Computers were built to take over jobs from people when they were created in the 1950s. Humans saw their silicon protégés as the ultimate tools for mankind.
Recently a lot of headlines hit the mainstream media about AI and the lists of jobs that it may take over. A few stories created existential angst suggesting they might overtake their human creators and take control of the whole show…
Rather than get worked up about it, people should be aware that, right now, software is running much of the world. It already helps run our stock markets, internet, power networks, media feeds, military, communications and so on. And this is an onward march.
Now, with the explosion of connectivity, big data and machine learning, computers are definitely entering a new era of capability, one that is beginning to resemble a human form of intelligence where machines can ‘learn for themselves’.
Machines, like people, can now learn from information and, with so much information and processing power available, they are beginning to address problems we never thought possible.
Deep Blue
One of the biggest public leaps forward in AI was the chess rematch between IBM Deep Blue and Gary Kasparov played in 1997 in New York.
The game of chess has a set of rules, but it is extremely complex and it was thought to be a uniquely human skill. However what AI shows us is that there are very deep and complex patterns underlying most things and that it can understand them. Compared to modelling everyday life, the moves that are possible in chess are very small, however it was still beyond the ability of computers to solve chess or to know how many moves there actually are.
It falls under a class of computing problems known as ‘NP’ problems (nondeterministic polynomial time). They are solvable, but it would take computers a very, very long time to solve. The Shannon number puts the number of moves available in chess at 10 to the power of 120 — which is more moves than atoms in the universe.
Due to the complexity of the game of chess, Kasparov was able to state earlier in his career: “Chess is very safe from computers”. Kasparov won the first encounter with Deep Blue in 1996 but lost games, so by 1997, facing an upgraded Deep Blue in a 6 series game, doubts were beginning to set in. To quote Kasparov before the 1997 rematch: “In one respect, I think I am trying to save the dignity of mankind by playing in this match. In another respect, it is a team of research scientists who created this computer system and they are really my opponents. Let’s see what they've come up with in terms of hardware and software to challenge the power of the human mind.”
When Kasparov lost game 6, Deep Blue took the series by 3 1/2 to 2 1/2 — the first time ever that a machine had beaten the world’s greatest chess player. Looking back now, Kasparov salvaged dignity by taking games from this machine. With sheer number crunching, and without any ‘human intelligence’, Deep Blue was able to beat Kasparov. It did this by using AI to examine over 200 million moves a second and, without being intelligent, crunched the outcomes of moves. With data and processing power, Deep Blue was able to strike a blow for silicon over its carbon creator.
In the future we might look back to this game as the first major sign that computers will bit by bit take over jobs that we thought required human intelligence. No one believed computers would ever beat the best player in the world at a game as complex as chess — so where will all this end?
DeepAgent
Two years ago, IRP Commerce began a research programme to experiment with AI in ecommerce with a project called ‘DeepAgent’. We constructed a machine learning framework and began to work out what we could do with it. What we needed to do was to supply it with information to see if we could make it do something useful — in our case to increase ecommerce sales.
The results were spectacular. With data and AI we realised we could predict who the buyers were on a website and, subsequently, intervene when someone might be a buyer and needed a nudge.
We were able to systematise these machine learning insights to indicate actions that would further increase sales. It was enlightening to see that in all the areas we were looking at — from conversion rates, to automated bidding on PPC, to who should be doing marketing services for which client — DeepAgent was ultimately going to outperform people. By adopting AI technologies like DeepAgent to do the work (as opposed to people), more sales and profits were guaranteed.
As I see it, many of the challenges in ecommerce share a similarity with chess. They are in the ‘NP’ class of computation problems. We cannot be sure if a move is right or wrong because the problem cannot be solved in a time period. However, one move can be a much better move than another — and some people are very good at playing the ecommerce game.
In the end, as with chess, AI machines will ‘play ecommerce’ even better than the best people.
So what is the future of AI in ecommerce? Like most other areas of work, the future will see increasing AI automation and machines taking over what people used to do. From online marketing, to conversion, to email marketing, to linked selling, every major area in Traffic, Conversion and Retention faces automation by AI augmented systems.
Increasingly, people will be there just to keep an eye on things and to oil the AI ecommerce machine.
