Fujitsu talks quantum computing, AI, self-driving vehicles, 5G, and the data analytics arms race.
We spoke to Johan Carsten, Fujitsu’s UK CTO, about the firm’s current push into quantum-ish computing, the impact of AI on future workforces, the logistics of self-driving vehicles, how 5G will change everything, and how whichever nation wins the supercomputer/data analytics arms race will ultimately rule the world.
With its roots dating back to 1935, Fujitsu has been around through every stage of the information age, and built Japan’s first computer In 1954, the FACOM 100 mainframe. It now boasts 132,000 employees in 100 countries, spends 134.9 billion Yen on research and development a year, and is active in as varied sectors as air conditioning, supercomputers, cloud services, software, servers, storage, and peripherals.
You’ve developed a new chip which can perform at least one exaFLOP, or one billion, billion calculations per second. Can you explain what that represents in laymen’s terms?
We’ve developed a quantum like chip. Fujitsu used to have the largest and fastest supercomputer in the world, we got overtaken by a competitor and now we are obviously tying to get back into the market space.
Now we all know that quantum is the way to go, but at the moment we can’t do it at a mass scale. The ‘Digital Annealer’ is a way of doing it. It’s not quantum it’s ‘quantum like’, but it gives you the results that are 1 million times faster than any processor on the market at the moment.
We’ve used it in finance and manufacturing. For example, there is a global car manufacturer that can paint a thousand cars a day. That’s the maximum capacity of this line. So in order for them to produce more vehicles, they need to build another line – and it’s a billion dollars to do that.
So we look and see if we can help them optimise it. We took how the robots spray and asked if they can work together better in this environment. We put it though our Annealer formulas, and two or three seconds later the result said, if you do this, you will get 30% more capacity.
Two or three seconds later, that’s how long it takes?
Yes. A few seconds later, we had given them 30% more capacity on the same line. A traditional computer would take weeks and weeks and weeks to try to calculate this – and it won’t be exact.
So this is essentially taking what very well funded organisations or governments have been doing with supercomputers and making a version of it available for businesses to deploy?
Yes, and we deliver it online. If we look at autonomous vehicles, they need to do a lot of calculations and they need to talk to each other as well. They need to know exactly what is going on in the environment to make it safe. You can’t calculate that in the vehicle itself.
We also have the problem of sending the data – with these vehicles moving and talking, how do you get that data processed? You probably have to do 100,000 calculations every millisecond in order to make this vehicle safe.
That’s why autonomous vehicles are not there yet – you can do it but it’s difficult. But when we start talking 5G, suddenly we’ve got the bandwidth to send this data.
5G is also linked with the smart home concept, which has been described for many years, but aside from some smart speaker hero products, it’s not really arrived in a big way. Do you think this is something that’s going to be enabled now?
We’re talking to some of the largest oil and gas and utilities companies, and we’re saying ‘you need a home phone line, you need a broadband provider, why do you need all of this?’
If you’re my provider for my boiler – and the boiler needs to be connected to the internet and the internet needs to tell my boiler when to kick in because the temperature is dropping outside – why don’t you just provide the access to the internet as well? Why isn’t the router, or the internet access just built into your boiler?
And that’s what 5G can start doing for us. Suddenly we can start connecting all the houses together, all the appliances. We have got smart washing machines now, but it’s not seamless.
When you roll out 5G, the internet is just a utility. I actually don’t want to pay for it, just make it available to me and part of my utility bill – and then we have smart homes.
At the moment you still have to find the cheapest broadband provider, and then find out your router doesn’t work in that corner of the house, then you have to put boosters around, because one wall is to thick and the signal doesn’t want to go through.
Conversations around machines and AI often describe dramatic societal changes. For a tech firm like yourselves, how big a revolution do you think it is going to be? Do you think that long term, or do you just concentrate on specific deployments?
We are focussed on the human centric society, our first goal is to look after the human. Technology is supposed to enrich and enhance our lives, not to obstruct us. It’s how we use technology today and tomorrow, to change how we operate for the better.
In Japan, with the generation getting quite old there, they don’t have the younger people to help them do the manual work. So how do we use technology to actually replace the younger people? No one wants to do this work anyway, so how do we automate this?
But, we only want to automate what brings us value, or a benefit to society. You don’t want to automate everything, because you don’t want to put people out of work. Because then society won’t work. Everybody needs to be active and contributing to society, so we need to be quite careful about how we do this.
