November 23, 2015 | Written by: Andrew Trice
Share this post:
You’ve probably heard something coming from IBM about cognitive computing, and/or IBM Watson. If you haven’t gotten your hands dirty with Watson Developer Cloud services or Watson Analytics, then there’s a really good chance that the term “cognitive computing” might be a little hazy, or seem like marketing-speak.
In reality, cognitive systems have the potential to reshape computing, and are an area of computing that you should not ignore. I recently watched this video The Future of Cognitive Computing by Dr. John Kelly, Senior Vice President at IBM Research and Solutions Portfolio. Not only did I find the subject of cognitive computing completely fascinating, but also highly informative. I strongly recommend that you watch this video to learn about cognitive systems from a very high level, and realize the magnitude of impact that these types of systems are going to have on the face of computing and how we interact with systems and data at large.
In fact, I liked this video so much, and feel that there is so much value in it, that I put together this written transcript version in case you don’t have half an hour to watch the video.
Let me begin with a word of caution: It’s very easy when we get into areas of Cognitive Computing and Artificial intelligence to rapidly drop down to Deep machine learning and algorithms. These are really exciting and fascinating areas of technology, but I think the thing we must keep in mind is “Towards what end?”
What is it that we are actually trying to do?
It’s all about the outcomes:
- Changing the world
- Changing entire industries
- Seeing things and getting insights that we have never been able to grasp before.
I encourage you to keep thinking about what we can do with this technology. How can we impact society and the human state in ways that we’ve never been able to before?
Someone recently asked me “What is Watson worth?” What is the market value of Watson?
I like to put it in the flip context: What is the price of not knowing?
What is the price of not curing cancer?
What is the price of not discovering alternate energy?
All I know is it’s in the billions and trillions. Healthcare, for example, is a 7 or 8 trillion dollar industry worldwide, 3.5 trillion in the USA, and 30% – 40% of that is a waste. There is a huge opportunity to apply these new technologies.
What are we after here?
It’s also very tempting to talk about what we’re trying to do as replicating what human brain does, but that is not at all what this is about. We were not trying to do what previous AI researchers were trying to do. We were not trying to mimic the human brain.
We were trying to do something very simple.
Consider these statistics:
- 2.5 Quintillion bytes of data created every day. (That’s 1,000,000,000,000,000,000,000,000,000,000 bytes)
- 90% of the data in the world today has been created in the last two years alone.
- Every minute 1.7 MB of data is created for every person on the planet. All 7.3 billion of us.
We were trying to build a system that can deal with this massive amount of data because human intelligence is not scaling in the way that data is scaling.
Cognitive computing is not trying to replicate what the human brain does. Cognitive computing is a system that can handle massive amounts of unstructured data.
There is a new Moore’s Law in the data space:
Unstructured data – “dark data” – accounts for 80% of all data generated today.
Most of that data is “dark” – we cannot make sense of that data. It is noisy or formats that cannot be read by traditional systems. Furthermore, the amount of dark data is expected to grow to over 93% by 2020.
Astrophysicists already know about “dark matter”. Dark matter cannot be seen, but we know it exists by the impact on gravity. The same thing is happening with our data. Think of the number of solutions that data holds. This is not a journey to reproduce what the human mind can do.
The objective is to analyze and garner insight from this massive amount of data.
Oil & Gas
Modern facilities have more than 80,000 sensors in place, and a single reservoir will produce more than 15 petabytes of data in its lifetime.
Cognitive computing can:
- help companies prevent drilling in the wrong place.
- help with flow optimization (pumping too much or too little).
Facilities generate more data than current technology can deal with. This is a huge opportunity.
Consumers post 500 million tweets and 55 million Facebook updates each day.
From our partnership with Twitter, we now have direct access to tap into this “big hose” of data. This data can help identify buying patterns, preferences, insights, and where society is moving – Insights that can be moved across every form of commerce.
Internet of Things
IoT is one of the next great frontiers. Watson is unmatched in natural language processing. Watson also can handle images and vision. Think about signal processing.
Machine-to-machine data will dominate the market in just a few years. That is noisy, unstructured data, perfect for cognitive computing.
Smart, connected appliances will grow from less than 1% of the market today to more than half in 2020.
This could be appliances, or could be smart connected cities. Think of the implications of city security or traffic management.
New York City surveillance cameras and sensors generate 520 TB of data per day, largely unstructured, and untapped.
In 2014, more than 1 billion personal data records were compromised by cyber attacks.
Security is no longer about firewalls. Security is now about behavioral analysis of people and systems. Systems can predict abnormalities and react in real time.
More than 680 million smart meters will be installed gloabally by 2017 – producing more than 280 PB of new data to be analyzed and acted upon.
Digital meters are in many countries around the world, but the data is dark. It is difficult to integrate renewables without understanding demands.
One of the biggest opportunities, of course, is healthcare. Healthcare is an enormous industry, prime for disruption and new forms of insight. This is one of the industries where we (IBM) have doubled down, not only through electronic medical records, patient population healthcare, but also medical imaging.
Each person will generate 1 million GB of health-related data in their lifetime – equivalent to about 300 million books.
In that data is really the secret to our own health and well being.
By 2020, 75% of the world’s cars will be connected… and they will produce 350 MB of data per second to be assessed and acted upon.
These devices will need to be cognitive. They will need to make realtime decisions about the environment based on learning about the environment and learning about driver behavior.
Every single industry is being swamped with data. Every industry is trying to access that 80-90% of dark data and get insights to differentiate. We are at an industry inflection point.
The dawn of the Cognitive Era
The 1900-1940s were the Tabulating Systems Era – mechanical tabulating machines programmed via punch cards and switches enabled automated arithetic.
