Artificial Intelligence And Intelligence

AI has made tremendous strides in the past 15 to 20 years. In going from a constraints and rules based system in which “expert systems” were hailed as the future, to today’s probabilistic and stochastic systems we’ve come a long way. There are many current and future uses for AI and they are far too many to list.

Some recent work has been done in gait simulations in which everything from dinosaurs to people walking with osteoarthritis or cerebral palsy (crouch gait) has been modeled. Strategies have been proposed to lessen the load on the knee using one model. But the strategy was difficult to adapt and I believe only the author of the study was able to successfully and readily use his “intoe” strategy. (Note: Lateral wedging on orthotics has also been found to lessen medial compartment forces in the knee. And that does not need you to study at Monty Python’s Ministry of Silly Walks, to walk like Young Frankenstein or to sing “Walk This Way” while you move about.)

Biorobotics, a new field of robotics, has been used to model a variety of animal methods of locomotion. Snakes, cats, fish, and human like robots have been created for this purpose.

The October 7th edition of Science Magazine features an interesting article which suggested that artificial intelligence products were needed to make connections and correlations for novel ideas on research projects. (Gill, Y et. al. 2014; 346:171-172). This is an excellent idea.

An intelligent system should be at the core of many systems. It should be at the core of all EHR (electronic health records). Some of the advocates, designers, and marketers of these products seem to be first interested in market penetration and are proud of achieving a government defined Level 2 of “meaningful use”. However meaningful use is losing its cachet since it doesn’t add intelligence, thought, or many things that are helpful to a medical record. Insted it adds more data points but not meaning. (ICD 10 coming to a healthcare provider near you next year will also add an incredible amount of not very useful datapoints and has been precicted to cause many physicians to leave private practice.)

AI could be helpful. AI could have assisted by creating a red flag notice on the Electronic Health Record that the patient with a headache and severe stomach pain had just arrived from Liberia. It was entered into the medical record of the hospital but set off no alarms.

Of course the final failure was a human one. The doctors, nurses, and residents should have made connections. It really doesn’t take a computer to put information like that together. The close partner of the patient who brought him to the hospital is said to have informed 3 people that he had come from Liberia, Africa. He did not have a Texas accent. He had more than one classical sign of Ebola or other serious illness. He rated his GI upset at 8/10, he had a fever and a headache and did not have signs that he had a sinus infection as some articles stated.

In this case an intelligent system could have made up for the unwise human conclusions and actions. But there are a few simple lessons:

1) We need to be intelligent and make connections. A diverse knowlege base is helpful.
2) We need to exclude data points that don’t make sense and are suspicious and not likely.
3) We need to have an intelligent core to our EHRs. One in which probabilty and hidden Markov Processes are used will be far better than just using Natural Language Processing (NLP) but it will be harder to implement.
4) The humans are the final arbiters of the decision making process and need to think and understand their own thinking process to produce optimal decision making and to determine the optimal next step in treating their patients.

“The world faces deep problems that challenge traditional methodologies and ideologies. These challenges will require the best brains on our planet. In the modern world, the best brains are a combination of humans and intelligent computers, able to surpass the capabilities of either one alone.” Well stated by Gill (2014).

AI-ScienceMag-small
AI Connections via the Hanalyzer.

 

References:

Gill, Y et. al. Amplify scientific discovery with artificial intelligence.  Science Magazine 20 14; 346:171-172

Ijspeert, Auke. Using robots to emulate and investigate agile locomotionScience 10 October 2014: Vol. 346 no. 6206 pp. 196-203.

EDx: CS169.1x Software as a Service (SaaS) – Rotten Potatoes Setup

Something is rotten about these potatoes.

The following information is for those signed up for the course at EDx.org: CS169.1x.

The subdirectory “rottenpotatoes” and the files for the program  were just plain missing from the VM – vdi files supplied and downloaded.

The instructions for downloading and setting up Virtualbox and the VM image are found at the beta.saasbook.info web site. Text and a video screencast are there and are pretty clear. Except the files needed to run the program were missing.

I set up Ubuntu Linux as described which worked just fine, accessed my network, surfed the web using Firefox and all looked well.  But, there was no subdirectory containing the test assignment program “Rotten Potatoes” – a takeoff on Rotten Tomatoes, the movie review site.

Solution:

Since it is a github project:

server$ git clone git://github.com/saasbook/hw2_rottenpotatoes.git
server$ cd hw2_rottenpotatoes/
server$ rails server
Some gem files were missing so:
server$ bundle
Then it turned out the “movie” table was missing so the following did the trick:
server$ rake db:migrate
server$ rails server
The gist of this was found thanks to eekbix.wordpress.com. My version is shorter, but was all I needed to do. Hopefully this may save a few others some hours of tinkering and wondering.
Rotten Potatoes – Working like a charm.