Manoj Saxena is the executive chairman of CognitiveScale and a founding managing director of The Entrepreneurs’ Fund IV (TEF), a $100m seed fund focused exclusively on the cognitive computing space.
Saxena is also the chairman of Federal Reserve Bank of Dallas, San Antonio branch and Chairman, SparkCognition an Austin based cognitive security and safety analytics company.
Prior to joining TEF, Saxena was general manager, IBM Watson, where his team built the world’s first cognitive systems in healthcare, financial services, and retail. Earlier he founded, built and sold two Austin based software startups.
Saxena will be speaking at Gigaom AI in San Francisco, February 15-16th. In anticipation of that, I caught up with him to ask a few questions about AI and it’s potential impact on the business world.
Byron Reese: How intelligent do you think a computer can become?
Manoj Saxena: I think they can become super intelligent. They already are. In certain areas, they have far exceeded human brain capacity. Now they are super intelligent, they are not super conscious. So I separate intelligence from awareness and consciousness. So I think intelligence is here, has been here for decades. And you know with the advent of cheaper computing power in the cloud and more access to cloud through mobile, I think that intelligence is going to get more and more pervasive and will basically be woven into all aspects of our life. You know, how we work, how we live and how we play is all going to be changed through computer intelligence surrounding us. I actually talked about this notion of as a species Homo Sapiens are dead. Home Digitalis is the future because we will be surrounded by intelligence and amplified and augmented by intelligence.
Do you believe that in AGI, general intelligence is possible to build?
I think it is possible but we are probably at least 40 or 50 years away from it. You know artificial general intelligence which essentially you could argue that you know parts of it. Google, you could argue as the beginning of an AGI kind of like a mega brain or Watson in certain domain is the beginning of an AGI but through AGI covering all forms of human knowledge and human pursuits, I’ve read a study on it that even if you ask the specialist in the AI field, the average response was that we are looking at 2050 or 2060 by the time we will attain AGI. I think the most exciting part is not AGI but ASI, Artificial Specific Intelligence.
Well I think a little bit of AI can go a long way. There were few big revelations when I was running IBM Watson. First, you don’t need to build an AGI to drive humanity forward or to transform businesses. A little bit of AI when applied to targeted consumer engagement or industry specific business processes can have exponentially huge impact. The second insight I had when I was running Watson is, the real interesting part about AI and machine intelligence is not asking the question of a machine, but it’s the machine telling you what question to ask. You know there is three types of information in this world: there is stuff you know, there is stuff you know you don’t know, and there is stuff you don’t know that you don’t know. The real interesting part of machine intelligence is the third bucket where the machine taps you on the shoulder and says, hey you got to check this out.
I don’t want to get bogged down in definitions or anything, but can you please explain the distinction between machine intelligence, artificial intelligence and cognitive computing?
Yeah. So artificial intelligence is sort of the uber category. Artificial intelligence is like saying ‘software’. It’s the broadest definition which includes multiple types of technologies and techniques: machine learning is one, computer vision is another one, and cognitive computing is yet another one. There are many other types of AI. So at a top level, AI is the super category and then within that, machine intelligence is application of AI where machines start learning and start getting smarter on their own. So it could be a thermostat, it could be a traffic light or it could be a mobile app. Any of these can get smarter. Cognitive computing is that specific part of AI that relates to mimicking the human brain in terms of how we understand, reason, decide, and learn as a human being.
And recently, you know Stephen Hawking has mentioned that AGI may be an existential threat. Elon Musk says things like, maybe we are just a boot loader for the machine intelligence and that’s the next step in evolution. Bill Gates is concerned about what a general intelligence could do. I would ask two questions. One, why do you think that so many obviously very smart people are worried about it and second, do you share that worry?
I think there is some truth to that worry that I share. But I also think there are other scenarios that are in my opinion overhyped and overinflated in terms of machines as the new digital overlords.
My car’s system is kind of acting up here. This is a good example of why I am not too worried about machines being our own overlords because you can’t even get the damn phone to work in your car or your autocorrect to work on your cell phone as someone said. Having said that, today we already are at a point where machines are running our lives. There are millions of us that entrust our lives to computers today by allowing a computer to land our plane and seem very comfortable doing so. And that will slowly expand that we could only get more prevalent as we start giving more and more trust to machines. Robotic surgeries of eyes or blood vessels are other good examples.
On the other hand, there is a lot that we don’t know about how the human brain works and it will be hard to replicate that in a machine. There is much to be learned around our own consciousness, compassion and instincts work for example so in that sense we are very far away from the worry of a new digital race of computers.
