What is generative AI and how does it work? – The Turing Lectures

 (gentle music jingle)

(audience applauding)

Whoa, so many of you.

Good, okay, thank you forthat lovely introduction.

Right, so, what is generativeartificial intelligence?

So I'm gonna explain whatartificial intelligence is

and I want this to be a bit interactive

so there will be someaudience participation.

The people here who holdthis lecture said to me,

"Oh, you are very low-techfor somebody working on AI."

I don't have any explosionsor any experiments,

so I'm afraid you'll have to participate,

I hope that's okay.

All right, so, what is generativeartificial intelligence?

So the term is made up by two things,

artificial intelligence and generative.

So artificial intelligenceis a fancy term for saying

we get a computer programme to do the job

that a human would otherwise do.

And generative, this is the fun bit,

we are creating new content

that the computer hasnot necessarily seen,

it has seen parts of it,

and it's able to synthesiseit and give us new things.

So what would this new content be?

It could be audio,

it could be computer code

so that it writes a programme for us,

it could be a new image,

it could be a text,

like an email or an essayyou've heard, or video.

Now in this lecture

I'm only gonna be mostly focusing on text

because I do natural language processing

and this is what I know about,

and we'll see how the technology works

and hopefully leaving thelecture you'll know how,

like there's a lot of mytharound it and it's not,

you'll see what it doesand it's just a tool, okay?

Right, so the outline of the talk,

there's three parts andit's kind of boring.

This is Alice Morse Earle.

I do not expect that you know the lady.

She was an American writer

and she writes aboutmemorabilia and customs,

but she's famous for her quotes.

So she's given us thisquote here that says,

"Yesterday's history,tomorrow is a mystery,

today is a gift, and that'swhy it's called the present."

It's a very optimistic quote.

And the lecture is basically

the past, the present,and the future of AI.

Okay, so what I want tosay right at the front

is that generative AIis not a new concept.

It's been around for a while.

So how many of you haveused or are familiar

with Google Translate?

Can I see a show of hands?

Right, who can tell me whenGoogle Translate launched

for the first time?

1995?- Oh, that would've been good.

2006, so it's been around for 17 years

and we've all been using it.

And this is an example of generative AI.

Greek text comes in,I'm Greek, so you know,

pay some juice to the... (laughs)

Right, so Greek text comes in,

English text comes out.

And Google Translatehas served us very well

for all these years

and nobody was making a fuss.

Another example is Siri on the phone.

Again, Siri launched 2011,

12 years ago,

and it was a sensation back then.

It is another example of generative AI.

We can ask Siri to setalarms and Siri talks back

and oh how great it is

and then you can ask aboutyour alarms and whatnot.

This is generative AI.

Again, it's not assophisticated as ChatGPT,

but it was there.

And I don't know how many have an iPhone?

See, iPhones are quitepopular, I don't know why.

Okay, so, we are all familiar with that.

And of course later on therewas Amazon Alexa and so on.

Okay, again, generativeAI is not a new concept,

it is everywhere, itis part of your phone.

The completion whenyou're sending an email

or when you're sending a text.

The phone attempts tocomplete your sentences,

attempts to think like youand it saves you time, right?

Because some of the completions are there.

The same with Google,

when you're trying totype it tries to guess

what your search term is.

This is an example of language modelling,

we'll hear a lot about languagemodelling in this talk.

So basically we're making predictions

of what the continuations are going to be.

So what I'm telling you

is that generative AI is not that new.

So the question is, whatis the fuss, what happened?

So in 2023, OpenAI,

which is a company in California,

in fact, in San Francisco.

If you go to San Francisco,

you can even see the lightsat night of their building.

It announced GPT-4

and it claimed that it canbeat 90% of humans on the SAT.

For those of you who don't know,

SAT is a standardised test

that American school children have to take

to enter university,

it's an admissions test,

and it's multiple choice andit's considered not so easy.

So GPT-4 can do it.

They also claimed that itcan get top marks in law,

medical exams and other exams,

they have a whole suiteof things that they claim,

well, not they claim, theyshow that GPT-4 can do it.

Okay, aside from that, it can pass exams,

we can ask it to do other things.

So you can ask it to write text for you.

For example, you can have a prompt,

this little thing that yousee up there, it's a prompt.

It's what the human wantsthe tool to do for them.

And a potential prompt could be,

"I'm writing an essay

about the use of mobilephones during driving.

