TechTalk ep 16: The one where Anand Rao talks to us about AI (Part 2)

Paramita: Hello, welcome to PwC Luxembourg TechTalk. Today, we continue our conversation with Dr. Anand Rao as he answers some of our burning questions on all things AI.

Paramita: I promised my colleagues because when they knew that you were coming and that I'm going to interview you, my colleagues they were quite interested in asking you a couple of questions. So the first question comes from Andreas Braun. He's actually a senior manager in AI here. So he asks: in Europe, we often have this doom and gloom view that we are desperately behind in AI compared to the US and China with no hope left. Do you agree completely? Or from your outside perspective are there some European strengths to leverage? Because I think and he is right somehow because they say that you know we are good in research but when it comes to commercialisation we are not that good.

Anand: Yeah I know there's been a lot of talk about China and US as being the sort of the two big countries of AI. Yes they are pouring in a lot of money. I mean US is sort of much more naturally entrepreneurial. And China is pouring a lot of money, a lot of research and so on. But I wouldn't take that as a sign that there is "no hope for EU". And I think to me Europe has always taken a very measured stance and more thoughtful stance around some of these issues.

And to me AI is just starting and again no one is to say that there is one or the other which has the sort of the right approach. There is again a spectrum of different views and values. Again just as a society we don't all have the same kind of cultural values and cultural norms. There's nothing wrong with that. That's what I think makes humanity very much unlike robots... not the same thing. So we all have our individuality and we all have very different cultures. And I think you're actually seeing I think in a very interesting way those things being now discussed with respect to AI. That's why I don't believe there'll be one monolithic view of AI. And when some of the great minds in fact talk about humans versus machines they think of one AI as if the AI is like a God with all the right answers and somehow we as humans are going to be subservient to this one Oracle called AI. I don't think that happens and that that's likely to happen as well. Just as the Europeans will mould the AI to their liking, to their values so much as what the Chinese would or the Americans would. So in fact, there will be multiplicity of AI reflecting the culture, the values and the norms. And again after all we're talking about now AI, AI ethics, and AI having certain values, having certain ethical code and so on and whose ethical code? Is there one humanity's ethical code? To some basic level yes. But there are also differences in that view and that's what you see reflected. So going back to this question I think... I don't think there is anything wrong with EU having a more measured approach thinking about its citizens, maybe a different set of values being part of the AI. And I know people worry about AI is all about data and of course China has a billion plus people more or less monopolies. Therefore, they can have lots of data. That's true. But the behaviour that they're capturing is Chinese behaviour. It's not so much Luxembourg behaviour or French or German behaviour. And when it comes to French and German I think the companies there can understand more of the behaviour of the French and the German and the Luxemburg countries right. So in that sense I think you would see more of a flourishing of multiple AIs with its own little changes and quirks in terms of what we think rather than one big thing called AI. So I wouldn't be too worried about EU missing out the boat. It's a long journey and I think there's nothing wrong in having very different kinds of vehicles to get there.

Paramita: Just like an extension of humanity.

Anand: Yeah very much so.

Paramita: So the second question comes from Pauline André. She asks: as the primary objective of AI should be supporting society and humanity, what roles governments can or will play in the future of AI?

Anand: Yeah. One thing that is very encouraging in this time around is a number of countries are now looking at AI very seriously. So again we are almost losing track of the number of countries that are announcing AI strategies but there are more than 30 plus countries now are very active in AI strategies and I think that some of the EU directives almost every European country is meant to produce a sort of a strategy document which all goes into EU. So in that sense that number is going to expand and explode beyond 50 to 60 countries at least having a nationally AI strategy. Now when it comes to a national AI strategy I think again we as PwC have been helping a number of countries, at least 20 plus the last time I counted, where we are helping the countries think through what are the different issues related with national AI strategies. And countries are looking at attracting investment, doing more R&D in AI, attracting the businesses and also creating the skills of the people both at the graduate level and at the high school level as well. So people are now sort of seriously considering what should our children be learning in the future in this AI dominated world. So those are all very interesting and important questions. We're also looking at people who are dispossessed due to job losses, how can we retrain them. That's something I see EU doing more often than some of the other countries. So in that sense I think that's one of the big parts. And then also we are looking at ethics what are some of the data that needs to be protected, citizens' data that needs to be protected. So everything around regulation, fairness, bias that we went through all of those things are part of what national AI strategies are actually looking at. So in that sense I think they have a very important role to play in both the opportunity side of AI as well as the risk side of AI. In fact we are just publishing and releasing a paper today from PwC Luxemburg and global which is targeted very much at is AI the next avenue for national competitive advantage which we believe it is and we talk about all of these different policy areas that people are looking at, countries are looking at in terms of AI.