What is needed for sure are some general principles and governance by which we as a race put this powerful technology to work for the betterment of society. I am currently engaged in some discussions with industry and local leaders around AI ethics and moral responsibilities to prevent both real and perceived threats from an AI apocalypse.
So what are you trying to do with CognitiveScale?
What we focus CognitiveScale on is deep practical applications of machine learning in industry. So what CognitiveScale builds is the notion of industry digital brains. They have taken AI and applied it into three verticals in commerce, in healthcare and in wealth management. We call it health, wealth and commerce. We are using AI to transform how patients manage chronic conditions and chronic diseases. We are using AI to manage how shoppers are experiencing their journey with the retailer and how financial advisors and investment advice is being delivered to end users. So we focus on transforming the experience of a user through a mobile phone or a browser that creates an experience like that of the traffic and map application Waze.
Waze is a good example of an existing cognitive app. You know it’s an app that is able to source a lot of data both structured data and unstructured data and it’s an app that guides you through the journey and optimizes your experience and outcomes. It knows you, it knows what’s around you and it gets you to your destination in the most efficient fashion. So what CognitiveScale is doing is they are building products for health, wealth and commerce that create a Waze-like experience that lets a patient manage their diabetes or their cancer or their obesity by guiding them through their journey. It lets a shopper manage the journey of shopping for an event and it helps a financial client manage the journey of investment advice because we believe that patient shoppers and financial clients, they all go through a journey and these applications take a regular mobile app and they put a little digital brain behind it and those applications start acting like Waze.
And where are you in your product lifecycle?
CognitiveScale has launched two products. One is called Engage for the customers, the other is called Amplify for business processes. So Engage transforms how a customer experiences the company and Amplify makes every employee your smartest employee. They are about a 100 people and they have been in existence for about 3 years. They have 20 customers and global brands you know everything from Barclays to Nestle to Macy’s to Dow, Eli Lilly, MD Anderson. So they’ve got a tremendous technology and client proof points and have a very strong deal pipeline. Off to a good start but much more needs to be done.
Do you believe that computers will become conscious?
Well yes and no. So, yes but it depends on how you define consciousness. So with consciousness, there are two problems. One is, there is no consciousness detector today. So we don’t really have a model that says, okay what’s the level of consciousness in a particular human being. Now there are some models that are based on anatomy like doctors use to know if you are comatose or not or behavioral people use to figure out whether you are mentally capable or not but there is no proper sort of a continuous consciousness detector that we can use to measure a person’s or a computer’s consciousness or lack of it. So one is a problem of measurement.
And second is a problem of applicability because only, I think, 5% of the human brain is used around consciousness or consciousness-related activities and it may very well be that consciousness may be outmoded and may be outdated by computers that get super intelligent and are able to do tasks much more efficiently and maybe the relevance of what consciousness is needed for is a lot more limiting.
For example, how does it matter for a computer or when does it matter if the computer can sense the pain of a young wife who lost her husband, or if the computer can sense the joy and laughter of a child on a beach, or why someone would throw themselves in front of a running train to save a baby, right? So these are the kinds of things while they are important, they may not be as relevant in the grander scheme of things for the progress of humanity with machine intelligence. So those are the two issues. Therefore, in part yes, you could say the computers will get self-aware but I think unless we have a proper consciousness detector, it will be very hard to formally answer the questions if computers can become conscious.
What is your take on the Chinese room problem, which argues that computers can’t really ever be truly intelligent? [Note, this is a classic argument against the possibility of a general AI put forth by the philosopher John Searle. It is worth looking up in Wikipedia. But the basic idea is that because a computer is completely mechanistic, it simply follows programming. No matter how clever it looks, it doesn’t really understand anything.]
I think there is a lot of truth to that statement as long as you assume that computers are being built on the Von Neumann architecture. Under the present architecture the Chinese room problem that you are talking about is true. You can say the computer is only parsing things together. However, if you look at evolution of quantum computers, [it might be different.] When you ask a current computer what is 1 plus 1, it will get you 2. When you ask a quantum computer what is 1 plus 1, it will take all numbers on the right and all numbers on the left and it will give you all kinds of answers. So it won’t just add up 1 plus 1, it will add up 1 plus 5, five million plus 3 and on both sides and then it picks a particular quantum event. So a lot of theories [suggest] that the human mind is a quantum machine, and that the reason we make connections across things which may not have any logic to it. There is a big stream of expertise and thinking that believes that the human mind operates not like a traditional computer but more like a quantum computer. So if you took that approach then I think the answer could be, ‘yes AGI is possible.’
Great. We’ll leave it there. Thank you for taking the time to talk today.
Manoj Saxena will be speaking at Gigaom AI in San Francisco, February 15-16th.