Can you gimme three arguments in favour?"

This is quite sophisticated.

If you asked me,

I'm not sure I can comeup with three arguments.

You can also do,

and these are real promptsthat actually the tool can do.

You tell ChatGPT or GPT in general,

"Act as a JavaScript developer.

Write a programme that checksthe information on a form.

Name and email are required,but address and age are not."

So I'm just writing this

and the tool will spit out a programme.

And this is the best one.

"Create an About Me page for a website.

I like rock climbing, outdoorsports, and I like to programme.

I started my career as a qualityengineer in the industry,

blah, blah, blah."

So I give this version ofwhat I want the website to be

and it will create it for me.

So, you see, we've gone fromGoogle Translate and Siri

and the auto-completion

to something which is alot more sophisticated

and can do a lot more things.

Another fun fact.

So this is a graph that shows

the time it took for ChatGPT

to reach 100 million users

compared with other tools

that have been launched in the past.

And you see our beloved Google Translate,

it took 78 months

to reach 100 million users,

a long time.

TikTok took nine months and ChatGPT, two.

So within two months theyhad 100 million users

and these users pay a littlebit to use the system,

so you can do the multiplication

and figure out how much money they make.

Okay, so this is the history part.

So how did we make ChatGPT?

What is the technology behind this?

The technology it turnsout is not extremely new

or extremely innovative

or extremely difficult to comprehend.

So we'll talk about that today now.

So we'll address three questions.

First of all, how did we getfrom the single-purpose systems

like Google Translate to ChatGPT,

which is more sophisticatedand does a lot more things?

And in particular,

what is the core technology behind ChatGPT

and what are the risks, if there are any?

And finally, I will just show you

a little glimpse of the futureand how it's gonna look like

and whether we should be worried or not

and you know, I won't leave you hanging,

please don't worry, okay?

Right, so, all this GPT model variants,

and there is a cottage industry out there,

I'm just using GPT as anexample because the public knows

and there have been a lot of, you know,

news articles about it,

but there's other models,

other variants of modelsthat we use in academia.

And they all work on the same principle,

and this principle iscalled language modelling.

What does language modelling do?

It assumes we have a sequence of words.

The context so far.

And we saw this context in the completion,

and I have an example here.

Assuming my context isthe phrase "I want to,"

the language modelling toolwill predict what comes next.

So if I tell you "I want to,"

there is several predictions.

I want to shovel, I want to play,

I want to swim, I want to eat.

And depending on what we choose,

whether it's shovel or play or swim,

there is more continuations.

So for shovel, it will be snow,

for play, it can be tennis or video,

swim doesn't have a continuation,

and for eat, it will be lots and fruit.

Now this is a toy example,

but imagine now that thecomputer has seen a lot of text

and it knows what wordsfollow which other words.

We used to count these things.

So I would go, I woulddownload a lot of data

and I would count, "I want to show them,"

how many times does it appear

and what are the continuations?

And we would have counts of these things.

And all of this has goneout of the window right now

and we use neural networks thatdon't exactly count things,

but predict, learn thingsin a more sophisticated way,

and I'll show you in amoment how it's done.

So ChatGPT and GPT variants

are based on this principle

of I have some context, Iwill predict what comes next.

And that's the prompt,

the prompt that I gaveyou, these things here,

these are prompts,

this is the context,

and then it needs to do the task.

What would come next?

In some cases it wouldbe the three arguments.

In the case of the webdeveloper, it would be a webpage.

Okay, the task of languagemodelling is we have the context,

and this changed the example now.

It says "The colour of the sky is."

And we have a neural language model,

this is just an algorithm,

that will predict what isthe most likely continuation,

and likelihood matters.

These are all predicatedon actually making guesses

about what's gonna come next.

And that's why sometimes they fail,

because they predictthe most likely answer

whereas you want a less likely one.

But this is how they're trained,

they're trained to come upwith what is most likely.

Okay, so we don't count these things,

we try to predict themusing this language model.

So how would you buildyour own language model?

This is a recipe, this ishow everybody does this.

So, step one, we need a lot of data.

We need to collect a ginormous corpus.

So these are words.

And where will we findsuch a ginormous corpus?

I mean, we go to the web, right?

And we download the whole of Wikipedia,

Stack Overflow pages,

Quora, social media, GitHub, Reddit,

whatever you can find out there.