Paramita: And just for our listeners if you want to consult the white paper, you can do that on our website. So the next question comes from Mary Carey. She says: if robots make everything and nobody has a job where will the money come from to pay for the products and services? And should robots that do the manual jobs, the jobs that we do, should they be taxed?

Anand: Yeah all very interesting questions. So now as I said in 1956 when the researchers, all very eminent researchers and founding fathers of AI came together, they said in 20 years all the jobs that humans do can and will be done by AI and in some cases even better is what they said. We still haven't quite reached there. Again now we are still saying that we'll have lots of jobs automated and that people will lose jobs. My own view on this is yes so there is definitely going to be automation. There will be definitely lots of tasks that will be automated but that doesn't mean that the jobs will disappear. So we need to be careful on how we look at jobs versus how we look at automation of the tasks. Let's say I have a hundred things to do in my job. And if the AI is automating 20 of those tasks so out of the 100 tasks let's say it's automates 20 tasks. Now it is often people think oh that means take five of these individuals. Each one gets 20. So you can eliminate one person's job. That's not... doesn't really work that way. So now if you remove 20 tasks from my my schedule I won't be using that time for doing something else. I'll be using that time more fruitfully to generate value to our clients. So maybe I'll I'll talk to more people, I'll talk to more clients. I'll be doing something useful so that 20 percent time that you have given me I'll be using it to add value. That's the same whether I'm in PwC, whether it's a bank manager or whether it's a retail person they'll interact with people more. Now that happens with the 20 percent time saving. Now if you do the reverse, if 80 percent of my tasks have been automated then trying to fill in 80 percent of new value maybe difficult. So there you can argue that if you combine a few people you can get rid of one. The truth is somewhere in between. So it's not just the 80 percent task automated or 20 percent that's automated. It's somewhere 30-40 percent. So the amount of people that are going to be released from automation I think is going to be much lower than what people are predicting. And also humans are very ingenuous in sort of creating various tasks. So we can fill up the time with whatever activities that we have. So we see more that the nature of the jobs changing. So people will interact with other people more or they might be using some of these tools to make things better. So you will see the nature of jobs changing as much or more than actually jobs getting lost.

Anand: So I'm not that much of a pessimist in terms of suddenly we'll have a whole lot of people losing their jobs and social unrest and so on. I think that's less likely a scenario, could happen, but that's less likely of a scenario. Now what should we do should we tax the robots? A lot of economists have been looking at this particular issue. Yes AI is likely to increase the divide between the people who have the knowledge or have the money versus people who don't. So the gap is going to actually increase... the income gap and the wealth gap between the haves and have-nots are likely to increase and not just within countries but across countries as well. So again some of the earlier questions on dominance of certain companies and certain countries are likely to be there. And that's why the governments will step in and make sure that those kinds of things don't happen. To me it's better to create the wealth and then look at the redistribution of that wealth than to actually prevent some of that value creation. So by taxing these robots we are essentially artificially trying to prevent some of these gains or the value from being added. In other words, we should add value through robotics, through AI and all of those and then say how do we equally distribute that value from various philosophical perspective whether it's sort of more socialistic a rearrangement or a more capitalistic rearrangement of the value.

Paramita: A question that is linked to this comes from a colleague who wants to stay anonymous. So who will financially benefit from the AI? Is it going to be the top one percent?

Anand: If we don't do anything, if the governments don't do anything then I think there is a natural tendency for the top one percent or even the one percent of the one percent so to be the beneficiaries of the gains just by the way I think that the whole system works. Now if we want to avoid that that's where I think the governments need to step in, one, and then there needs to be a much wider movement of different organisations consciously calling into question some of these aspects. So for example, the World Economic Forum has been very vocal about the digital divide or we call it the intelligence divide, digital is too broad, the AI divide or the intelligence divide I think becomes quite prominent. So how do you avoid that. How do you make sure that the benefits of what the companies do, what the individuals do, what the countries do are made explicit and are essentially spread across the globe. For example we have been working on AI for Earth along with the World Economic Forum and then we also have done a similar work for Microsoft. Microsoft's initiative which they announced a month back and that goes into how AI can be used very specifically on various aspects of climate change, biodiversity and all of the beneficial uses of AI. So that way people are looking at again the AI for good movement looks at the UN Sustainable Development Goals which is around reducing poverty, increasing equality of gender all of those sort of 17 different SDGs and seeing how AI can help. So there are various moves to actually have the distribution of the AI more broadly. I think that's probably more of a way to make sure that this doesn't go to just the one percent that is likely to benefit.

