best way, and minimising – reducing

as much as possible, the number of errors

it makes.

And success can be the result of a stroke

of luck, when something unexpected happens

purely by chance, although so far that hasn’t

helped AIs to write good chat-up lines – the

flattering remarks people make to get to know

someone they find attractive.

And AIs don’t know much about cooking oysters

either!

That’s all from us from this programme.

Be sure to join us again for more topical

discussion and vocabulary at 6 Minute English

for BBC Learning English. Bye for now!

Bye.