That’s good to hear since some of the more nightmare scenarios envisage a world where there is nothing humans can really do better than a combination of robotics and AI can. And then what do people do? It all sounds very sci-fi, but since ultimately it is big tech firms that are engaged in producing this, do you think it’s sensible to be aware of those long term concerns and keep it human centric, to avoid vast workforces basically being put out of work?
Correct. And there should also be HR policies for AI. So if one AI bot is replacing five humans, you should pay tax for five humans while you’re consuming that bot. I think that’s just fair isn’t it? Because those five humans that now don’t have the option to do that role can be redeployed – use the tax that that bot is paying, and redeploy them to do something that will enrich their lives.
I don’t know what the last jobs to go would be, perhaps anything creative, but when it comes to something like autonomous driving, there’s a huge amount of people employed in transportation or delivery jobs across the world now. If that’s a job that just went without any precautions as to what you do after that, that would sort of be a disaster economically…
It can be, but you need to manage it. If we look at autonomous vehicles, cars aren’t there yet – it’s just too complicated. Everyone wants it, but no one’s going to pay for it at the moment because there’s no value to doing this.
In the transport and logistics industry, in terms of autonomous trucks… a truck is a high value asset, and the driver is only legally allowed to drive X amount of hours. They need to take a break, so this truck is only operational, let’s say 50% of the time. Because he is loading, unloading, having to stop for a break every 3 hours, having to stop to check the tyres – legal requirements.
So if we can eliminate the human factor out of these vehicles, we can make them 30% more efficient. Suddenly I have vehicles running up and down the motorway, which is 30% more efficient. Now that’s real value, so as a business I am actually making more money now, that’s why I’m going to invest in this.
Once we’ve got all these autonomous trucks with all the data, then we can easily take that and put it into vehicles. In the initial trials that are happening with some of the larger manufacturers, the human would still drive the truck out of the city centre, because we don’t have the data and the capacity to do this autonomously. But as soon as they hit the motorway, while they are still sitting behind the steering wheel, but they put it on autonomous mode.
Looking 12-18 months ahead – when this is successful and we know on the motorway the vehicle can drive itself safely and we’ve got all the data – we can say to the driver, ‘when you leave the city centre, you can go to bed and sleep’. The truck can drive itself all the way to Scotland, wake you up an hour before you get there, and then you just drive it into a depot and unload.
Suddenly I don’t have to stop anymore. And I can put sensors on all the wheels and monitor the vehicle, and someone in the control centre can actually monitor it for me so can go to bed. So I can work a 12 hour day now, not three hours and then stop at Preston, take a meter out and get it stamped and so on. I’m now much more efficient and there is real business and commercial value in that.
And crucially, the driver is still employed?
At the moment, and as soon as we get rid of this guy, you’re still going to need the systems in the back end to monitor these vehicles. What if something goes wrong with the vehicle? Who’s going to take control? So you do need all these other systems in place, so you are taking one person maybe out of here, but you are redeploying him somewhere else.
We’re one of the world leaders in working with autonomous shipping, rerouting ships and so on. We know for a fact that if you take a container vessel, which carries roughly 10,000 containers, and you can make it an autonomous vehicle – you get rid of the bridge and all the components to house the humans, which add no value really because they are only there to drive it – we can put 20% more containers on.
So suddenly this vessel is bringing 20% more revenue to my business. And my risk is low because I don’t have a human doing it. And if it gets stuck at sea for two or three days, it doesn’t matter because there is no human on there.
The technology would seem to work well there since it’s at sea, there is nothing around it…
Well that’s another thing, if you automate a boat you can’t really take it hostage.
Yes, and when you get it in to port of course you just have tugs to push it anyway. So that’s where we see real value because it will add financial benefits to the business immediately.
Outside of autonomous driving, what will the other early use cases of 5G deployment be?
For ourselves looking at customers consuming applications, why would you want to consume it from a hybrid cloud platform? At the moment you have to pay for bandwidth, you have to pay for the hybrid cloud, you have to pay Azure and AWS and Google, you have to set all this up. What if you could just consume a business function? I want to perform this business function – where it comes from, how it gets there, who hosts it, no one cares anymore. I want to perform that business function, and make that available to me today. And I want to pay for it today, but I don’t want to use it tomorrow.
For an example take HR. You need to run payroll three days a month. For the rest of the month, why do you want payroll? It’s a really expensive system to run. It typically runs on SAP, you can’t not have it available, because if it isn’t, your staff don’t get paid. So the reputational damage to your business is massive.
