Is Norman, The 'Psychopath AI' One Step Too Far?
In insight / By Mark Flynn / 11 June 2018
The week’s news has once again been dominated by AI. But there’s a good reason behind that due to some very interesting developments in that sector.
The first and perhaps most worrying thing to come out of the world of AI this week was ‘Norman’. Now admittedly creating a psychopathic AI sounds like a terrible idea but the creation of Norman’s is designed to help to raise awareness of growing issues within AI research.
Norman was purposely fed data from some of the darkest corners of the web and was then asked what it could identify from inkblot drawings. The results were compared with another AI that was fed normal data. The ‘normal' AI identified the images as flowers, birds and people whereas Norman identified the same images as people being brutally murdered.
Now, this experiment might seem a waste of time but it helps to highlight the issue of bias within AI. Most assume AI will be unbiased as it’s a machine, but we’ve seen with the Norman experiment, AI will carry forward any bias within the data it is trained on.
The other big newsworthy item that relates to AI came from Ben Broadbent, a deputy governor of the Bank of England. In his interview with the Telegraph he talked about how the British economy appears to have entered a climacteric phase.
A climacteric economy is one that has become less potent and is generally experiencing slow productivity growth. The term is particularly used to describe the slowdown of productivity growth that occurred in the 19th century which some historians believe was caused by the shift from steam to electricity.
As such Mr Broadbent believes that we may have reached the peak productivity provided by the most recent industrial revolution and driven by the internet. He believes that we must now wait for the next industrial revolution which may come from AI.
As the wheels of the next potential industrial revolution begin to turn it is important to look at how others are using AI to their advantage. Starbucks is using AI to boost performance in a number of telling ways for example.
When considering AI, the first thing to look at is where the data for the AI comes from and if it might have any biases. The sources Starbucks use is their app and reward scheme, both of which have millions of users which helps remove bias thanks to a large data pool.
The most obvious use for AI is in personalising the customer experience by offering deals based on things they like or on things people with similar tastes have liked previously, as well as incorporating local weather reports and the date in order to give the customer the best experience possible in real time.
A less obvious use of AI in Starbucks is used when choosing new store locations. Deciding to open a new Starbucks is a complicated process that AI helps to simplify greatly. Evaluating and combining several different types of data such as proximity to other stores, the demographics of the area and the traffic around the site all helps to simulate not only how well the store would perform but also how it would affect other local stores.