In 1997, IBM Deep Blue won the reigning world chess champion Garry Kasparov.

In 2011, IBM Watson earned the price of $1 million by wininng a Jeopardy! contest against former winners.

These are impressive achievements of Artificial Intelligence.

This project tries to continue in this same spirit. It consists of two different programs:

1st program: Using Watson's Visual Recognition to aid a surgeon to locate the area affected by cancer and which must be extirpated.

It is necessary to identify cancer visually because it must be extirpated entirely, if not it would grow again. (There are other methods to dealing with cancer too, like directed chemotherapy and radiation.) In the case with surgery, if provided with high-resolution photography and trained, Watson's Visual Recognition can aid a surgeon to perceive the borders of the area where the cancer is, and, if there are multiple areas affected by cancer, then Watson can help with an optimal strategy of the order of removal of the various locations of the tumour in the 3-dimensional body (human organ), e.g., first extract this location of cancer close to the surface of the organ, then approach to extract a second location which is more profound, and so on extracting deeper into the organ, in a manner that is the least invasive (a geometrical-path optimization problem, where the targets -tumours- must be extracted there are constraints in the path, like arteries, nerves, and other small anatomical structures throught the 3-dimensional organ).

Cutting short the time of a surgery is not only beneficial for the patient, but it also cuts short the risk of a blood clot, which can, by itself, cause death. As it is mentioned the following article, 'each year, over 500,000 hospitalizations and 100,000 deaths are associated with blood clots':

http://www.medicalnewstoday.com/articles/286398.php

Besides, surgeons are humans and surgeries take several hours, during which time surgeons must concentrate profesionally in their work while remaining standing up. These are some opinions of surgeons themselves on how they deal with the exhaustation from lengthy interventions:

https://www.quora.com/What-do-surgeons-do-if-they-get-exhausted-in-middle-of-a-surgery

So it is better for surgeons to be aided by Visual Recognition, like the one Watson gives, to cut short the time a surgery takes.

For example, on the cancer:

Original cancer

this is the image of the contours this program finds with a resolution of only 30x30 tiles (the maximum resolution is to divide the original photography in 99x99 tiles):

Watson contours on the cancer

As said above, one condition for this project is high-resolution photography. The source code repository has images of how this Visual Recognition of cancer with Watson works, besides README files with explain in a formal, technical manner, how it does it. The background of this program is that I had to do an exposition several months ago where it was analyzed the increase in the cost of personal health care. An introduction to this issue is: https://en.wikipedia.org/wiki/Health_insurance_costs_in_the_United_States

2nd program: tries to follow the ideas of Jane Jacobs for Urban Planning.

Urban Planning is a discipline that combines several specialities in itself: art, finances, ergonometrics, transportation, etc, and other activities that affect every-day human life. An Urban Planner must know them all, if not he/she would be a poor planner.

Urban-planning theorist Jane Jacobs, in her book 'The Death and Life of Great American Cities', posed a problem on Urban Planning: besides all the different specialties that this discipline comprises (some are in the paragraph above), if it does not bring interesting variety to humans (ie., if its settlement is monotonous), then that urban planning project will fail in the long-term, and the city itself may die. So, in chapter 7 of this book, she says that the point of interesting "variety" is the most essential one in an urban planning project, over all the other points. As an example of a failed urban planning project is the Pruitt–Igoe neighborhoud in St. Louis, Missouri, which cost $36 million in the early 1950s [around 315 million current dollars] over 57 acres, but had to be demolished less that 15 years later [https://en.wikipedia.org/wiki/Pruitt%E2%80%93Igoe].

Of course, there are also finances, materials employed in the construction, and other aspects to take into account, many of them having confounding, hidden variables meaningful to the urban planning project, so it is very sensitive. So Watson can help with its Trade-Off Analytics module to take into account these aspects.

But, again, the main point is, as Jane Jacobs said, the variety of art. So Watson must learn to appreciate art, and feel, 'wow!' We simplified this to teach Watson the [Artistic] Theory of Colors.

The Theory of Colors is how an ordered sequence of colors is perceived by a person, and the associated beauty in this sequence, and is an old topic in art and visual design, first dealt with scientifically by Isaac Newton, and then by many others, even the German writer Goethe wrote a book on this theory with precisely this title [https://en.wikipedia.org/wiki/Theory_of_Colours]. There are resources on this well-establish theory in the Internet, and one of the best books is 'The Interaction of Color', by Josef Albers.

We did this by using Watson's Visual Recognition API (V.R.) on a set of colored dominos. Our client program generates a set of dominos, each piece with an ordered sequence of colors according to the Theory of Color, and then trains Watson (a new classifier in Watson V.R.) that some subsets of dominos (of ordered sequence of colors) have certain values of sentiment of beauty, and others subsets another. So it is a mapping from ordered sequence of colors to values of sentiments of beauty, so instead of Watson V.R. saying 'this input is a tiger', to 'this input is beautiful'. These values of beauty, according to the Theory of Colors, are the ones that an Urban Planner can use with the Watson Tradeoff Analytics, where he/she adds the finances available, construction materials, etc.

Below are three ordered sequence of tiles (dominos) to train Watson to appreciate their numerical estimation of beauty:

An ordered sequence of colors An ordered sequence of colors An ordered sequence of colors

Of course, there are many details in this project. First, the short time it was developed, so it focused more on the Theory of Colors and associated beauty. Another is that the value of beauty is a sentiment on which an Urban Planner must agree beforehand and not alter later on -except in the case of intelligent homes, which can change colors later on (we did not try to use some of the Sentiment and Emotion Analysis modules of Watson, like Alchemy, because we didn't have time, but we should have). Another detail is the use of the pieces of dominos painted with ordered sequence of colors, which are then visually recognized: a more optimal approach would be to tell Watson, "the ordered sequence of colors 'Blue, #307D7E, Green' has this value of beauty", but note that one color is specified by its RGB value, #307D7E, and not by an English name, so it is to ask Watson to perceive that 'this input tensor of colors must map to this output tensor of values of beauty', and we tried to do this with PNG images of dominos into the Visual Recognition API, but a more optimal approach would be another way around into the Tensor Mathematical Library in the Watson core, and not generating PNG images files.

As in the previous application of cancer surgery, the source code repository has more technical details that can be given in this short summary.

These results from using Watson to aid Urban Planning are related to other arts, like sculpture and 3D-printing with the SolidPython or another 3D-Printing library (e.g., the Blender library, https://www.blender.org/): it is evident that if Watson learns the Theory of Colors, then, it is the first step to it using one of the before-mentioned 3D-printing libraries to build small-scale Urban Planning models.

Also, from Watson appreciating the Theory of Colors, there could be completely different uses as a computer keyboard whose keys blushed while a person types on them, when the keys [keyboard] receives from Watson a real-time sentiment analysis on the text it is being typed. Or the keys of a piano changing colors by the musical harmony. These are possible, but more distant, applications of Watson learning the Theory of Colors. Urban Planning is a more comprehensive discipline.

Share this project:
×

Updates