The advertising industry is grappling with disruptive forces and I was privileged to spend some time at GREY South Africa recently, talking about the technologies behind some key trends. The so-called fourth industrial revolution has seen the rise of cyber-physical systems like Tilley Lockey’s new bionc arm and a host of new capabilities that are rapidly changing the way we live and work. Some of the change is painful, especially for large companies, according to Accenture’s CEO Pierre Nanterne, “Digital is the main reason just over half the companies on the Fortune 500 have disappeared since the year 2000.” This a scary thought for established business models, even in advertising, a highly creative and innovative industry that have new challenges keeping brands leading in ever more disruptive and competitive markets.
We also spoke about exponential companies and what they do that makes them successful, whether by design or by experiment – usually the latter. Early digitisation and deceptive growth creates customer and product platforms that can scale quickly and provide useful solutions, which often leverage the network effect. The more people that use a platform, the more beneficial it is to everyone else; Facebook is only useful because everyone else uses it, Waze only succeeded because it could aggregate vehicle info from a growing number of users and send it back to the rest of the users which achieved compound growth. Points of discontinuity from old to new business models are common in the case studies we looked at: Kodak / Instagram, Wikipedia / Encarta, Whatsapp / SMS, Apple / Blackberry and others. I’ve written previously about the lessons we can learn from South Africa’s big disruptors.
Our discussion on AI was very intense and led to questions that people like Max Tegmark and Elon Musk are asking. Can we keep AI beneficial for humanity? Is machine learning taking us on a dangerous path to self-aware cyber-physical systems? This used to be science fiction, but now it’s main stream narrative. Chess programs were around since the 1950s, but were popularised through their success against Gary Kasparov and the Deep Blue tournament. AI robotics expert Alexander Moravec figured out in 1988 already that, “it is easy to make computers exhibit adult level performance on intelligence test or playing checkers, and difficult or impossible to give them the skills of a one-year-old when it comes to perception and mobility” AI research is decades old but is now a hot topic – only because it has become relevant, interesting and disruptive – possibly to the whole of humanity. I’ve written previously about IBM’s Watson and the future of personalised technology.
Google’s Deep Q-Learning AI program was told to beat an old Atari Breakout game but was not given any assistance. When it started playing, it did not even know the object of the game which is to knock out bricks from a wall to earn points; even the concept of a paddle or ball was unknown to the software. Unlike Deep Blue which was taught the rules of chess, Q-Learning was only able to “see” the screen and given a singular outcome to achieve; maximise the score. In his excellent book Life 3.0, Max Tegmark called it one of his “major jaw drops” when he saw how quickly this machine could learn the game and then beat it. It took about 10 minutes of basic training, 2 hours to become an expert and after 4 hours of training, it devised a score-maximising strategy that even the human programmers had not figured out – check the video!
GREY’s team of creative experts wanted to know how soon machines would be achieving human level intelligence in how we think, respond and learn in new situations. Max Tegmark puts it best when he reflected on the Atari match, “There was a human-like feature to this that I found somewhat unsettling: I was watching an AI that had a goal and learned to get better at achieving it, eventually outperforming its creators. DeepMind’s AI was growing more intelligent in front of my eyes (albeit merely in the very narrow sense of playing particular game).” Since then Google DeepMind also beat Korean legend Lee Sedol 4-1 at the notoriously complex game of Go in 2016. This was much sooner than anticipated and really pushed AI into the headlines where it has also attracted attention and questions about jobs, wars and medicine. Many of these topics Tegmark addresses in his book in which he also talks about his pioneering work on “Beneficial AI” and ensuring it becomes a positive force for the good of humanity.
After his controversial battle against the machines, Garry Kasparov provides some hope when he expertly reworded the Moravec paradox, “we are least aware of what our minds do best”. Hubert Dreyfus called it “thoughtless mastery” – the human mind’s ability to learn, intuit and apply without deliberate thought. We also know some things machines don’t do well, but this list is getting shorter; team work, patient care, sales and creativity for example. The team at GREY advertising agency was particularly pleased their tradecraft of creativity is safe; machines aren’t currently producing brand campaigns – well, not yet.