Sunday, 6 January 2008

Benefits and Perils of AI

In our discussions we exalted AI. Do you think that there are also dangers involved? Please discuss the benefits and the perils of artificial intelligence.

In I, Robot, one of the characters finds himself in grave danger on the surface of Mercury. The robot that has carried him onto the surface "said stupidly, 'you are in danger, Master'" (Asimov, 1968). At the start of the twenty first century the human race is still a long way off creating artifical intelligence that is this... intelligent. We do benefit from some early advances in the field, but I must agree with much of what science fiction writers predict - AI holds more perils for us than benefits in the long term.

The neural network is the most commonly implemented form of AI, "constructed from many individual processors in a manner that models networks of neurons in living biological systems" (Brookshear, 2007). These have two major benefits - they allow for extremely sophisticated models to be built from data that has a non - linear relationship and they can be trained to learn how the data in the model is ordered.

Two fields where neural networks are becoming increasingly are medicine and finance. "Using artificial neural networks, it can be monitored a lot of health indices (respiration rate, blood pressure, glucose level) or can be predicted the patient response to a therapy. Artificial neural networks have a very important role in image analysis, too, being used together with processing of digital image in recognition and classification" (Albu & Ungureanu, 2005). It is often the case that a patient will have a range of symptoms and a range of lifestyle choices and a neural network can quickly identify illness and treatment based on this set of seemingly non - related data. It has been proved that AI is better than humans at working the stock market. In a review of the book Using Artificial Intelligence to Improve Real-World Performance Lou Mendelsohn states "the authors show that the neural network approach yielded better performance than MDA [a traditional statistical model]" (2008). It can't be too long - with the cost of computer hardware falling and its power increasing - before AI increasingly replaces people in making financial decisions. When I apply for a loan, a credit card or car insurance it's not an underwriter who decides if I'm a good risk any more - it's a neural network.

All this improves accuracy and removes human error. But therin lies the problem - as the requirement for intelligent machines increases, the need for the humans that they are replacing will decrease. Already the manufacturing industry has cut "as many as 10 million jobs involving physical labor and repetitive the coming years, a large number of first-level jobs in service industries related to customer service, help desk and directory assistance will be lost due to the advent of intelligent systems" (CNET, 2004). Artifical intelligence will mean a shrinking labour market, and while this was once restricted to jobs at the lower rungs of the employment ladder I am sure that so called 'skilled' jobs will eventually simply be a case of sitting and making sure that the machine works properly. This is a scary thought indeed.


Asimov, I. (1968) I, Robot London: Granada Publishing Ltd.

Brookshear, J. G. (2007) Computer Science, An Overview (9th Ed.) Boston: Pearson Education Ltd.

Albu, A. & Ungureanu, L. (2005) Artificial Neural Networks in Medicine [Online] Timisoara: Politehnica University of Timisoara
Available from (Accessed 6th Jan 2008)

Mendelsohn, L. (2008) Neural Networks in Finance and Investing [Online] Florida: Market Technologies
Available from (Accessed 6th Jan 2008)

CNET (2004) Smart systems will erase jobs, report warns [Online] CNET Networks Inc
Available from,-report-warns/2100-1022_3-5247644.html (Accessed 6th Jan 2008)

No comments: