1. Lecture Two: One World
Series: Understanding Cognitive Systems
Jim Spohrer (IBM)
Director, Understanding Cognitive Systems
December 7, 2016
http://www.slideshare.net/spohrer/understanding_two_20161207_v1
12/12/2016 Understanding Cognitive Systems 1
2. Today’s Talk: One World
• Lecture One (Recap): Worlds Apart
• Biological and Digital Cognitive System Entities
• Build them:
• BCS interact to learn through experience
• Adult: 10M minutes (gains commonsense reasoning and knowledge)
• Expert: 2M minutes, includes 10K hours of deliberate practice
• DCS programmed
• Today: HW/SW building blocks (requires programmer)
• Tomorrow: Master Algorithm (Domingos), like BCS
• Lecture Two: One World
• DCS beginning to interact to learn
• Environment (simulated, real)
• With other DCS and BCS
• Future: Socio-technical system design
• Fitting DCS into BCS personal and professional lives
• Smart/wise service systems
12/12/2016 Understanding Cognitive Systems 2
Jim Spohrer
Director, Understand Cognitive Systems
IBM Research
(Bio: Maine, MIT, Verbex, Yale, Apple, IBM, …)
This presentation provides an initial conceptual framework for “Understanding Cognitive Systems” – this presentation can be downloaded from slideshare.net/spohrer
Spohrer, J (2016) Understanding Cognitive Systems. A CSIG weekly speaker series presentation, also for Marty Apple. URL: http://www.slideshare.net/spohrer/understanding_20161127_v8
I’m Jim Spohrer, I work at IBM and I am the presenter, and in today’s short talk, I will briefly cover what is a cognitive system (entity) – both biological and digital. Then I will briefly discuss how to build, understand, and work with digital cognitive systems – and how this is steps towards a next generation cognitive curriculum, including types of digital cognitive systems.
Biological cognitive system entities… and human intelligence…
The best explanation of what a biological cognitive system entity is can be found in Terrence Deacon’s book- the Symbolic Species – the co-evolution of language and the brain. All easily recognizable biological cognitive systems from ants to wolves to crows to dolphins to monkeys to people have brains that have co-evolved with symbol systems - chemical, visual, auditory – that individuals of the species you as a type of language for communicating and coordinating reasoning and interactions and the accumulation of knowledge for successful multi-generational living in an environment. Less sophisticated languages and brains deal primarily with the physical world, but more sophisticated languages and brains are needed to deal with the social and in the case of people, the cultural and institutional world, of large numbers of others of their species living in close proximity with each other.
Responds to environment (decision making – model of physical world, tasks), learning (memory – model of self), language (communications – model of other), institutions (collaboration, competition – model of institutional world, trust)
Is the entity the individual, the collective, or the species? Or all of these levels of organization?
Bionic ant design: https://www.youtube.com/watch?v=voNBzuI7IJ4
Technological cognitive system entities, or what we will call digital cognitive system entities – are most easily explained from science fiction examples like Star Trek’s Mr. Data – an individual smart machine or AI (Artificial Intelligence). Artificial Intelligence (AI = MI) versus Intelligence Augmentations (IA = HI + MI) = Human Intelligence + Machine Intelligence
Versus a Tony Stark and Jarvis in Iron Man – which is an example of an Augmented Intelligence that combines biological human intelligence with technological machine intelligence into a system. In Augmented Intelligence – human computer interactions are often aided by augmented reality symbol systems that allow a blending of the physical and virtual or cyber worlds, so that people in the physical world can directly interact with bits of information in the machine intelligence or AI world. So it is likely that over time augmented intelligence and augmented reality interfaces will co-evolve as complementary technologies and system components.
Important property of Augmented Intelligence, must increase HI by use and not diminish it. Same as a parent and a child – the child must become more capable over time, and more dependent
People take 10 million minutes of experience to go from conception to adult in today’s society with rights and responsibilities.
Dogs, and some other biological cognitive systems, make the journey in about 500,000 minutes of experience (half a million minutes of experience).
Building an expert BCS takes 10,000 hours of deliberate practice, or around 2 million minutes of experience.
Building an expert DCS is still very heard work, and the rule-based approach of expert systems can be brittle and the bottom-up data-driven deep learning approaches can take an incredible amount of data.
n the 60 year history of the new field called Artificial Intelligence (AI) the building blocks have been getting better, and now “cognition as a service” is being offered by industry and non-profit players.
As the lower level building blocks get better and easier to use, value migrates to higher and higher levels….
Ultimately, building an expert BCS (person) or DCS (assistant) is not simply about the hardware/software/data/experience levels but about the design and transformation of socio-technical systems.
Where is the variety? Hardware and even software standardizing into modules and algorithms…. Data will standardize next into categories and types…. Experience is where the uniqueness is, and variety and variability, and identity.
Moving from science fiction to what exists today in 2016 – in the marketplace, we can see that more and more companies are developing intelligent assistants and augmented intelligence systems with technological components – in the cloud, on smartphones, in robots, and in cars, or as embedded in existing communication apps or other devices that people use when interacting with others or the world.
Here is what I tell students....
... to try to provoke their thinking about the cognitive era:
(0) 2015 - about 9 months to build a formative Q&A system - 40% accuracy;
- another 1-2 years and a team of 10-20, can get it to 90% accuracy, by reducing the scope ("sorry that question is out of scope")
- today's systems can only answer questions, if the answers are already existing in the text explicitly
- debater is an example of where we would like to get to though in 5 years: https://www.youtube.com/watch?v=7g59PJxbGhY
- more about the ambitions at http://cognitive-science.info
(1) 2025: Watson will be able to rapidly ingest just about any textbooks and produce a Q&A system
- the Q&A system will rival C-grade (average) student performance on questions
(2) 2035 - above, but rivals C-level (average) faculty performance on questions
(3) 2035 - an exascale of compute power costs about $1000
- an exascale is the equivalent compute of one person's brain power (at 20W power)
(4) 2035 - nearly everyone has a cognitive mediator that knows them in many ways better than they know themselves
- memory of all health information, memory of everyone you have ever interacted with, executive assistant, personal coach, process and memory aid, etc.
(5) 2055 - nearly everyone has 100 cognitive assistants that "work for them"
- better management of your cognitive assistant workforce is a course taught at university
In 2015, we are at the beginning of the beginning or the cognitive era...
In 2025, we will be middle of beginning... easy to generate average student level performance on questions in textbook....
In 2035, we will be end of beginning (one brain power equivalent)... easy to generate average faculty level performance on questions in textbook....
http://www.slideshare.net/spohrer/spohrer-ubi-learn-20151103-v2
By 2055, roughly 2x 20 year generations out, the cognitive era will be in full force.
Cellphones will likely become body suits - with burst-mode super-strength and super-safety features:
Suits - body suit cell phones
Cognitive Mediators will read everything for us, and relate the information to us - and what we know and our goals.
Think combined personal coach, executive assistant, personal research team....
The key is knowing which problem to work on next - see this long video for the answer - energy, water, food, wellness - and note especially the wellness suit at the end:
https://www.youtube.com/watch?v=YY7f1t9y9a0&index=10&list=WL
Do not be put off by the beginning of the video - it is a bit over hyped and trivial, to say the leasat... but the projects are really good if you have the patience to watch.
In early 2016, IBM did an experiment to see what types of digital cognitive systems is employees where most interested in building.
The experiment was called the “Cognitive Build” and over 250K employees took part in some way.
I analyzed the top 400 projects, and with the help of ideas about Types of cognitive systems - co-created with the help of Don Norman and Paul Maglio – was able to identify five major types of cognitive systems that differ in the complexity of the models they possess.
Tools are the simplest since the do one thing well
Assistants are slightly more complex since they can help with a range of tasks.
A collaborator is more complex, in that the best collaborators really know they user well.
A coach must know the user even more deeply, not just today’s capabilities, but aspirations for the future.
And finally a trusted mediator can take actions on behalf of the user – to help people interact better to co-create more value in complex interactions that require reasoning about institutional arrangements and laws, social and cultural conventions.
These five types of digital cognitive systems and the types of models and capabilities they require are summarized in this table.
Relate to learning, perception, reasoning, interaction, and knowledge – five parts of an AI course.
So how far are we from the master algorithm? How far are we from having digital cognitive systems learn more like biological cognitive systems, rather than people programming with the rapidly improving building blocks?
To begin to answer these questions, let’s look at the work of Prof. Peter Abbeel (UC Berkeley) – people may note that he has one of the top rated AI/Machine Learning MOOCs in the world fyi…
One of my heroes and mentors – Doug Engelbart (1925-2013)
Doug and I had several conversations about the relationship between augmentation theory and service science. I wish we could have had many more.
Before connecting augmentation theory to service science, I have to travel through some technical areas that are closer to my first two degrees physics at MIT and artificial intelligence at Yale university – but I promise you, I will connect this to service science and smarter service system research agenda….