Review of AlphaGo and AI Ethics Documentaries
Our team of enthusiastic young individuals set out to understand how AI was being applied and perceived in the modern world. For this, we primarily watched two documentaries — one on AlphaGo and the other on AI Ethics in general. Below you can find their summarized reviews and analysis.
AlphaGo Documentary
The 2,500-year-old Chinese board game called ‘Go’ was a major challenge for Artificial Intelligence, justifiable by having more board configurations than atoms in the observable universe (Krieg, Rosen, Proudfoot & Kohs, 2017). Directed by Greg Kohs, this intriguing documentary unveils the journey of how a small team of engineers from London managed to do the inconceivable.
After a jaw-dropping 5–0 win against Hui, AlphaGo started making headlines. The loss shocked Hui as a machine had never defeated a human player in any match before (Krieg et al., 2017). Skeptics weren’t convinced with the result and AlphaGo got a stronger competitor to take it head-on in another 5-game match.
Hundreds of millions tuned in to watch the battle between Lee Sedol and the AI, which was now stronger than ever. Lee, known for his creative play, something that is thought of as exclusive to human intelligence, was confident he wouldn’t lose to a machine (Krieg et al., 2017). After the nail-biting match, AlphaGo was the ultimate champion and people started to understand the true potential behind this technology. The event marked a major achievement in the field of general AI.
During Match 2 against Lee, AlphaGo made a controversial play — Move 37. This move was one that proved the AI was not simply playing against a fixed set of rules and rather appears to create its own algorithm and mimic understanding in a way that is indistinguishable from the real thing (Metz, 2016). AlphaGo displayed what seemed to be intuition that seemed to exceed human intuition (Krieg et al., 2017). Move 37 can be regarded as the moment humans and machines began to evolve together (Krieg et al., 2017).
What was so fascinating was that Move 37 was not part of the 30 million human Go moves that were used to train the deep neural network (Krieg et al., 2017). Then how did AlphaGo learn to play it? AlphaGo itself calculated that the probability of a human playing that move was 1 in 10,000 (Krieg et al., 2017).
Despite the odds, having played the move it showed that the machine had started to think on its own. That is why Lee and everyone watching was taken aback by that play. They didn’t understand the logic behind it. However, in the end, AlphaGo demonstrated its superiority because that was the instrumental move that got it the victory.
Similarly, in Game 4, just before lee Sedol played Move 78, AlphaGo calculated it’s chances of winning at 70 percent. A few moves later, the percentage drastically dropped and Lee had won his 1st game against the machine. The infamous “wedge move” effectively cut AlphaGo’s defenses in half (Metz, 2016).
The AI had taken so long to play its turn after Move 78 that the creators started to think it had gotten lost deep in its own algorithm that probably contained a bug (Krieg et al., 2017). The next move played was horrible and it was evident that even AlphaGo had a weakness and Lee Sedol had found it (Krieg et al., 2017).
AlphaGo is a neural network that mimics the way the human brain learns and picked up Go by mimicking human play and then playing against old versions of itself to improve (Krieg et al., 2017). The AI tries to maximize its chances of winning without giving importance to the margin (Krieg et al., 2017). The 3 main components of AlphaGo are the Policy Network, Value Net, and Tree Search (Krieg et al., 2017). Each played an important role in the decision-making process from training to evaluating board positions to predicting possible outcomes.
This stunt opened the world’s eyes to a new discovery. The AI has proven its success in perhaps the most complex game ever conceived. Lee had expressed how he had a new-found interest for the game after playing with AlphaGo and Fan Hui mentioned how the AI had given him a new perspective to the game and taught him many new moves (Krieg et al., 2017). The machine did a very human thing and played a moved that asserted its superiority to its developers. A seed has been sowed that will grow into something bigger in the future where AI tech could drive a new kind of scientific research, where machines point humans toward the next big breakthrough (Metz, 2016).
AI Ethics Documentary
The second documentary we came across covered interesting applications of AI and the ethical implications resulting from it. We were also introduced to a few influential people who contributed to establishing an AI Code of Ethics.
Jürgen Schmidhuber, a German Computer Scientist, predicts that AI will spread from Earth to the cosmos and that AI could be sent and received in the speed of light (Schneider, 2019). AI was claimed to be the most important development (Schneider, 2019). A great example is how a woman’s life was saved in Japan when IBM’s Watson detected a rare Leukemia in 10 minutes, where a human would’ve otherwise taken 2 weeks (Schneider, 2019).
When questioning whether AI can further be used to take care of patients, it raises concerns about whether a robot can give the same emotional support like that of a human. Another story of a man named Kondo who married Miku, a hologram, had made headlines (Schneider, 2019). There may be some positive aspects in this considering that Kondo otherwise found it difficult to find love and had to turn to an inanimate character that stimulates the feelings and dialogues of a real wife. However, can we say that this is the right thing to happen? How much self-deception should be allowed in society?
A major downside of robotics and AI comes into picture when assessing its potential contribution to warfare, the spread of diseases, and weapons of mass destruction. It is a possible reality but not ideal to implement in society (Schneider, 2019).
Back in the day, biologists managed to get biological weapons banned (Schneider, 2019). The physicists, on the other hand, failed (Schneider, 2019). That is why nuclear weapons exist (Schneider, 2019). If we do not learn from our past and repeat the same mistakes, what would have otherwise been a gift, would instead become a tragedy.
Google Duplex is another really interesting application that has impressed audiences around the world. The entire idea of an AI making calls just as if it were a human, seems fascinating. However, in the Brussel’s panel related to social AI, Metzinger had called for guidelines for interaction between AI and people (Schneider, 2019). He called for a ban on AI systems that don’t see themselves as such when dealing with humans and says that AI should never be allowed to manipulate the people who use it (Schneider, 2019).
AI could also have a potentially positive impact on political systems since it is less susceptible to bias, conflicts of interest, bribery and other issues that regularly inhibit human politicians’. While every development in the field of AI can be put to positive use, it can be misused without proper guidelines or regulation. Ethics experts say we need to deal with this situation before it gets out of hand (Schneider, 2019). What they propose is safety engineering. As stated in the documentary, “The development of AI must be subject to strict ethical guidelines. Otherwise, we may become slaves to our own technology. (2019)”
References
Krieg, G. (Producer), Rosen, J. (Producer), Proudfoot, K. (Producer) & Kohs, G. (Director). (2017). AlphaGo [Documentary]. United States: Moxie Pictures.
Metz, C. (2016). What the AI behind AlphaGo can teach us about being human. Retrieved from https://www.wired.com/2016/05/google-alpha-go-ai/
Schneider, U. (Producer & Director). (2019). Paradise or Robocalypse? The curse and blessings of Artificial Intelligence [Documentary]. United States: DW Documentaries.
