Author: Katarzyna Krzywicka Edited by: Inês Barreiros
We are undeniably witnessing the rise of Artificial Intelligence. Ever since we have impatiently watched IBM’s Deep Blue defeat world chess champion Garry Kasparov, it was clear that a computer is something more than a spreadsheet-processing box. It became a powerful tool, with analytical capabilities even the most talented humans cannot dream of. With such rapid technological progress, however, computers have made a leap from their analytical safe space to elusive world of speech and emotions. In other words, it seems that everything we have learned about computers is wrong.
After Kasparov’s memorable defeat in 1997, subsequent victories of supercomputers over current human world champions continued to shake human confidence and honour as the machines mastered skills which we obliviously perceived as reserved for humans. In 2011, IBM Watson stunningly outperformed human world champions in Jeopardy! - a game requiring linguistic intelligence as well as detailed general knowledge. In 2016, Google DeepMind’s AlphaGo outcompeted 18-time world champion Lee Sedol in a hugely popular strategy game in Korea. Due to the almost infinite number of possible moves in the game (more than the number of atoms in the universe!), it usually requires not only sophisticated analytic and strategic skills, but also some intuition and experience. This means that AlphaGo was not processing all possible moves and choosing the most appropriate – like it was done in the chess challenge nineteen years ago – but it had effectively learned from its own experience. Using the so-called “reinforcement learning” AlphaGo had used all the knowledge it had from the previous games to develop a strategy which led it to its victory.
Although in the past we learned about AI from TV shows, nowadays it no longer serves entertainment purposes only. Giving rise to big data management and deep learning models, it is increasingly used in technology, business and healthcare industries. Especially healthcare - in desperate need for innovation and increase in efficiency - is soon to be revolutionised by AI, by improving its efficiency and cost-effectiveness. It is also potentially a long-awaited solution for rising costs of healthcare and social care in the ageing population.
Over the past two decades, heavy investments from the biggest technological giants played a major role in the rapid acceleration in development of the AI technology in healthcare. IBM Watson for Oncology, constructed to help physicians make evidence-based decisions, has been taught to collect and “understand” specialised medical knowledge. It analyses medical records and imaging results, correlates it with knowledge from more than 1.5 million patient records and a million pages from medical journals, to provide the physician a diagnosis. It also issues a recommendation on further tests and advices on treatment options according to the medical history of a particular patient. Watson Oncology is currently being trained by the clinicians in the Memorial Sloan Kettering Cancer Center. It is claimed to be 85 to 95% accurate and it continuously improves itself based on the received feedback. DeepMind Health Project developed by Google, also collects and processes patient data to obtain a comprehensive database of diseases. On the other side of the block, Bio Model Analyzer introduced by Microsoft is developing a detailed model of cellular mechanisms and interactions leading to cancer progression. In the primary care sector, Babylon app provides an alternative to a visit to a General Practitioner – it simulates the consultation, correlates the symptoms with a disease from the database, offers treatment and ensures timely follow-up. The field of imaging does not stay behind either - Enlitic Inc. have produced a software which, in 2015, detected lung cancer with 50% higher accuracy than a group of qualified radiologists. Countless companies and startups are monitoring diseases, collecting genome information and discovering drugs– and there is much more to come.
Big data collection and analysis are the natural habitats of a computer. But the potential of machine learning is far beyond its traditional fields. For instance, more and more AI is finding applications in human-dominated fields such as psychology and psychiatry. In 2015, NeuroLex Diagnostics introduced AI technology that uses automated speech analysis for 100% accurate prediction in determining which patients with schizophrenia will develop psychosis. Machine learning has learned to perform “semantic analysis” on “two markers of speech complexity - the length of a sentence and how many clauses it entails”. The company is currently working on recognising change in speech patterns in patients hospitalised with depression or psychosis. This is expected to improve the efficacy of the treatment and shorten the inpatient time. In the future, it is expected that AI might aid in recognising specific patterns of behaviours among individuals and aid in early diagnostics of mental disorders such as depression or bipolar disorder.
Even more notable, however, is a recently developed startup X2AI, currently working on Tess AI – an app providing accessible mental health services. The system provides its users with personalised “psychotherapy, psychological coaching, and even cognitive behavioural therapy”. Michiel Rauws, the CEO of X2AI, claims that it is based on “emotion algorithms” and “accuracy of the conversation algorithms” – “which understand the meaning behind what people say”. He explains that Tess adjusts its tone depending on the conversation and it adapts to its interlocutor and keeps track of the conversations. As opposed to its human alternative, it is not, however, tired, prejudiced or judgmental. In the future, Rauws sees not only mental health, but also the entire healthcare as more accessible, “sustainable and affordable”.
On a final note, it should be highlighted that although the progress in AI has perhaps exceeded our expectations, its use still requires a considerable amount of time to meet the standards of clinical care. Computers, as advanced as they are, remain to adapt to the complexity of the human world. A recent attempt to “humanise” the famous IBM Watson - which involved teaching the computer “The Urban Dictionary” - resulted in on-air embarrassment when the system spat out curse words and insults. Therefore, despite the imminent progress in robot social intelligence and the compelling perspective of computers performing our marriage counselling, we should not forget the value of human presence, human voice and human touch.