Man versus Machine Talking Business


Download Subtitles

SRT

ASS


Man versus Machine

Karishma Vaswani talks to experts about the challenges and opportunities of machine learning and artificial intelligence.


Similar Content

Browse content similar to Man versus Machine. Check below for episodes and series from the same categories and more!

Transcript


LineFromTo

Imagine a future where machines permeate every aspect of our lives.

:00:00.:00:07.

That reality is frankly already upon us.

:00:08.:00:10.

Artificial intelligence is in our homes, in our cars,

:00:11.:00:14.

But, how will artificial intelligence or AI changed the way

:00:15.:00:20.

we do business in Asia bastion Mark and what challenges will it bring?

:00:21.:00:25.

All of that and more on this edition of Talking Business.

:00:26.:00:51.

There is so much talk about artificial intelligence these

:00:52.:00:57.

And deep learning, another phrase that is often thrown about.

:00:58.:01:03.

We'll be speaking with a panel of experts in just a moment

:01:04.:01:06.

Artificial intelligence is a blanket term for any kind of technology

:01:07.:01:14.

or software that enables machines to perform tasks on their own.

:01:15.:01:20.

It is very much a part of our everyday lives.

:01:21.:01:23.

From something as simple as automated factories

:01:24.:01:27.

to intelligent personal assistance like Google Assistant,

:01:28.:01:29.

But AI is slowly evolving into something much more complex.

:01:30.:01:39.

Deep learning is one branch of machine learning that teaches

:01:40.:01:47.

computers to process information on several different

:01:48.:01:50.

layers at the same time, much like neural networks

:01:51.:01:54.

Part of that is for machines to learn for themselves

:01:55.:01:59.

by the accumulating and extracting patterns from huge amounts

:02:00.:02:02.

But the eventual goal is for machines to incorporate

:02:03.:02:14.

aspects of human intuition and emotional intelligence

:02:15.:02:16.

As we teach machines to be more independent of human beings,

:02:17.:02:21.

we are also asking questions about what could potentially go

:02:22.:02:24.

wrong, and more importantly, what all who the smart machines

:02:25.:02:30.

of the future will be held accountable to.

:02:31.:02:36.

Artificial intelligence or AI is already being used

:02:37.:02:46.

in many industry and business processing today.

:02:47.:02:49.

Deep learning within AI is really the next frontier.

:02:50.:02:51.

But what kinds of possibilities does this open up for businesses?

:02:52.:02:56.

With us we have Modar Alaoui the founder and CEO of deep learning

:02:57.:03:00.

It makes, amongst other things, Eyeris is a deep emotion recognition

:03:01.:03:10.

software that reads facial micro-expressions in real-time.

:03:11.:03:12.

Vasilios Vonikakis, a research scientist

:03:13.:03:18.

with the Advances Digital Sciences Center.

:03:19.:03:31.

And the chief technology officer of Coin Tech, dealing with deep

:03:32.:03:34.

learning and automation in the transport and

:03:35.:03:36.

Thank you, gentlemen, for joining us today,

:03:37.:03:38.

to talk about artificial intelligence and deep learning.

:03:39.:03:41.

I'd like to start with you, Modar. What is deep learning?

:03:42.:03:44.

How would you describe it to our audience?

:03:45.:03:46.

In image recognition, deep learning enables algorithms

:03:47.:03:48.

to recognise the same face from different angles,

:03:49.:03:54.

and in text recognition, it would try to identify the text

:03:55.:03:59.

In, for example, speech recognition, it would try to identify

:04:00.:04:06.

the same keyword or phrases from different accents.

:04:07.:04:08.

So, it needs a lot of data to be able to do this.

:04:09.:04:12.

Typically, a lot of data but a lot of quality data as well

:04:13.:04:15.

so not just quantity, but also the quality of the data.

:04:16.:04:17.

Yuris, in your business, can you explain how you use deep

:04:18.:04:21.

So, we do a couple of things, basically.

:04:22.:04:27.

One of the things we do is we do recoup addition of passing trains

:04:28.:04:32.

and different kinds of things about it, what is normal,

:04:33.:04:40.

what is not, starting from very simple things like numbers,

:04:41.:04:42.

which was initially how we started it.

:04:43.:04:45.