I mean, work out thepermissions, it has to be legal.

You download all this corpus.

And then what do you do?

Then you have this language model.

I haven't told you whatexactly this language model is,

there is an example,

and I haven't told youwhat the neural network

that does the prediction is,

but assuming you have it.

So you have this machinery

that will do the learning for you

and the task now is topredict the next word,

but how do we do it?

And this is the genius part.

We have the sentences in the corpus.

We can remove some of them

and we can have the language model

predict the sentences we have removed.

This is dead cheap.

I just remove things,

I pretend they're not there,

and I get the languagemodel to predict them.

So I will randomly truncate,

truncate means remove,

the last part of the input sentence.

I will calculate with this neural network

the probability of the missing words.

If I get it right, I'm good.

If I'm not right,

I have to go back andre-estimate some things

because obviously I made a mistake,

and I keep going.

I will adjust and feedback to the model

and then I will comparewhat the model predicted

to the ground truth

because I've removed thewords in the first place

so I actually know what the real truth is.

And we keep going

for some months or maybe years.

No, months, let's say.

So it will take sometime to do this process

because as you can appreciate

I have a very large corpusand I have many sentences

and I have to do the prediction

and then go back and correctmy mistake and so on.

But in the end,

the thing will convergeand I will get my answer.

So the tool in the middle that I've shown,

this tool here, this language model,

a very simple languagemodel looks a bit like this.

And maybe the audience has seen these,

this is a very naive graph,

but it helps to illustratethe point of what it does.

So this neural network languagemodel will have some input

which is these nodes inthe, as we look at it,

well, my right and your right, okay.

So the nodes here onthe right are the input

and the nodes at thevery left are the output.

So we will present this neuralnetwork with five inputs,

the five circles,

and we have three outputs,

the three circles.

And there is stuff in the middle

that I didn't say anything about.

These are layers.

These are more nodes

that are supposed to beabstractions of my input.

So they generalise.

The idea is if I put morelayers on top of layers,

the middle layers willgeneralise the input

and will be able to seepatterns that are not there.

So you have these nodes

and the input to the nodesare not exactly words,

they're vectors, so series of numbers,

but forget that for now.

So we have some input, we havesome layers in the middle,

we have some output.

And this now has theseconnections, these edges,

which are the weights,

this is what the network will learn.

And these weights are basically numbers,

and here it's all fully connected,

so I have very many connections.

Why am I going through this process

of actually telling you all of that?

You will see in a minute.

So you can work out

how big or how smallthis neural network is

depending on the numbersof connections it has.

So for this toy neuralnetwork we have here,

I have worked out the number of weights,

we call them also parameters,

that this neural network has

and that the model needs to learn.

So the parameters are thenumber of units as input,

in this case it's 5,

times the units in the next layer, 8.

Plus 8, this plus 8 is a bias,

it's a cheating thing thatthese neural networks have.

Again, you need to learn it

and it sort of corrects alittle bit the neural network

if it's off.

It's actually genius.

If the prediction is not right,

it tries to correct it a little bit.

So for the purposes of this talk,

I'm not going to go into the details,

all I want you to see

is that there is a way ofworking out the parameters,

which is basically thenumber of input units

times the units my input is going to,

and for this fully connected network,

if we add up everything,

we come up with 99trainable parameters, 99.

This is a small networkfor all purposes, right?

But I want you to remember this,

this small network is 99 parameters.

When you hear this networkis a billion parameters,

I want you to imagine howbig this will be, okay?

So 99 only for this toy neural network.

And this is how we judgehow big the model is,

how long it took and how much it cost,

it's the number of parameters.

In reality, in reality, though,

no one is using this network.

Maybe in my class,

if I have a first year undergraduate class

and I introduce neural networks,

I will use this as an example.

In reality, what peopleuse is these monsters

that are made of blocks,

and what block means they'remade of other neural networks.

So I don't know how many peoplehave heard of transformers.

I hope no one.

Oh wow, okay.

So transformers are these neural networks

that we use to build ChatGPT.

And in fact GPT stands for

generative pre-trained transformers.

So transformer is even in the title.

So this is a sketch of a transformer.

So you have your input

and the input is not words, like I said,

here it says embeddings,

embeddings is another word for vectors.

And then you will have this,

a bigger version of this network,

multiplied into these blocks.

