Should intelligent decision-making be handed over to Artificial Intelligence?

As more enterprises around the world start investing in Artificial Intelligence (AI) and its subsets Machine and Deep Learning, Image and Voice Recognition, Artificial Neural Networks, etc, scientific and business leaders have an interesting question to answer. Should AI be used solely to ease navigation through big data and augment human capabilities, or should it be allowed a more independent role in intelligent decision-making? Before getting AI fully involved in that sophisticated process, a few other important points need to be addressed. Experts in a panel discussion at the AI Expo in London in April said AI should have clear definitions of what data to collect and what to use it for. It also needs to overcome its current bias, caused by insufficient training data.

With the Internet of Things (IoT) connected devices already producing large amounts of data at great speed, the industry is faced with a new challenge – how to identify the valuable bits. ‘’The easiest way is to automate the analysis of the aggregated datasets’’, according to Bjorn Skou Eilersten, Chief Technology Officer at Milestone Systems.  At present this is happening in the cloud, but experts say hybrid AI scenarios, where AI systems do some processing in the cloud and some processing locally, are also possible and will depend on the nature of each application. The current stage of AI automation was referred to as ‘’probing time’’ by some experts at the Expo, as questions related to raw data gathering and its transformation into useful, structured information are still being answered.

Discipline in data collection

This makes some experts doubt AI’s ability to live up to its expectations. But there are others who claim it is already providing the discipline needed for collecting the correct data. Patrice Slupowski, Chief IoT Officer and VP Digital Innovation at Orange talked about an algorithm, developed by his company, that deals with more than 15 million sets of data to predict risk of deterioration in patients with rheumatoid arthritis.  Rheumatoid arthritis is a long-term condition that causes pain, swelling and stiffness in the joints.  Monitored patients wear activity-measuring wristbands similar to the commercially available fitness-tracking wearables. Orange’s AI is trained on the data harvested by those devices. According to Slupowski, the algorithm is now capable of predicting the onset of the flare-ups, allowing for better and more timely treatment in each individual case. The algorithm is already deployed in some hospitals in Paris.

Despite the growing number of AI applications, its progress is still far from steady. According to Bjorn Eilersten it will take another two to three years for AI to combat and eliminate one of its major hurdles - bias. ‘’We just need to keep training our AI without stopping. I believe the general solution is a couple of years away’’, Eilersten added. Meanwhile Orange revealed that it is already tackling the issue by hiring psychologists to ‘’teach AI common sense and make it understand human behaviour … Everyone is preoccupied with AI cutting jobs, but this is how it helps create employment’’, Slupowski explained.

There were some conflicting opinions on the question how self-sufficient AI should be allowed to become in its evolution. ‘’Would you board a plane that is fully automated, with no human crew?’’, Bjorn Eilesten asked the audience at London Olympia. He insisted that there should always be a room for crucial human intervention in the process of intelligent decision-making. Simon Fabri, Technical Director Heating Controls and Connected Home at Schneider Electric also declared support for an ‘’appropriate governance’’ of AI. According to Orange’s Slupowski, however, a more independent AI could help eradicate human errors in sectors like healthcare and road safety.  He stated that in the next five to ten years human driving will be banned completely and autonomous vehicles could help make road accidents history.

Chatbot diagnosis of medical conditions ... and more human-like AI for autonomous vehicles

At present, there are already AI-driven applications with certain levels of independence. For example, Babylon Health, a UK-based company, and Tencent of China have developed AI-based chatbots to deliver personal health assessments, treatment advice and health records for users in China with limited access to healthcare. Patients provide answers to the questions posed by the chatbots, which identify illnesses and recommend (or not) a visit to a doctor. The US-based start-up iSee is working on an AI for self-driving cars that relies on the cognitive approach people use to understand the world. iSee aims at making future autonomous vehicles better prepared for the unpredictable and complex nature of road interactions and ultimately capable of independent decision-making. All current developments in the AI world point towards its high potential when it comes to intelligent decision-making situations, but the experts at the London AI Expo confess it is yet to be fully trained to understand and replicate the complexity of human logic and reason.

[Image licensed by Ingram Image.]

Add this: