During the previous lecture we briefly discussed how Google’s Artificial Intelligence has beaten a human top player in the game Go, which was a major breakthrough in the field of AI. Recently, I came across an article related to this subject (Economist, 2017), which discusses how advances in AI can read people’s faces by spotting patterns in large amounts of data. Since AI is able to analyse more pictures than a human potentially could, AI is able to analyse subtle differences in facial structure, that a human cannot. This goes as far as being able to predict sexuality (gay versus straight) based on being shown 5 pictures of a person, with an accuracy of 91% for men and 83% for women. Shown the same pictures, humans predictions were only accurate 61% and 54% respectively. It is expected that AI systems will soon be trained to accurately read people’s face on other intimate traits as well, such as IQ and political views. Based on this article, I would like to discuss two controversial questions with you:
- Safety / Privacy: Is it ethical to allow machines to read people’s faces and discover traits that other people, or even the person itself might not be aware of? Besides the concern that AI is still not 100% accurate, I believe that certain personal details are private and acceptance should first be asked when reading a person’s face. The reason for this is that complete transparency about personal information could be easily be abused. Employers may for example screen on several traits when assessing job candidates, health insurance providers may more easily assess people’s health when accepting clients, governments may disadvantage certain groups in society, etc.
- Future employment: When AI already outperforms humans in several tasks, will AI eventually exceed human capability on every single task and what will be consequences of this? According to recent article published by Dailymail (2017) this will indeed happen and will not even be that far away in the future. They claim that there is a 50% chance AI will outperform humans in all possible tasks within only 45 years. Furthermore, they expect machines to take over all human jobs within 120 years. This will range from translating languages (2024) to working as surgeons (2053). If this would indeed happen, I believe that consequences would be huge, since this causes changes in the entire society, including people’s daily life, employment, law, security, etc., although it would be hard to quantify how huge the changes will be, especially since they would intervene with other technological advancements.
What do you think of these two questions raised and do you have examples of any other AI advances that are particularly interesting/promising?
References:
Dailymail. (2017). AI will outperform humans in all tasks. Retrieved from Dailymail.com: http://www.dailymail.co.uk/sciencetech/article-4560824/AI-outperform-humans-tasks-just-45-years.html
Economist. (2017). Advances in AI are used to spot signs of sequality. Retrieved from Economist.com: https://www.economist.com/news/science-and-technology/21728614-machines-read-faces-are-coming-advances-ai-are-used-spot-signs
