The State of the Art in Demon Summoning, a.k.a., Artificial Intelligence.
Jana's relationship with AI is a 25-year journey from using genetic algorithms to predict the maximum theoretical conductivity of certain polymers, to seeing if neural nets could help distinguish Paula Abdul's voice from that of a backup singer's, to expert systems that at this moment decide which driver is behind the wheel of any of tens of thousands of semi-tractor trailers. Today, the most frequent AI question she gets, is whether Elon Musk is right: are we summoning the demons with AI? She'll share with you a sorted version of machine intelligence to help you decide for yourself. She admits it will not be the sordid version of AI presented by people she respects greatly, like Mr. Musk, Mr. Hawking, and Mr. Gates. You'll learn the current state of machine learning, the players and their tribes, what's possible today and a bit into the future, and where she worries and where not, from the perspective of someone who considers daily if she is, indeed, in the belly of the beast. She'll be delighted to get your feedback and experience in questions during and after.
7. 90% of the world’s
data was created in
the last 2 years
@jeggers
8. It’s a signal to noise problem
Perfect for AI
One morning (and it will be soon), when everyone wakes up as a
writer, the age of universal deafness and incomprehension will
have arrived.
Milan Kundera, 1978
9. Machine-to-machine communication and learning also
help managers increase their
capability and capacity and the speed
of their decisions.
The potential uses have barely been scratched, and
the growth opportunities of this bend
in the road can be immense for those
who seize them.
Ram Charan, The Algorithmic CEO
Fortune, January 22, 2015
@jeggers
WIIFY
10. Breaking the magicians code on ML
• Definition: algorithms learning from data;
etymology <1960 AI and statistics>
• Common words: supervised, unsupervised, neural
networks, Bayesian methods, deep learning,
pattern recognition, singularity
• Next generation of: filtering and segmentation,
rules-based systems, statistical analysis, expert
systems
It’s just maths @jeggers
16. Image by Eyewire
Our goal:
General-purpose
recommendation engine
Our approach:
Neuroscience + ML
17.
18. Homework: What is your problem?
• Do you have loads of “unemployed”
data?
• Have you segmented your data or
business in order to solve your
problem(s)?
• Does your problem fit AI/ML?
• Can you define a Proof of Concept?
• Where in your network is expertise?
@jeggers
19. More sunlight
More patience
More diversity
Don’t undervalue you
Don’t succumb to fear
What I
think about
when I think
about the
future of AI
@jeggers
Editor's Notes
Give Roomba example
Most AI is for a specific task. We use AI all the time in our daily lives, but we often don’t realize it’s AI. John McCarthy, who coined the term “Artificial Intelligence” in 1956, complained that “as soon as it works, no one calls it AI anymore.”
We do not have any examples of “general AI”, and we aren’t any where near super intelligence, i.e., smarter than us. ;-). We need less theory (& hype) and more reality.
That said, growth has been faster in the last 10 years than the previous 40.
Superintelligence is predicted by the experts to be by 2060.
* Within thirty years, we will have the technological means to create superhuman intelligence. Shortly after, the human era will be ended.
—"The Coming Technological Singularity" (1993) by Vernor Vinge (SDSU)
* Fermi paradox – we should have been visited by someone who created superintelligence already.
While you can say that superintelligence could render us obsolete, it could also be the thing that makes us immortal. It could go either way. Again, kind of like a kid! ;-)
Camps: There are really only two camps that reach the court of public awareness:
* Team Kurzweil: Superintelligence is Nigh, and will give us everlasting life
* Team Hawking: Superintelligence is Nigh and it will kill us all
I want to create a new party and ask you to join
* Stockdale Paradox:
You must retain faith that you will prevail in the end, regardless of the difficulties.
AND at the same time…
You must confront the most brutal facts of your current reality, whatever they might be.
* Team Stockdale: Enjoy and drive our current reality with the understanding that there’s some chance it might now work out.
Why do you even care?
… and that rate is only increasing.
Think about how much data is behind each of these points.
One man’s noise is another man’s treasure
Use of algorithms has already become an engine of creative destruction in the business world, fracturing time tested business models and implementing dazzling new ones.
the ability to acquire and apply knowledge and skills.
While Machine learning is statistical analysis, it is a more active version. This is more about continual analysis or constant learning versus one and done. Then one again and done again.
Mostly the same algorithms – some advances like back propagation. Overall, just evolution.
Expert systems.
170 AI researchers on Wikipedia
1. From Wikipedia "Artificial intelligence"
2.1 Goals
2.1.1 Deduction, reasoning, problem solving
2.1.2 Knowledge representation
2.1.3 Planning
2.1.4 Learning
2.1.5 Natural language processing (communication)
2.1.6 Perception
2.1.7 Motion and manipulation
2.1.8 Long-term goals
2.1.8.1 Social intelligence
2.1.8.2 Creativity
2.1.8.3 General intelligence
2. From my AI textbook in college (Russell and Norvig)
Point of interest: Norvig is now Director of Research at Google! Russell stayed in academia (at Berkeley).
natural language processing to enable it to communicate successfully in English (or some other human language);
knowledge representation to store information provided before or during the interrogation;
automated reasoning to use the stored information to answer questions and to draw new conclusions;
machine learning to adapt to new circumstances and to detect and extrapolate patterns.
computer vision to perceive objects, and
robotics to move them about.
Also mentions agents and logic later in the section.
3. A somewhat slanted view (slide from a researcher in England) but speaks to the multi-dimensional nature of it: 1. AI has inter-related scientific and engineering goals.
2. AI has its roots in several older disciplines: Philosophy, Logic, Computation,
Cognitive Science/Psychology, Biology/Neuroscience, and Evolution.
3. Major sub-fields of AI now include: Machine Learning, Neural Networks,
Evolutionary Computation, Vision, Robotics, Expert Systems, Speech
Processing, Natural Language Processing, and Planning.
4. Major common techniques used across many of these sub-fields include:
Knowledge Representation, Rule Systems, Search and Learning.
Next group:
Cornell, Princeton, Harvard, Toronto, UCL, CalTech, Edinburgh
Facebook growing, SRI decreating
More like Cliques w/ AI being new to business
Jaron Lanier: “Humans aren’t biological computers”.
We need more minds and more ideas in this space.
Diversity in thinking is going to help us raise a well-rounded child, and we do need to confront social issues too. What are the social norms?
Not to mention ethicists, ethnographers, etc.
An example of what we are doing…
Not l33t
I started with the question most posed to me… are you creating killer robots. I’ll end telling you what I think about when I think about this…
Haruki Murakami
More sunlight, less streetlight – regulation scares the crap out of me!
Regulation isn’t going to help and honestly it is scarier
Did I mention this scares me?
Much less controllable than nuclear bombs
We want focus on business!!!!
Patience for results – it is early, stop the hype
We are raising a child… and we might have some terrible twos.
Worried that this will remain l33t, worry it will get regulated.
Stop focusing on when