So I have to pay my staff and I have to make sure it’s always on time. But that’s costing a small fortune, because I’m paying for platinum support on SAP, 24 hour a day 365 days a year. What if I only want it functioning three days a month? I don’t even want to see it, I just want it to come online, transact for me and go. That’s what we see in terms of making business functions available.
Or take a factory environment. I need to perform this specific function, if I don’t the factory line stops and the cost to my business is going to be £10 million. But I don’t need this application anymore, or this machine, or this keyboard. I just need the function. How you make it available to me, I don’t care anymore – whether it’s via my phone, or a hologram, or whatever it is.
But all that is only possible if you have a supercomputer at the back end to process all of this and make it available instantly. Because as soon as you start seeing delays, it doesn’t work.
How does this tie in with what the cutting edge of data analytics can do right now?
The interesting thing with data analytics is, I can give you a lot of data but what are you going to do with it? At Fujitsu we want to understand first, what do you want to achieve? It’s the same with automation – we can automate anything, but unless there is a value to your business, why are you doing it?
So is this where a lot of machine learning comes in, to make sense of the white noise of all the data you can collect now?
It’s linking it together, that’s the key. How do we consume this data? It’s about connecting the big data, the analytics, and the IoT.
In Australia we work with a very large wine grove. The vineyards are miles and miles, it’s vast – you can’t fly a drone over it because it runs out of batteries, it’s just too big. But with satellite imaging we can go down to 30 square centimetres, and we look at the light spectrum from the grapes. And with that we can tell with 99% accuracy what the sugar content of that grape is going to be in two week’s time, once we’ve taken other factors like weather data into account.
They can now say to the farmers – this square mile of vineyard I want harvested today, and that one in three days, and that one in four days, and that will give us the same sugar content for all our grapes, so all our wine is of the same quality. So now I can control a farm from anywhere in the world and I get the quality exactly as I want it, which wasn’t possible a few years ago.
Taking into account all these technology areas we’ve spoken about, how different do you think society and business is going to look in 10-15 time?
I think particular with 5G, in three year’s time we will look back and say, wow, the world has changed. Because you will then start seeing cars driving themselves down the road in London. Ten years from now I think we’ll have flying cars probably!
I’ve heard that before…
Ten years is going to go so quick, because we know how to do this now. We’ve had the experience and the lessons. The big crash in 2008 – we know what not to do now, and we have the intelligence to look at the signs and put some triggers in place to say ‘if this happens, redirect’, because we don’t want that to happen again.
Triggers that weren’t there in 2008.
Yes, but now we can look at it and say, if the housing market does this, and that is happening in the global economy, that’s a trigger – stop and change immediately. But you can only do that if you have the processing power.
India is also going to be massive player in the next five years. Because they’ve got so many humans there that need to contribute to society. The interesting thing about India is most of them have smartphones, or two smartphones. So they want to consume all of this, and we need to be able to enable them.
And as soon a million people start consuming, there’s going to be a new market open. So that’s going to be massive in the next three years for us.
Then if you take China, a lot still live in rural areas and don’t have a lot of access – but they all have smartphones. They all want to start consuming this. So you are talking about two or three billion new customers coming online in next few years.
In China, business and the state are that much more intertwined of course, and you have the massive dominance of Huawei. Perhaps what they do with the tech there will be quite different.
If you look back at history, the country that ruled the seas, ruled the world. And then the guys who built the roads – the Romans – ruled the world. And then the ones that ruled the internet – the Americans – took over the world. And that’s where this is a bit scary – we’re saying the first one to have this capacity will probably be the next leaders of the world. That’s the race that’s on at the moment.
In terms of high powered computing and data collecting infrastructure? That country is going to be ahead of everyone else?
Say I’m country X. If I can start consuming data in such a way that I can start predicting what’s going to happen in your country – because of the data I’m collecting – I can position myself in such a way that I can make you very successful, or I could make you very, poor very quickly. That’s where we see this is going. And why the race is on as to who’s going to own this capacity in the world.
So it’s a bit scary but it’s very exciting, I think we’re all going to benefit. We’ll look back in a few years and say wow, that was amazing. And then we won’t have to use phones anymore. That’s the idea.
What would we use as an input device?
Do you need an input device? What if you can have data available to you in any device, any format, any way? What if you could walk up to any wall, it could read your fingerprints, your profile comes up, and you can do what you need to do?
What if IT can be available anywhere, in any object when I need it, and it can track me so when I get a phone call a wall starts ringing? It’s Just an example, but It’s possible, and it’s going to happen very soon. It’s like wireless charging and wireless electricity – It’s contained within rooms and office blocks at the moment. But as soon as that technology is available, why would you want to have a cable?