Archive for October, 2015

Artificial Intelligence Post Number 21

October 29, 2015

I just finished reading “The Master Algorithm: How the Quest for the Ultimate Learning Machine Will Remake Our World,” by Pedro Domingos. This is a very difficult read. The author covers certain areas in great (and confusing) detail, and is contradictory in other areas. For example, despite the concerns and warnings from many very smart people about AI, the author states that there is absolutely no risk of artificial intelligence computers taking over. But he then proceeds to list some ways that computers MAY take over, including The Master Algorithm getting into the hands of a bad person. What are the odds of THAT happening? He also says that there is a risk of people getting so confident of computers always being right that they begin to follow the AI computer like following a god? Again, what are the odds of THAT happening? Have people ever really done that?

The author believes that it is important that The Master Algorithm be discovered BEFORE individual algorithms get developed for specific narrow problems, because otherwise the detail of those individual algorithms will become too complex to incorporate into one Master Algorithm. Again, the author spends chapters talking about the progress already being made in these individual learning algorithms, including areas like self-driving cars, medical care, etc. He doesn’t say how we are supposed to stop this progress until someone identifies an overriding Master Algorithm. And he does say that the individual algorithms on specific subjects will be more all-encompassing. He just feels that required computer power will be overwhelming when someone tries to put everything together into one Master Algorithm package equivalent to a brain. He doesn’t mention breakthroughs like IBM’s TrueNorth computer chip.

The author then lists the competing philosophies being applied in the development of a Master Algorithm, which he calls “the five tribes of machine learning.” The five tribes are symbolists, connectionists, evolutionaries, Bayesians, and analogizers. I won’t even attempt to describe the details of each, which is the main content of the book.

I read this book because the author Pedro Domingos is a professor of computer science and is a winner of the SIGKDD Innovation Award. The book is very recent (published September 22, 2015) and is rated fairly well on Amazon. So I thought that I would be getting very up-to-date information. But I don’t feel it helped much in this blog’s quest to see the progress of AI in general, and when we should expect to see dramatic changes as the result of AI. I actually think that my fictional book “Artificial Intelligence Newborn” does a better job in laying out a possible AI future, especially given that Domingos gives a zero chance of AI taking over!


Artificial Intelligence Post Number 20

October 18, 2015

I would like to give a brief update of how much effort (money) is being put into artificial intelligence by hugely wealthy companies. First, let’s look at Apple. Just this month Apple purchased Percptio and VocalIQ. Perceptio makes image recognition technology for smartphones. VocalIQ is developing technology that helps computers understand everyday human speech. Does Apple want these companies’ expertise to help in the development of their rumored self-driving electric car that is supposed to come out in 2019?

Apple is not alone! In a 10/12/2015 event in San Francisco, IBM hosted a meeting to discuss AI. In this meeting IBM was blowing their own horn, emphasizing that they were working on every element of AI, with Watson and the TrueNorth chip offered as evidence. They also discussed how they can get insights from unstructured data. Earlier this year they noted that in the healthcare industry, much patient information is saved in text format and seldom used in analysis. Some estimates are that 80% of health information is unstructured (like in physician notes and patient surveys) and therefore not used. IBM Watson Content Analytics addresses this source of unused information to give a more complete view of patients’ needs and appropriate treatments.

Google recently invested in DFKI, a German AI lab. In 2014 they also paid $500 million to buy the UK company DeepMind. DeepMind’s Mission: “Solve intelligence!…We build powerful self-learning general purpose algorithms.”

Musk, Zuckerberg and Kutcher are investing in a company called Vicarious, which has raised over $100 million. Vicarious’ goal is to build a system that will have general intelligence matching a human.

Many of these companies would pursue AI even more aggressively if they could hire people with the required abilities. Google has publicly stated that their biggest challenge in AI is finding people with the right digital skills.

The last thing I want to emphasize is that despite all this effort by well-funded firms, I believe that there is at least an equal chance of AI breakthroughs coming from a quant playing with his computer in his basement and using the cloud for his required computer power. His motivation for developing self-learning AI programs will be the same that many hackers have of breaking into “secure” computer systems. It’s an ego thing!