And, then, moving on to identifying different kinds of issues.

:04:46.:04:53.

Lately, we also use this to identify things which, from video feeds,

:04:54.:04:56.

what is happening in the city, like how many pedestrians, cyclists,

:04:57.:04:59.

cyclists with helmets, cyclists without helmets,

:05:00.:05:00.

with child seats, without trial seats in the back, transport

:05:01.:05:03.

all the information you can get from any video stream,

:05:04.:05:05.

transferring this into actually valuable real-life data.

:05:06.:05:07.

Vasilios, how do you see deep learning being useful in Asia?

:05:08.:05:11.

Because you are working here at the advanced science

:05:12.:05:14.

How far along has the research and the technology become advanced?

:05:15.:05:29.

Basically, there is a lot of research in this field

:05:30.:05:31.

going on in the local universities and in many different research

:05:32.:05:34.

I have seen a lot of start-ups, actually, utilising this type

:05:35.:05:37.

of technology, start-ups that have to do with observer recognition,

:05:38.:05:40.

in order to help shoppers find similar looking items from a big

:05:41.:05:45.

database and going directly and buying this item online.

:05:46.:05:51.

It has also been used in autonomous vehicles, a lot.

:05:52.:05:57.

This is a very ongoing research in this direction.

:05:58.:06:02.

Modar, where do you see the most advanced usage

:06:03.:06:07.

Image recognition still holds a number of challenges,

:06:08.:06:15.

simply because we cannot map the entire world in a very

:06:16.:06:22.

short period of time, nor can we do it that efficiently.

:06:23.:06:27.

So, I do believe that is an area that is also ripe for research.

:06:28.:06:30.

Not in Asia, but in the world in general.

:06:31.:06:33.

What I'm interested in is the deep learning that is trying to break.

:06:34.:06:44.

Because I understand having developed from machine learning,

:06:45.:06:53.

as a result of the limitations of machine learning,

:06:54.:06:55.

would it be fair to say that there are limitations

:06:56.:06:58.

I would say that deep learning is just the next

:06:59.:07:01.

step, not even close to full artificial intelligence.

:07:02.:07:03.

The limitations are set there are regarding the ideas

:07:04.:07:13.

about what are the tasks that we want deep learning

:07:14.:07:15.

Do we want to recognise images? Do we want to recognise speech?

:07:16.:07:20.

What kind of images are we searching for?

:07:21.:07:22.

So, it is not about making decisions.

:07:23.:07:24.

It is actually about doing specific tasks at a certain level,

:07:25.:07:27.

so precision is also important, to understand what is relevant

:07:28.:07:29.

and what is an acceptable level of precision, for example.

:07:30.:07:31.

So, to follow up on that, how is it being used in real life?

:07:32.:07:35.

If you can talk specifically to some industries, what kinds of functions

:07:36.:07:38.

or what kind of people is deep learning replacing?

:07:39.:07:40.

Yes, some examples of the things we are doing.

:07:41.:07:44.

Like, when you talk about on the railroad,

:07:45.:07:48.

passing over the train or the wagon between two companies or two

:07:49.:07:51.

motocross and points or something like that,

:07:52.:07:56.

we actually have a bunch of people running around doing

:07:57.:07:59.

different kinds of checks, visually checking it,

:08:00.:08:09.

doing this stuff 24 hours a day, seven days a week at all the time.

:08:10.:08:21.

So, what we do, we actually managed to replace them,

:08:22.:08:24.

to take the necessity of way, and this is a very simple

:08:25.:08:27.

and but there are still thousands of people who are actually

:08:28.:08:29.

So, Vassilios, currently, deep learning isn't taking over

:08:30.:08:32.

the entire supply chain of the labour force employed

:08:33.:08:34.

Do you see that changing in the future?

:08:35.:08:46.

Do you see more businesses employing deep learning and replacing people

:08:47.:08:49.

I think that I can see in the future a lot of jobs being displaced by AI.

:08:50.:08:56.

So, many people are talking about new

:08:57.:08:58.

Yes, for sure, but I think the rate at which this happens will not be

:08:59.:09:04.

as great as the rate with which the jobs become obsolete.

:09:05.:09:07.

So, for example, if you take autonomous vehicles,

:09:08.:09:10.

actually went this bull happen, you can say that a lot of jobs

:09:11.:09:15.

will be displaced there, like truck drivers, taxi

:09:16.:09:23.

Yes, I can add on top of this as well, a lot

:09:24.:09:30.

You talk about security guards in the malls and everything else,

:09:31.:09:34.

that is another industry where there are a lot

:09:35.:09:36.

of people working for it, they are considerably not

:09:37.:09:38.

the highest level of jobs, they are not that technology

:09:39.:09:41.

intensive, and these are very likely to be either replaced or optimised

:09:42.:09:44.

as part of applying more and more technology solutions.

:09:45.:09:47.

Modar, do you see this becoming a sort of long-running industry

:09:48.:09:49.

trend for the future or will deep learning, as we have been

:09:50.:09:57.

discussing, hit a barrier and another breakthrough has

:09:58.:10:00.

As in everything, technology will definitely be a barrier,

:10:01.:10:04.

Just like what we have seen with machine learning and other

:10:05.:10:15.

But I do have to say that deep learning is still in its infancy,

:10:16.:10:19.

there is still a long way for it to mature, and to disrupt a number

:10:20.:10:23.

of industries and lives before we can see it being potentially

:10:24.:10:26.

So, as it is now, deep learning requires a lot of data and a lot

:10:27.:10:41.

of good quality data, as you mentioned.

:10:42.:10:43.

So, as it is now, deep learning requires a lot of data and a lot

:10:44.:10:48.

of good quality data, as you mentioned.

:10:49.:10:51.

So, many people are working towards making this technology able

:10:52.:10:54.

to generalise better with less number of data,

:10:55.:10:56.

How can you train in the neural network with fewer examples in order

:10:57.:11:01.

to generalise as well as it will be with a large area of data assets?

:11:02.:11:07.

Another direction is how to make these deep learning networks learn

:11:08.:11:10.

by themselves unsupervised without having to tell them exactly.

:11:11.:11:13.

But on that point of unsupervised learning, we have seen incidents

:11:14.:11:17.

already just recently of when unsupervised learning

:11:18.:11:22.

or an attempt to make this kind of technology more independent,

:11:23.:11:29.

One example that comes to mind, for example, the hate tweets.

:11:30.:11:36.

That is an example of trying to educate a chat bot, if you will,

:11:37.:11:40.

to do a certain thing, has ended up having the adverse

:11:41.:11:47.

Supervised is here to stay they because at least we know that,

:11:48.:11:55.

you know, ground truth data is being labelled basically by human

:11:56.:12:03.

beings, therefore we are stirring up the level of accuracy there.

:12:04.:12:12.

Unsupervised is still going to have a certain number of question

:12:13.:12:15.

marks around it at least for the foreseeable future.

:12:16.:12:17.

There is a lot of effort being conducted there.

:12:18.:12:19.

One thing I wanted to add to this also is that it is not just that

:12:20.:12:24.

deep learning technology which has to advance, we also have to add arms

:12:25.:12:27.

The ability to interact with people, the ability to see the world,

:12:28.:12:41.

so we talk about all sorts of vision development which include also

:12:42.:12:44.

sound, image, video, whatever you are processing that

:12:45.:12:45.

So, it has to be hand-in-hand to become better, so it is not just

:12:46.:12:50.

Hold that thought, gentlemen, we are going to be talking

:12:51.:12:54.

about deep learning in just a few minutes again but before that,

:12:55.:12:57.

let's take a tongue in cheek look at deep learning

:12:58.:13:00.

Deep learning, it is the latest phrase that I'm going to impress

:13:01.:13:07.

But it will have applications that might actually surprise you.

:13:08.:13:15.

I've come to the seat of learning here at Trinity College Dublin,

:13:16.:13:18.

to find out about one of the surprising ways in which deep

:13:19.:13:21.

Deep learning, it is a great phrase, isn't it?

:13:22.:13:28.

It combines two wonderful words, learning, what

:13:29.:13:31.

You take away key learnings from the meeting, use it

:13:32.:13:36.

at the foot of the master, learning lots of wisdom and deep,

:13:37.:13:51.

Deep Thought was the famous computer in the Hitchhiker's

:13:52.:13:53.

