George Lee, chief information officer for the Investment Banking Division at Goldman Sachs, discusses the disruptive potential of artificial intelligence.
This podcast was recorded on March 4, 2015.
The information contained in this recording was obtained from publicly available sources and has not been independently verified by Goldman Sachs. Neither Goldman Sachs nor any of its affiliates makes any representation or warranty as to the accuracy or completeness of the information contained in this recording and any liability as a result of this recording is expressly disclaimed. This recording should not be relied upon to evaluate any potential transaction. Goldman Sachs is not giving investment advice by means of this recording, and this recording does not establish a client relationship with Goldman Sachs.
Copyright 2015 Goldman Sachs. All rights reserved.
This is an unofficial transcript meant for reference. Accuracy is not guaranteed.
This is excellent.
is it Goldman Sachs? Were people from our firm share their insights on developments currently shaping markets industries in the global economy? I'm Jake Stewart Obeid of corporate communications here at the firm across industries technology is transforming the way virtually every company does business while it's hard to keep up with the breakneck pace of innovation. Today's guest George Lee comes as close as anyone to be able to do that. George of the co chairman of the global technology, media and telecom group of Goldman Sachs and the chief information officer for investment banking division he's here today to discuss artificial intelligence. One of the.
significant developments in technology. Today, George welcome to the programme thanks Jack Great, be here, George you're, immersed in these issues on a daily basis, with some that leading acknowledges in the world bullets that step back for a second,
our listeners might not be as familiar with an artificial intelligence. How do you define it? How do you explain to the lay person what artificial intelligence means to the technologists
You know the way I think about. It is really a rise in attempts to stimulate human level, cognition recognition and even reasoning at some level
in computing and without rhythms and new things that some sort of fantastical, but that the truth is we experience at every day, in certainly on
in technology, not mine, things like you, know, advanced advanced, online search
placement of online advertisements, recommendation
engines in e commerce. All those things are fundamentally artificial, intelligence applications, the ability of fish
send for me when I'm looking for something on a search engine, exactly that's a great example, because it leverage
the fact that it seen so many searches that start out with the beginning of the string you entered that it can pattern match from that huge data set and supply you an answer that has some level of intuitively ass to its interest, the interplay, because sometimes you begin to types of the
starts guessing for you think we'll gush I've been meaning to look for that, but maybe that hinder us you're you're, absolutely right. They see that they see that phenomena by the way. This is one of the paradoxes of artificial intelligence today, which is you know something that originally defined as a very ambitious artificial intelligence from machine learning application once it becomes commonplace, stopped being a high and starts just being regarded as a feature. So we definitely see that.
with think about the historical evolution of artificial intelligence, not like it's a new term machine running those phrases been around for while we ve had some false brings on the eight hour front, word seem likely. Technology was just about to break through and and then it had a wall where we now this different. Yes, it s by someone said those are the most dangerous terms and in technology and financial,
this time it's different, that having been said it does feel like we are in a fundamentally different place in your right. The bank, the term artificial intelligence goes back some sixty years and there have been lots of. Sir
judges of productivity in this area of study and then a set of disappointments. Today, the ability to have the processing power and speed that's afforded by the cloud and new technologies and the ability to access and create huge amounts. It
of because of mobility and big data, those two things are fundamentally new assets for the evolution of ay. I that I think, create the promise of vaulting it to a new level. So, let's talk about the connection a little bit between big data and artificial intelligence machine learning. What does that mean for the the shape of innovation in the sector? I would say: big data is a natural complement to the development of artificial intelligence today, because in a sense, big data demands artificial intelligence, the size and scale of the data sets that are getting created, really defy any human ability to comprehend such vastness, and so many
friend, data points and pieces of information, and so the ability to harness technology to crawl across those data sets and try to draw out meaning in insight is
really really important, and so there very mutually dependent phenomenon, which is to say a
thrives in scenarios where there's big data and the availability
big Data allows you to train computers by giving them a lot of information that allows them to be more intelligent and drive more insight. George start ups have the ability to use this technology, but also established companies how are startups looking at disrupting current business models, current ways of doing business and then how're established companies thinking about incorporating artificial intelligence and big data into ensuring that their one
at the head of the storm in those trying to disrupt their industries. This is not just purely a silicon valley phenomenon by any means, so some of the areas-
fastening around this or health care diagnostics? So, if you think about a
capable artificial intelligence set of algorithms,
you, can create a machine that has almost infinite knowledge with respect to medical matters, case histories, academic papers, etc and allow it to try to pattern match what detects
in the person being on the phenomenon being diagnosed against that infinite knowledge, and while that
May or may not be a better than humans is certainly a great compliment to human intuition in coming up with possible diagnoses, fur
care issues, and so that is an area. We see a lot of innovation so places where you could reduce, basically, the risk of medical error to which has been a huge problem for the industry. In a way better
about what a doctor does he scanning his memory banks to try to match the symptoms there being detected to a set of possible outcomes, are possible disease states that match those symptoms and he's taking advantage of all his experience in what he reads and has seen overtime
a lot of that functionality actually can be provided by the computer and then delivered to that doctor can help shape and and leverage that asset
in another area that much talked about are self driving cars or assisted driving scenarios which,
really are very sophisticated machine learning and artificial intelligence application. You think about it: the scanners on the top.