Paramita: You just answered another question there because another question that was linked to this topic was how... is AI going to respect environment and respond to urgent ecological needs and particularly the refuse, reduce, reuse, recycle, returned to Earth principles?

So I think you kind of answered that question... AI for good...

Anand: That's right. AI for good is a big group of people, almost a movement. You can see that coming in from different directions.

So obviously the UN is funding various initiatives. You have number of corporates including PwC, Microsoft, Google and a number of other large companies are supporting that initiative. And you also see countries very specifically taking up some of these objectives and promoting the AI for Earth kind of use cases. And there have been some very interesting examples of how AI can be used. For example with all the satellite imagery we can really see the growth of cities or the growth of forests in any particular area. And then based on that take whatever directions to increase the forest cover. For example we as PwC, we looked at that specific thing in India and specifically in Mumbai we looked at how the construction areas within the greater Mumbai is expanding and then encroaching upon all of the other rural areas. And essentially looking at how slums are growing. So the inequality all of that can be essentially got from satellite images. People are looking at the night time image from the satellites and saying how has it changed in the past 10 years, 20 years to sort of plot the economic development of those various countries. So there are a number of ways in which we can use AI satellite imagery and drones for example, at the imagery that's coming from drones for very much the common good, the beneficial good. And people are using various AI algorithms to catch poachers for example in Africa. So some very interesting applications of AI all targeted towards these kinds of issues. I'm quite hopeful that with all of these things we will at least find a way to again may not be equally distributed but at least find more equitable distribution of the benefits of AI rather than just having it being in one or two companies or individuals.

Paramita: I would have loved to go on with this conversation but we have more questions to go through. One more from Luis Salerno... he asks: isn't the rise of AI and the promise of automating almost everything somehow linked to the generalised idea that business translates into growth, endless growth no matter what is at stake?

Anand: Yes this is a very interesting premise. So that we will start automating more and more of the work. And what is left for humans to do is one of those questions. And if you actually look at the evolution of mankind humans over the past what about 2000 - 3000 - 4000 number of years you'll see that initially it was almost 100 percent of the population was required to find food. So either hunt or use agriculture. And then of course when the Industrial Revolution came in that percentage decreased rapidly and more than 60-70 percent of it was around producing goods and only 25-30 percent of the population was needed to produce agriculture. Now that percentage has come down even further. I believe it's around 7 or 8 percent of the population of the globe is sufficient to feed everyone. The problem is more distribution of that food as opposed to manufacturing or generating that food. Now most of this is into the manufacturing and more into the service industry now as more and more service job gets automated people are wondering what will be next.

And to me I think we'll always find things that we want to do and what we are seeing emerge is people to people business. So where people are interacting with other people that is something that would be on the rise for example entertainment. So people are spending lots of money in going to concerts, movies and so that is in entertainment business. Same thing with games. So yes you might for argument's sake you might want to watch a robot team robo-soccer between two teams that that will soon wane you'd still want to watch people. So one team play against the other. And again you see the salary increases of people in the games or the entertainment industry have actually shot up quite a lot over the past 20-30 years since we have had the media. So in that sense people are now looking at taking extreme sports for example which we never imagined that that we could do in the past. So we will pick up things which are uniquely human and leave more of the mundane things for the machines. So I don't think we would get to a stage where people will be wondering what do I do now. If more and more of those things are provided for you then we'll be using our mental and physical power to do things which are interesting and which we find enlightening and happy, make us happy. So I think in that sense we'll continue to do the things that we do.

And there's another notion around is the endless growth. One of the private equity leaders in the US mentioned this, which I think, is very interesting. He basically made the point that progress is deflationary. In other words if the technology is always progressing if it progresses then it is essentially going to bring the cost down. So there'll be a time when we are essentially... we can't be going for growth growth growth all the time because we would pretty much have everything that we need just on this planet at least. So one can argue if you're interested in space that that's probably the time but there's nothing much to "do here" to get that growth. So you'll go into the moon and Mars and explore and then I think it becomes more endless in terms of the capacity. Whereas on earth we might reach a point where we have got everything that we want. Again "Abundance" is another book by one of the singularity people who have written this and which very much goes around the cost curve coming down for a whole host of different things and which we have seen even within our own lifetime. The notion of bandwidth so communicating from a person across the world. I still remember in 2000 and it was a dollar a minute to talk to India and everyone will watch the time for three minutes, five minutes, 10 minutes no more. Now it is pretty much fifteen dollars. You can keep talking forever if you want.

Paramita: There's Facebook!

Anand: Yeah there is Facebook right. So it is always connected so. And if you are into WhatsApp then it is literally free to essentially stay connected. Again that's just one aspect but more of that information based and bandwidth based kind of things are pretty much almost free.