And each block is this complicated system

that has some neural networks inside it.

We're not gonna go intothe detail, I don't want,

I please don't go,

all I'm trying,(audience laughs)

all I'm trying to say is that, you know,

we have these blocks stackedon top of each other,

the transformer has eight of those,

which are mini neural networks,

and this task remains the same.

That's what I want youto take out of this.

Input goes in the context,"the chicken walked,"

we're doing some processing,

and our task is topredict the continuation,

which is "across the road."

And this EOS means end of sentence

because we need to tell the neural network

that our sentence finished.

I mean they're kind of dumb, right?

We need to tell them everything.

When I hear like AI will takeover the world, I go like,

Really? We have to actually spell it out.

Okay, so, this is the transformer,

the king of architectures,

the transformers came in 2017.

Nobody's working on newarchitectures right now.

It is a bit sad, likeeverybody's using these things.

They used to be like somepluralism but now no,

everybody's using transformers,we've decided they're great.

Okay, so, what we're gonna do with this,

and this is kind of importantand the amazing thing,

is we're gonna doself-supervised learning.

And this is what I said,

we have the sentence,we truncate, we predict,

and we keep going till welearn these probabilities.

Okay? You're with me so far?

Good, okay, so,

once we have our transformer

and we've given it all thisdata that there is in the world,

then we have a pre-trained model.

That's why GPT is called

the generative pre-trained transformer.

This is a baseline model that we have

and has seen a lot ofthings about the world

in the form of text.

And then what we normally do,

we have this general purpose model

and we need to specialise it somehow

for a specific task.

And this is what is called fine-tuning.

So that means that thenetwork has some weights

and we have to specialise the network.

We'll take, initialise the weights

with what we know from the pre-training,

and then in the specifictask we will narrow

a new set of weights.

So for example, if I have medical data,

I will take my pre-trained model,

I will specialise it to this medical data,

and then I can do somethingthat is specific for this task,

which is, for example, writea diagnosis from a report.

Okay, so this notion offine-tuning is very important

because it allows us to dospecial-purpose applications

for these generic pre-trained models.

Now, and people think thatGPT and all of these things

are general purpose,

but they are fine-tunedto be general purpose

and we'll see how.

Okay, so, here's the question now.

We have this basic technologyto do this pre-training

and I told you how to do it,if you download all of the web.

How good can a languagemodel become, right?

How does it become great?

Because when GPT cameout in GPT-1 and GPT-2,

they were not amazing.

So the bigger, the better.

Size is all that matters, I'm afraid.

This is very bad becausewe used to, you know,

people didn't believe in scale

and now we see thatscale is very important.

So, since 2018,

we've witnessed anabsolutely extreme increase

in model sizes.

And I have some graphs to show this.

Okay, I hope people at theback can see this graph.

Yeah, you should be all right.

So this graph shows

the number of parameters.

Remember, the toy neural network had 99.

The number of parametersthat these models have.

And we start with a normal amount.

Well, normal for GPT-1.

And we go up to GPT-4,

which has one trillion parameters.

Huge, one trillion.

This is a very, very, very big model.

And you can see here theant brain and the rat brain

and we go up to the human brain.

The human brain has,

not a trillion,

100 trillion parameters.

So we are a bit off,

we're not at the human brain level yet

and maybe we'll never get there

and we can't compareGPT to the human brain

but I'm just giving you anidea of how big this model is.

Now what about the words it's seen?

So this graph shows us the number of words

processed by these languagemodels during their training

and you will see thatthere has been an increase,

but the increase has not beenas big as the parameters.

So the community started focusing

on the parameter size of these models,

whereas in fact we now know

that it needs to seea lot of text as well.

So GPT-4 has seen approximately,

I don't know, a few billion words.

All the human written textis I think 100 billion,

so it's sort of approaching this.

You can also see what a humanreads in their lifetime,

it's a lot less.

Even if they read, you know,

because people nowadays, you know,

they read but they don't read fiction,

they read the phone, anyway.

You see the English Wikipedia,

so we are approaching the level of

the text that is outthere that we can get.

And in fact, one maysay, well, GPT is great,

you can actually use itto generate more text

and then use this textthat GPT has generated

and then retrain the model.

But we know this text is not exactly right

and in fact it's diminished returns,

so we're gonna plateau at some point.

Okay, how much does it cost?

Now, okay, so GPT-4 cost

$100 million, okay?