self driving cars are gathering enormous amounts of data in real time, their matching that data against eight collective knowledge of what could possibly occur and making small predictions. It's also a great example of how artificial intelligence and machine learning benefit from Q
What are the facts and network effects in the sense of every mile or our that a self driving car drives it get smarter. It has more experiences, it observes more situations. It gathers more data, moreover, that data can be shared across a fleet of different cars, and so the experience of every car, the observations of there
a car can gather and cumulatively and mutually effect at any given car in that network, the power that could be quite substantial tickets,
like Goldman Sachs, had had a. We use artificial intelligence in the way we conduct business today, and how might we be using it
in the months and years to come, will it its first water. It is a field of technology where there's we have a lot of expertise in the firm there's, an enormous amount of work and study in development, going in two ways that we can harness machine learning, an artificial intelligence
Serve our clients better and more efficiently and one area where we certainly see lots of applications for this is, if you think about the day, to day at a Goldman Sachs Portugal in the operational domains, we collect absolutely vast amounts of data about securities transactions about you, operational exchanges with all of our clients. Those data sets are so vast that it's very hard for any human or any group of humans to crawl across,
and trotted detect areas where we can do better, we can serve Clive, better and more efficiently and with
I'll go with them and a vast data set. We can have machines crawl across that data set and come back to us and propose operational improvements that can allow us to work better with our clients, and so that's a huge area of emphasis for us today towards them in some great fascinating debates about artificial intelligence and its impact on society, and these are our debates between what I'd and technologists. These are debates between people who love technology whom innovate in the field, some of the smartest people about technology today, and yet they have very different emotional reactions and very different points of view about what the future holds talk a little bit about those debates and the extent to which each side is making a fair point. It is an important debate and one that's fashion, because some of the keenest observers and smartest people in technology coming out very clearly most in a polar sense on either side of this debate. The two fundamental schools are one the
extraordinary capability and promise of artificial intelligence and the fact that it can grow smarter and more capable every day and there's a sort of moors law. Phenomenon of exponential improvement of artificial intelligence is something that, if we wait to try to figure
how to build a guardrails later on it's going to be too late. You know, Elon Musk, I think, has been probably the most prominent observer of that and I think he's very passionate about the idea that we have to study and think about writing the rules of the road. If you will come sooner rather than later, and by the way you can, you can spin up some pretty scary scenarios when you extend this vast capability without a sense of ethics and reasonableness, Turnitin machines, one of the one of the interesting examples that people use is a scenario called the paperclip Maximizer, which is to say
if you had a computer with infinite capabilities, and you programmed it with a singular objective which is to maximize the production of paper clips. Imagine what it would do without the boundaries that come with human ethics, so it would immediately construct a snare. We drove all of its competitors out of business. It might
you're late markets to corner all the resources of the EU to make paperclip. It might try to protect against its own potential mortality by creating a copy of itself and disturbing it around the network. And you take all of these things to their extreme, and you can imagine a pit something as simple as a machine designed to maximize projection of paper. Clips could have
huge knock on effects in financial markets on the other side of the ledger, people which you know prominently include mark and recent as it is both for this perspective, which is that is a fair consideration. But that point in time is so far distant that we should concern ourselves or inhibit. The innovation in artificial machine learning today gum up the works. If you will, because the time in which there's any real threat on this basis is not ten years or
twenty years away, it's fifty years or beyond, and so therefore too much hand wringing over
Regulation today will just be an inhibiting innovation which and that innovation can be so leveraging so helpful, so important to the economy, the creation of jobs. Finally, instead of the destruction of jobs that we just should encourage it. When I hear you talk, George, I sometimes have a hard time to terminate what is artificial about the intelligence you described. So how do we distinguish between artificial intelligence, which is trying to mimic the human brain and intelligence itself? It's a great question, and you know it can lead to a debate on what is the nature of human intelligence and by that, by the way that debate
Frightened centre in some of the different approaches to artificial intelligence such I can talk about, but you don't some ways. It's you know is the cumulative of information knowledge or is there more to it and in one of the thought, Exe
payments on this attention. What somebody really promise
exciting new artificial intelligence strategies, harness the vastness of processing and processing speed it's available in the clouds.