And that's the kind of society we might be getting into. So it's very hard to look for growth in those kinds of industries. So the government might own some of those or large corporations might own some of those and it becomes part of the common good.

Paramita: So according to you think that that we would probably be doing things that we like doing more and more if we "don't have anything to do".

Anand: The majority of us will be doing that. I'm not sure if some of us would be bored but I don't mind being bored for a while given all the work that we do...

Paramita: The next question is. Well it's actually related to all this because you mentioned singularity. So one question comes from Rafael Junger. He asks: how far are we from singularity and does that notion even make sense?

Anand: It's interesting. So it depends on your definition of technology singularity. So the way Kurzweil and others define is that the humans merge with the machines and again left open as to how that happens. To me I think we'll constantly be interacting with technology. I don't believe that there is going to be a super intelligence anytime soon. So the whole concept around an AI that is out there which learns fast and learns how to learn and therefore within a short period of time before we as humans realise what it is it becomes super intelligent and I don't believe that is likely to happen at all. Definitely not any time in the near future because there is no theory around it, there's no one really working on it. So the technology doesn't exist for us to really build super intelligence. It's still a far way out.

But would we be using technology more integrated with us than what it is today? I believe definitely yes. So now everyone has a smartphone and I think the knowledge we have... the knowledge not only the wisdom that we have is all there in the smartphone... so far more than... Even 30 years back we never had anything which which was that powerful. So in that sense we are using the technology pretty much all the time. Now we still need that device and obviously it runs out of battery power every every few hours. But now we'll probably overcome some of those restrictions and we might be able to maybe communicate with other devices in other different ways. I think those things would be possible. So again reading and writing into memory is something that there is a lot of research in the bio world happening. It's not so much AI to me. It's much more into neuroscience and how people are working more deeper into how does the memory get recorded in the brain and how can we read the memory. So once we understand reading and writing of memory then I think we could start doing more things using the brain to connect to other devices. I know Elon Musk and Neuralink are working on some of those technologies. I don't think it's the next five year or 10 year kind of thing. But I believe that at some point it will come true. We should be able to essentially link into larger sources of data and then be able to do. So people often ask what does it mean then to work. So what value can we add as humans. I just look at in the 80s when I did my PhD thesis the only data or books that were available was what was there in the library pretty much on one shelf and a set of books and then that's what you used. So it was easier for me to look at those and write up something comparing contrasting. But now I have the entire body of human knowledge from start till now on Google. Now I need to somehow synthesise all of that and that's far more challenging than just when I had one shelf of books that I can sort of flip through and understand. So I think the task would become even more challenging if we can access every part of knowledge whether it's physics, chemistry, biology. But how do I actually put it together to make something meaningful becomes even more challenging in my view. So I think in that sense we'll be moving more and more towards value added than to basically lose anything that we won't be able to do.

Paramita: Just one last question, as an AI expert do you have anything that concerns you relating to AI?

Anand: Yes as an AI expert, the one thing that concerns me...

Paramita: Concern in the sense worries you...

Anand: Worries me... So my worry is sort of related to the notion of the hype in addition to the sort of the hype of AI it's also this view that AI can be very easily democratised and anyone can become an AI person or any AI expert. That's also a very troubling thing. Yes the use of AI I think needs to be democratised. But just because you are using a few tools doesn't necessarily make you a data scientist or an AI expert. And that's something that is sort of very worrying to me where people believe just like any other subject you can read a few things and then now you can become an expert. I think there is a difference because it is you need to really be building the code, understand the theory behind some of these to really understand the limitations and it's not like reading even a dozen or two dozen books on customer experience, talk to customers and then start doing customer experience. It's not the same. I think we need to give more importance to people who have actually done this as a profession over 5-10 years as opposed to someone who does a few courses and suddenly thinks that they are data scientists. So there is a danger there if you're not very careful we would have a sort of a reputation being questioned in AI while we might have the experts. So some of the non-experts might be viewed as experts or they might declare themselves as experts and be exposed to at least to some of our clients situations. That's definitely a worry of mine.

We need to be very careful on this whole citizens-led and democratisation of AI that we don't alienate the true AI people, the true experts and we sort of look at what they are doing as well and embrace them as opposed to just going this whole citizen and democratisation way of doing things.

Paramita: It was my pleasure to have one true expert here with me. Thank you so much. It was my privilege. Thank you so much.

Anand: Thank you. Yeah. Thank you very much. Thanks for having me.

Paramita: So that was the end of my conversation with Anand Rao. I absolutely enjoyed recording this particular episode and hope that you enjoyed listening to it. I’ll see you next week on a brand new episode of TechTalk. Until then, have an excellent week.