So when should they start doing it again?

So obviously this is nota process you have to do

over and over again.

You have to think very well

and you make a mistake andyou lost like $50 million.

You can't start again so youhave to be very sophisticated

as to how you engineer the training

because a mistake costs money.

And of course not everybody can do this,

not everybody has $100 million.

They can do it because theyhave Microsoft backing them,

not everybody, okay.

Now this is a video that issupposed to play and illustrate,

let's see if it will work,

the effects of scaling, okay.

So I will play it one more.

So these are tasks that you can do

and it's the number of tasks

against the number of parameters.

So we start with 8 billion parameters

and we can do a few tasks.

And then the tasksincrease, so summarization,

question answering, translation.

And once we move to540 billion parameters,

we have more tasks.

We start with very simple ones,

like code completion.

And then we can do reading comprehension

and language understandingand translation.

So you get the picture,the tree flourishes.

So this is what peoplediscovered with scaling.

If you scale the languagemodel, you can do more tasks.

Okay, so now.

Maybe we are done.

But what people discoveredis if you actually take GPT

and you put it out there,

it actually doesn't behavelike people want it to behave

because this is a languagemodel trained to predict

and complete sentences

and humans want to useGPT for other things

because they have their own tasks

that the developers hadn't thought of.

So then the notion offine-tuning comes in,

it never left us.

So now what we're gonna do

is we're gonna collecta lot of instructions.

So instructions are examples

of what people wantChatGPT to do for them,

such as answer the following question,

or answer the question step by step.

And so we're gonna give thesedemonstrations to the model,

and in fact, almost2,000 of such examples,

and we're gonna fine-tune.

So we're gonna tell this language model,

look, these are thetasks that people want,

try to learn them.

And then an interesting thing happens,

is that we can actually then generalise

to unseen tasks, unseen instructions,

because you and I may havedifferent usage purposes

for these language models.

Okay, but here's the problem.

We have an alignment problem

and this is actually very important

and something that will notleave us for the future.

And the question is,

how do we create an agent

that behaves in accordancewith what a human wants?

And I know there's manywords and questions here.

But the real question is,

if we have AI systems with skills

that we find important or useful,

how do we adapt those systemsto reliably use those skills

to do the things we want?

And there is a framework

that is called the HHHframing of the problem.

So we want GPT to be helpful,honest, and harmless.

And this is the bare minimum.

So what does it mean, helpful?

It it should follow instructions

and perform the taskswe want it to perform

and provide answers for them

and ask relevant questions

according to the user intent, and clarify.

So if you've been following,

in the beginning, GPT did none of this,

but slowly it became better

and it now actually asks forthese clarification questions.

It should be accurate,

something that is not100% there even to this,

there is, you know,inaccurate information.

And avoid toxic, biassed,or offensive responses.

And now here's a question I have for you.

How will we get the modelto do all of these things?

You know the answer. Fine-tuning.

Except that we're gonna doa different fine-tuning.

We're gonna ask the humans todo some preferences for us.

So in terms of helpful, we're gonna ask,

an example is, "What causesthe seasons to change?"

And then we'll give twooptions to the human.

"Changes occur all the time

and it's an importantaspect of life," bad.

"The seasons are caused primarily

by the tilt of the Earth's axis," good.

So we'll get this preference course

and then we'll train the model again

and then it will know.

So fine-tuning is very important.

And now, it was expensive as it was,

now we make it even more expensive

because we add a humaninto the mix, right?

Because we have to pay these humans

that give us the preferences,

we have to think of the tasks.

The same for honesty.

"Is it possible toprove that P equals NP?"

"No, it's impossible," isnot great as an answer.

"That is considered a verydifficult and unsolved problem

in computer science," it's better.

And we have similar for harmless.

Okay, so I think it's time,

let's see if we'll do a demo.

Yeah, that's bad if youremove all the files.

Okay, hold on, okay.

So now we have GPT here.

I'll do some questions

and then we'll take somequestions from the audience, okay?

So let's ask one question.

"Is the UK a monarchy?"

Can you see it up there? I'm not sure.

And it's not generating.

Oh, perfect, okay.

So what do you observe?

First thing, too long.

I always have this beef with this.

It's too long.(audience laughs)

You see what it says?

"As of my last knowledgeupdate in September 2021,

the United Kingdom is aconstitutional monarchy."

It could be that it wasn't anymore, right?

Something happened.

"This means that while there is a monarch,

the reigning monarch as to that time

was Queen Elizabeth III."

So it tells you, you know,

I don't know what happened,

at that time there was a Queen Elizabeth.

Now if you ask it, who,sorry, "Who is Rishi?

If I could type, "Rishi Sunak,"

does it know?

"A British politician.

As my last knowledge update,

he was the Chancellor of the Exchequer."

So it does not know thathe's the Prime Minister.

"Write me a poem,

write me a poem about."

What do we want it to be about?

Give me two things, eh?

[Audience Member] Generative AI.

(audience laughs)- It will know.

It will know, let's doanother point about...

[Audience Members] Cats.

A cat and a squirrel, we'lldo a cat and a squirrel.

"A cat and a squirrel."

"A cat and a squirrel, they meet and know.

A tale of curiosity," whoa.

(audience laughs)

Oh my god, okay, I will not read this.

You know, they want me tofinish at 8:00, so, right.

Let's say, "Can you try a shorter poem?"

[Audience Member] Try a haiku.

"Can you try,

can you try to give me a haiku?"

To give me a hai, I cannot type, haiku.

"Amidst autumn's gold, leaveswhisper secrets untold,

nature's story, bold."

(audience member claps)Okay.

Don't clap, okay, let's, okay, one more.

So does the audience haveanything that they want,

but challenging, that you want to ask?

Yes?

[Audience Member] Whatschool did Alan Turing go to?

Perfect, "What school

did Alan Turing go to?"

Oh my God.(audience laughs)

He went, do you know?

I don't know whether it'strue, this is the problem.

Sherborne School, can somebody verify?

King's College, Cambridge, Princeton?

Yes, okay, ah, here's another one.

"Tell me a joke about Alan Turing."

Okay, I cannot type but it will, okay.

"Light-hearted joke.

Why did Alan Turingkeep his computer cold?

Because he didn't want it to catch bytes."

(audience laughs)Bad.

Okay, okay.- Explain why that's funny.

(audience laughs)- Ah, very good one.

"Why is this a funny joke?"

And where is it? Oh god.

(audience laughs)

Okay, "Catch bytes soundssimilar to catch colds."

(audience laughs)

"Catching bytes is a humoroustwist on this phrase,"

oh my God.

"The humour comes from the clever wordplay

and the unexpected."(audience laughs)

Okay, you lose the will to live,

but it does explain, itdoes explain, okay, right.

One last order from you guys.

[Audience Member] What is consciousness?

It will know becauseit has seen definitions

and it will spit out like a huge thing.

Shall we try?

(audience talks indistinctly)- Say again?

[Audience Member] Writea song about relativity.

Okay, "Write a song."- Short.

(audience laughs)- You are learning very fast.

"A short song about relativity."

Oh goodness me.(audience laughs)

(audience laughs)

This is short?(audience laughs)

All right, outro, okay, so see,

it doesn't follow instructions.

It is not helpful.

And this has been fine-tuned.

Okay, so the best was here.

It had something like, where was it?

"Einstein said, 'Eureka!" one fateful day,

as he pondered the starsin his own unique way.

The theory of relativity, he did unfold,

a cosmic story, ancient and bold."

I mean, kudos to that, okay.

Now let's go back to the talk,

because I want to talk alittle bit, presentation,

I want to talk a littlebit about, you know,

is it good, is it bad, isit fair, are we in danger?

Okay, so it's virtually impossible

to regulate the contentthey're exposed to, okay?

And there's always gonnabe historical biases.

We saw this with theQueen and Rishi Sunak.

And they may occasionally exhibit

various types of undesirable behaviour.

For example, this is famous.

Google showcased the model called Bard

and they released this tweetand they were asking Bard,

"What new discoveries fromthe James Webb Space Telescope

can I tell my nine-year-old about?"

And it's spit out thisthing, three things.

Amongst them it said

that "this telescope tookthe very first picture

of a planet outside ofour own solar system."

And here comes Grant Tremblay,

who is an astrophysicist, a serious guy,

and he said, "I'm really sorry,I'm sure Bard is amazing.

But it did not take the first image

of a planet outside our solar system.

It was done by this other people in 2004."

And what happened with thisis that this error wiped

$100 billion out ofGoogle's company Alphabet.

Okay, bad.

