Humans Versus Artificial Intelligence
The complexity of the human brain, in relation to that of other species, is one of the biggest wonders in science. Today, neural networks compete alongside us with processing powers that can calculate 200 million potential outcomes per second. With the data trail we feed into it as we go through our lives, and the vast data sets scientists are training machines on, they are learning to compete with the greatest human minds at our own games.
In this story we explore how humans compare with machines in games, healthcare, art and emotional intelligence.
Games
AlphaGo, DeepMind’s Go playing AI, has been dubbed “The AI that has nothing to learn from humans”. Go is a complex strategy game over 3000 years old, with 10170 different board configurations. It played against itself thousands of times, getting stronger every time. It found inventive winning strategies that haven’t been played in hundreds of years, winning its first game 5–0 against European Champion Fan Hui, and later 4–1 against 18 world title winning Lee Sudol. DeepMind have since made more advanced versions; AlphaGo Master, AlphaGo Zero and AlphaZero.
Chess is another game where computers are revealing tactics overlooked by humans, for instance maintaining pressure until the opponent makes an error. IBM’s DeepBlue AI defeated world Champion Garry Kasparov 4–2 in 1997; which is unsurprising when, despite the greatest minds, AI can calculate 200 million moves per second and map 5 to 6 potential consequences of each move.
Machines are learning new strategies, showing potential for humans and machines to collaborate to each improve and become more intelligent together.
Healthcare
AI can compute vast amounts of data, for instance databases of symptom data, disease causes, test results, medical images, latest medical papers, doctors reports. This means it can spot patterns from data that a single human never could.
Over the past few years AI has become even more accurate at identifying disease diagnosis in patient scan images and health reports. In a study, it was found that AI could diagnose with 87% accuracy, which doesn’t seem like good enough, but the same study found that healthcare professionals diagnosed the same data with 86% accuracy.
The US Institute of Medicine report 1 in 10 diagnoses are wrong, leading to 80 thousand unnecessary deaths per year. An improvement of 1% accuracy with AI assistance could help limit this number.
One team at University of California are working with 1.3 million young patients to train an AI that can diagnose Glandula fever, roseola, influenza, chicken pox and hand-foot-mouth disease with 90–97% accuracy.
The benefit of AI is that it doesn’t get tired, hungry, or lose concentration, like a doctor would. The downside is that a lot of health data has ethnic bias, meaning training databases may not be diverse enough to work for all patients. However, it is the beginning, and currently showing signs of being a vital tool that can be used alongside doctor’s diagnoses to improve detection.
AI can be used beyond the detection process, but also to identify new treatments, because it can predict a number of outcomes. For instance, AtomWise works to predict how molecules could combine with a protein, in order to advise new medical treatments.
Art
Edward Belamy’s portrait sold for $432,500 at Christie’s in New York. The creativity was entirely AI; it was trained on thousands of images of portraits and generated Edward Belamy based on its knowledge of art. This makes us question if it was just another pastiche, like AI generated music based on works of Mozart and Beethoven, or if it is true machine creativity. This sparks a debate about what human creativity really is, and whether AI can ever truly have it. A key area artists and programmers are currently working with AI on is trying to get out more than they put in — to be truly creative and original, it must make something beyond that which it has been fed on. This is where artists are collaborating with AI. For example, artist Amy Ridler used many photographs of tulips, and combined it with a program that could fluctuate the images based on bitcoin prices. In another example, AI learnt Sougwen Chung’s painting style, an in combination with a robotic arm, paints alongside her creating a beautiful balance of machine and human.
AI may not have its own creative mind yet, but in combination with artists it can generate new levels of creativity.
Emotional Intelligence
Emotional intelligence, the ability to respond to emotional signals, is one of the few skills that define what it means to be human, in addition to creativity. But as with art, programmers are working on giving AI these human skills. Deep fakes already fool many with their close replication of human facial expressions and mouth movements in real time. But AI must respond to real facial expressions in order to be emotionally intelligent. If your car could sense you are unhappy, it might suggest a more scenic route or upbeat music. AI could help with treatment for depression, anxiety, and other emotion-linked conditions. It would use speech patterns, facial expressions, physiology, which are already being programmed by design teams working on facial recognition, in addition to physiological signals like breath and heart rate, by tracking blood flow from detecting the skin. If AI begins to respond to emotional and physiological cues, it opens many doors to healthcare solutions and becoming smart and closer to being human.
Our next story will take a closer look at how policy changes are keeping up with this rapid growth of technological innovation.
This was written by a researcher at a specialist data company. The Digital Bucket Company operates in the UK and works with clients in overcoming data challenges including privacy concerns.