Deep Blue was the first computer to beta human at chess,

:13:54.:13:57.

so combine the two of them, it has got to be good, right?

:13:58.:14:00.

It certainly sounds better than careful, now.

:14:01.:14:01.

Let's meet Dr Ahmed Salim at the Connect Research Institute

:14:02.:14:04.

He's developed, along with Dr Mohammed Elgary,

:14:05.:14:09.

a portrait painting technique which is based on deep learning.

:14:10.:14:11.

This is how Frank Gough would paint me.

:14:12.:14:19.

What we have here enables you to photograph and the van Gogh

:14:20.:14:25.

painting, the model by itself, produces these results.

:14:26.:14:28.

Look at me. A masterpiece.

:14:29.:14:30.

We are using these models to do feature extraction.

:14:31.:14:33.

Then we upload these features into our mathematical model.

:14:34.:14:36.

It will generate the painting that we are looking for.

:14:37.:14:41.

Professor Linda Doyle has a view on the bigger picture.

:14:42.:14:48.

Actually, it is true, the eyes do follow you around the room.

:14:49.:14:53.

Artificial intelligence is already actually embedded

:14:54.:14:55.

in the world around us, and people don't realise that.

:14:56.:14:58.

So, when they think of artificial intelligence, they tend to think

:14:59.:15:02.

about robots and science fiction and things they have seen on TV,

:15:03.:15:05.

whereas in fact in everything that you do, in Google,

:15:06.:15:08.

in Facebook, and in all the applications that

:15:09.:15:10.

you use, there is some kind of artificial intelligence there.

:15:11.:15:12.

If you have an algorithm that shows you how to avoid having an accident,

:15:13.:15:17.

there has to be some decision-making process that says A is more

:15:18.:15:25.

important than B so I'll bump into B rather than A.

:15:26.:15:31.

That have to be put in by humans somewhere.

:15:32.:15:33.

Yes, there is a lot of to think about when it comes to artificial

:15:34.:15:36.

intelligence and this iterative machine learning.

:15:37.:15:38.

So, when you think about artificial intelligence, what you are saying

:15:39.:15:42.

is that the algorithms running in the background that are learning

:15:43.:15:44.

behaviour and making decisions, but they don't do that in a vacuum,

:15:45.:15:47.

If you believe that human beings are a finite number

:15:48.:15:51.

of cells or whatever it is, trillions, eventually,

:15:52.:15:53.

we are going to make artificial intelligence that will be smarter

:15:54.:15:56.

We have to grapple with the ethical issues now because artificial

:15:57.:16:00.

intelligence isn't something of the future, it is

:16:01.:16:02.

Mm, where to immortalise myself on the wall?

:16:03.:16:09.

Of course, when you talk about anything like AI,

:16:10.:16:15.

there are still so many questions that remain because so much

:16:16.:16:18.

One thing I can say, now, though, is a rebuttal to anyone who ever

:16:19.:16:28.

If you like that, you can catch more of those videos

:16:29.:16:34.

Now, back to some serious business with our panel of experts.

:16:35.:16:43.

Today we have with us Modar Alaoui from Eyeris.

:16:44.:16:45.

Vassilios Vonikakis with the Advanced Digital Sciences

:16:46.:16:48.

Before the break, we were discussing the possibilities of deep learning.

:16:49.:16:58.

Now I want to talk about some of the fears that have cropped up,

:16:59.:17:01.

There is always this discussion about how machines

:17:02.:17:11.

Who is to be held responsible with the current technology

:17:12.:17:17.

that we have in place today for example, autonomous cars,

:17:18.:17:20.

if there is an accident when an autonomous car

:17:21.:17:22.

Maybe, Modar, you could pick that up.

:17:23.:17:26.

Yeah, well, there has to be a certain, you know,

:17:27.:17:29.

I do potentially think that there will have to be some sort

:17:30.:17:36.

I also leave that companies and governments and authorities

:17:37.:17:42.

will have at some point to decide how the AI should respond

:17:43.:17:51.

For example, the autonomous vehicle accidents, recently an announcement

:17:52.:18:00.

was made by one of the automotive companies, they made the choice

:18:01.:18:06.

for the AI that is going to be inside the vehicle to protect

:18:07.:18:11.

the driver at any given time, whether that protects the pedestrian

:18:12.:18:18.

or the person that is in front of the vehicle.

:18:19.:18:25.

the driver at any given time, rather that protects the pedestrian

:18:26.:18:27.

or the person that is in front of the vehicle.

:18:28.:18:30.

Who to blame and how to think about how AI would respond

:18:31.:18:34.

But doesn't this bring up a whole host of questions about how this

:18:35.:18:39.

kind of technology is going to be applied in daily life?

:18:40.:18:42.

Given the fact that currently it is on track to replace so many

:18:43.:18:47.

jobs, and we were talking about it before the break, it is going to be

:18:48.:18:52.

something that we are going to see more of in the future,

:18:53.:18:55.

shouldn't we be trying to answer these questions

:18:56.:18:57.

From my perspective, I would like to use a very

:18:58.:19:02.

One of the Ps would be for precision.

:19:03.:19:11.

That will differ from case to case, from application to application.

:19:12.:19:14.

Another one would be personal data protection.

:19:15.:19:15.

The third thing would be predictability which would be

:19:16.:19:21.

actually to do with the ability to predict something.

:19:22.:19:23.

And the fourth one would be the productivity issue.

:19:24.:19:27.

Where we say that machines can be more productive

:19:28.:19:29.

If you look at this perspective, then the question of ethics

:19:30.:19:33.

Because it is not just about autonomous vehicles,

:19:34.:19:37.

it is also about health care, it is also about diagnostics,

:19:38.:19:39.

and so, the way people perceive trust is actually changing

:19:40.:19:42.

And I think that is also one of the changes which isn't

:19:43.:19:49.

going to be about technology, isn't going to be about science,

:19:50.:19:52.

it is going to be one of the changes which will be about the perception

:19:53.:19:56.

of people, the way they see the world, the way they see

:19:57.:19:58.

the tech ologies they trust, and then of course we can talk

:19:59.:20:01.

principles about decision-making. into the deep learning,

:20:02.:20:11.

But the more complex systems get, the more demands there will be. How

:20:12.:20:19.

do we make the technology fair for everyone? Click for example Google

:20:20.:20:27.

along with Amazon and Facebook and Microsoft, if I were to mention

:20:28.:20:33.

these as the four or five horsemen companies, they have the ability to

:20:34.:20:40.

time they recognise that using it time they recognise that using it

:20:41.:20:47.

for the greater good of advancing humanity is of paramount role than

:20:48.:20:57.

to use it otherwise. What happens if or when an entity like this comes an

:20:58.:21:03.

independent freethinking entity? I know we are delving into the realm

:21:04.:21:08.

of science fiction here but, you know, it was far-fetched about five

:21:09.:21:11.

or ten years ago to talk about deep learning, who is to say it is deep

:21:12.:21:18.

-- far-fetched to talk about this? If you look at where we are now, we

:21:19.:21:23.

are still miles, not even Miles, millions of miles away from having

:21:24.:21:29.

an entity like this. Unless there is some hidden research which nobody

:21:30.:21:33.

knows about. Most of the research which is open is still focusing on

:21:34.:21:37.

specific tasks, it is still about being better in a very specific,

:21:38.:21:46.

narrow task. It is about being good about decision-making, being

:21:47.:21:52.

cognitive, making decisions on a larger scale, we are still very far

:21:53.:21:58.

away from this. Do you think we are getting closer to the stage where

:21:59.:22:01.

machines will be able to make decisions that currently humans do?

:22:02.:22:07.

Gradually, yes, but it will take some time. I expect that the things

:22:08.:22:13.

that we already have will become better and better over time.

:22:14.:22:17.

Usually, I think that weekend to overestimate what we can do in the

:22:18.:22:23.

short term. We tend to overestimate what we can do in the long run. I

:22:24.:22:30.

think, generally, AI will become better and better but it will always

:22:31.:22:35.

be applications specific, at least for the near future. Thank you

:22:36.:22:40.

gentlemen for joining us for this discussion on deep learning. That is

:22:41.:22:44.

it for this edition of Talking Business in Singapore.

:22:45.:23:03.

After a great and misty day, a murky night on the way tonight. Be wary

:23:04.:23:11.

and the Woakes, fog would

:23:12.:23:12.