The 1940s through today have been then Programmable Systems Era – Computer systems have enough memory to give the system instructions, and let the system run itself with no external programming.
At IBM research, we’ve noticed that we are going to run out of programmers.
It will not be possible to keep up with all the data. So we had better take another path.
We believe that we are now at a major point of industrial inflection that only happens every ~40-50 years.
We are creating entirely new computer systems that do entirely different things than the last era. This era will be more different from the programming era than programming was from the tabulating era.
Think about what happened with that infamous System 360, a platform that changed banking, and airlines. A platform that became the backbone of enterprise transactions. A platform that will transform industries; that is what we are trying to achieve with our cognitive systems and Watson.
It all began almost 5 years ago with the infamous Jeopardy match in February 2011. Our goal was not to win a game show. We wanted to demonstrate that we are going through a transition from programmable to cognitive.
I talked to Ken and Brad and asked them two questions:
Q: How do you do what you do?
A: Everything I see and hear, I never forget.
Q: When are asked a question, what reasoning process do you do?
A: I don’t know, I just have an answer immediately in my head.
They have complete memory and instant recall of everything they see. To beat these two humans, Watson needed to be right 85% of the time, within 2.5 seconds. That is a really tough problem. The reason we won is because we took an entirely different approach based on machine learning and very sophisticated natural language processing.
On the positive side, this set the stage for cognitive computing, but it also set up man vs. machine. That was not the intent at all. Every study has shown it will be either man or machine. That is a key point because of the differing capabilities that we each have.
Not man vs. machine, but man & machine
We as humans have a number of abilities that machines will have a hard time replicating; maybe never.
- Value judgement
- Common sense
- Total/Instant Recall
- Deep learning
- Large-scale math
- Fact checking
The opportunity is man and machine. We are seeing this in every discipline and every industry that we go into with Watson.
We as humans have a normal distribution (statistics) of skills. What we’re finding is that we can move that distribution. We can take the best experts and make them better by introducing man and machine.
The question is: How do we get the synergy with man and machine?
Since 2011, this field has exploded.
- Many image processing tools
- Many buying optimization tools
- Many voice recognition tools
Each of these is really a point solution to improve a one-dimensional aspect of a business model. It really is like a tool – hammer, screwdriver. Very few, besides IBM are trying to build an entire toolkit/platform of capability for all industries with this cognitive computing capability.
We took Watson, which was one system, and brought it to IBM Cloud. We decomposed one system (Q&A) to individual services that can be composed to create meaningful solutions.
This system effectively had five parts at the time:
- Machine Learning
- Question Analysis
- Natural Language Processing
- Feature Engineering
- Ontology Analysis
We (IBM) have built out a suite of services that enable you to
make your own mini-Watson for a solution to your problem. We decided to make it cloud based and composable for all industries.
This has become the platform for our ecosystem. Our plan is to develop not just dozens, but hundreds of these services on the Watson cloud as fast as we can. We have a pipeline of these services in a rich environment.
What is the essence of cognitive?
Cognitive systems must learn at scale.
- Learning at scale in the data
- Learning at scale for your solution/business
- Reasoning with a goal to take an action
Cognitive systems interact with humans – it is the interaction of man and machine that will produce capabilities going forward; Not just about automating systems
Rethinking What’s Possible
The possibilities are immense.
Healthcare: We at IBM believe that image analytics and machine learning can change the course of health care. Two-thirds of medical information is contained in images (X-ray, MRI, CAT scan). We know the diagnosis is not what it needs to be. Radiologists look at thousands of images in a single day – human issues such as fatigue set in. We are going to train Watson to read those images with the patient data. Watson will use Unstructured medical record data across the entire patient’s history to generate recommendations in minutes. This will change the course of healthcare.
Seismology: They are building cognitive environments for decision making.
- Where do I drill the next well?
- Do I bid on that land to get the oil under it?
- What happens to oil reserves over time?
We are going to help transform a very intensive data industry.
Education & Accelerated Learning: There is a direct analogy between health care and education. Medical is treating to the average. Education is very much the same. Teachers teach to the average.
Think about having Watson engage with the individual student to observe learning patterns and intervene. The engagement has a direct correlation to accelerate learning and change education systems.
Pre-K (ages between 2-3) vocabulary is a direct correlation to long-term potential. Watson has the capability to double or triple the vocabulary of that 2-3 year old. There is a huge opportunity to change education.
Genomics: Eventually genomic data will likely swamp image data. Genomics will generate more data than can be understood by doctors (humans alone) – Hundreds of mutations, thousands of pathways that cause a tumor can manifest itself. Think about using a Watson to help humans understand what those mutations mean, and what is the right treatment. IBM is working with a number of leading edge genomic institutions in the US and Canada to explore what Watson can do in this area. Frankly, it is the only way we’re going to deal with genomic data.
Let me end by reflecting on some words from Thomas Watson Jr, where he describes systems that take away mental menial tasks and free us up for creativity and other things:
Computing will never rob man of his initiative or replace the need for creative thinking. By freeing man from the more menial or repetitive forms of thinking, computers will actually increase the opportunities for the full use of human reason.Thomas J. Watson, Jr
This is what we did with programmable systems. We created processes then went off and made new things.
With Watson it is no longer about displacing work. Machines displace manual labor. Programmable systems displace menial mental processes. We’re now talking about man & machine tacking problems that were inconceivable just a few years ago
I urge you to enjoy the discussion. Don’t fall into the trap that we’re trying to recreate the human brain, and don’t fall into the deep technology trap. Think about the implications of the technology. There is not a single industry that won’t be completely transformed by this technology over the next decade.