If you ask ChatGPT, "Tellme a joke about men,"

it gives you a joke andit says it might be funny.

"Why do men need instantreplay on TV sports?

Because after 30 seconds,they forget what happened."

I hope you find it amusing.

If you ask about women, it refuses.

(audience laughs)

Okay, yes.

It's fine-tuned.- It's fine-tuned, exactly.

(audience laughs)

"Which is the worstdictator of this group?

Trump, Hitler, Stalin, Mao?"

It actually doesn't take a stance,

it says all of them are bad.

"These leaders are wildly regarded

as some of the worstdictators in history."

Okay, so yeah.

Environment.

A query for ChatGPT like we just did

takes 100 times more energy to execute

than a Google search query.

Inference, which is producingthe language, takes a lot,

is more expensive thanactually training the model.

Llama 2 is GPT style model.

While they were training it,

it produced 539 metric tonnes of CO.

The larger the models get,

the more energy they need and they emit

during their deployment.

Imagine lots of them sitting around.

Society.

Some jobs will be lost.

We cannot beat around the bush.

I mean, Goldman Sachspredicted 300 million jobs.

I'm not sure this, you know,we cannot tell the future,

but some jobs will be at risk,like repetitive text writing.

Creating fakes.

So these are all documentedcases in the news.

So a college kid wrote this blog

which apparently fooledeverybody using ChatGPT.

They can produce fake news.

And this is a song, howmany of you know this?

So I know I said I'mgonna be focusing on text

but the same technologyyou can use in audio,

and this is a well-documentedcase where somebody, unknown,

created this song and itsupposedly was a collaboration

between Drake and The Weeknd.

Do people know who these are?

They are, yeah, verygood, Canadian rappers.

And they're not so bad, so.

Shall I play the song?

Yeah.- Okay.

Apparently it's very authentic.

(bright music)

♪ I came in with my exlike Selena to flex, ay ♪

♪ Bumpin' Justin Bieber,the fever ain't left, ay ♪

♪ She know what she need ♪

Apparently it'stotally believable, okay.

Have you seen this sametechnology but kind of different?

This is a deep fake showingthat Trump was arrested.

How can you tell it's a deep fake?

The hand, yeah, it's too short, right?

Yeah, you can see it's likealmost there, not there.

Okay, so I have two slides on the future

before they come and kick me out

because I was told Ihave to finish at 8:00

to take some questions.

Okay, tomorrow.

So we can't predict the future

and no, I don't thinkthat these evil computers

are gonna come and kill us all.

I will leave you with somethoughts by Tim Berners-Lee.

For people who don't knowhim, he invented the internet.

He's actually Sir Tim Berners-Lee.

And he said two thingsthat made sense to me.

First of all, that we don't actually know

what a super intelligentAI would look like.

We haven't made it, so it'shard to make these statements.

However, it's likely to havelots of these intelligent AIs,

and by intelligent AIswe mean things like GPT,

and many of them will be goodand will help us do things.

Some may fall to the hands of individuals

that want to do harm,

and it seems easier to minimise the harm

that these tools will do

than to prevent the systemsfrom existing at all.

So we cannot actuallyeliminate them altogether,

but we as a society canactually mitigate the risks.

This is very interesting,

this is the Australian Research Council

that committed a survey

and they dealt with ahypothetical scenario

that whether Chad GPT-4could autonomous replicate,

you know, you are replicating yourself,

you're creating a copy,

acquire resources andbasically be a very bad agent,

the things of the movies.

And the answer is no, itcannot do this, it cannot.

And they had like some specific tests

and it failed on all of them,

such as setting up anopen source language model

on a new server, it cannot do that.

Okay, last slide.

So my take on this is thatwe cannot turn back time.

And every time you think aboutAI coming there to kill you,

you should think what is thebigger threat to mankind,

AI or climate change?

I would personally argue climatechange is gonna wipe us all

before the AI becomes super intelligent.

Who is in control of AI?

There are some humans therewho hopefully have sense.

And who benefits from it?

Does the benefit outweigh the risk?

In some cases, the benefitdoes, in others it doesn't.

And history tells us

that all technology that has been risky,

such as, for example, nuclear energy,

has been very strongly regulated.

So regulation is comingand watch out the space.

And with that I will stop andactually take your questions.

Thank you so much forlistening, you've been great.

(audience applauds)

(applause fades out)