And in order to make a decision, they will try a bunch of different approaches to a problem by the way. Solving a video game, for instance, is it as
ass area and they'll, try literally millions of scenarios and then slowly but surely or quickly, but surely eliminate failed strategies to determine the optimal path? And so the question is: is that intelligence, or is that just brute force,
and at some level does our brain work that way or doesn't it, which again leads to this big debate, an artificial intelligence which will artificial intelligence as it becomes to them.
in the next ten or twenty or thirty years. Will it really mimic the structure and function of
human brain or will be something different and entire
the innovative and new, and so there are a bunch of accurate
x and researchers and developers who believe that ultimately, the way to create artificial intelligence is by mimicking and improving the way that the human brain works at serve the neural networking approach. There are other research
who believe that machines will never be able to replicate that unique approached? I want to focus on what machines do well and find other innovative and wholly new approaches to problem solving some really interesting debate:
a little bit of what the future holds. We ve seen a lot happen in a very short period of time, but what can we expect to see over the next couple of years in space, it's hard to project much further than that, but for the next couple years, one of the trends that we might be seeing across
a range of industries. Night again, as I said in know, it's funny this artificial intelligence machine learning, while they seem in a very futuristic, are things that we experience every
in the way the we we interact with online resources in particular, and so one answer is artificial intelligence. Machine Lammle just continue
or progressively imbue all the activities that we,
do online in Kevin Kelly, who is an editor
Wired wrote a peace where we basically said what the future holds
a machine learning is that everything that we formerly electrified will now be what he called cognate ties facing me
smarter by the application of a little bit of ay. I and machine learning. He also observe
which, by the way, I can certainly agree with. Given my my position in Silicon Valley, is he even said you can't near the next ten thousand start ups in the world can be defined as take acts, some activity and add a eye to it, and so the
rise of big data. The rise of smart algorithms, a lot of the activities in Silicon Valley in the entrepreneurial start up community around taking every day, experiences applying more data to those and then trying to develop algorithms tat, just make them better, more efficient and more satisfying experiences for consumers and businesses. I think, for the foreseeable future. It's gonna be a great compliment to
human capability and that it will, it will leverage us and that a lot of what will do will be co. Work with these either machines or or algorithms, are one of the great examples of that
the smartest thing on the planet today is neither man nor machine. It's a combination of the two, so the world's greatest chess masters and the world's greatest chess supercomputers can both be beaten by a very good chess player, with the benefit of a laptop and a smart chess programme. So an interesting way of saying that you know when we work really closely with
genes and with technology we can actually achieve more than is possible, certainly on our own, but certainly today, more than can be achieved with a machine alone. George. Thank you very much for joining us today. A great discussion and left us with a lot to think about thanks Jack. This exchange is common
acts on Jake Seaward. Thank you very much for joining us today. Unless the spotlight
Was recorded on March for two thousand and fifteen, the information
contained in this recording was obtained from publicly available sources and has not been independently verified by Goldman Sachs. Neither Goldman Sachs nor any of its affiliates makes any representation or warranty as to the accuracy or completeness of the information contained
in this recording and any liability as a result of this recording is expressly disclaimed. This recording should not be relied upon to evaluate any potential transaction. Goldman Sachs is not giving investment advice by means of this recording, and this recording does not establish a client relationship with Goldman Sachs.
Transcript generated on 2021-10-15.