Ex Machina Revisited — Thoughts about an A.I. as a future marketing tool.

The eternal dream of mankind: creating artificial intelligence. But would a machine with the ability to save practically unlimited amounts of data, abstract and combine their logical relationships, learn from these and thereby improve itself to a nearly godlike mega-brain be a blessing — or a curse?

A few days ago I watched the wonderful movie Ex Machina, which is about exactly this (don’t worry, I won’t spoiler!) — the struggle of man against the A.I. he created. The intention of this movie made me think about the idea of using artificial intelligence not only for proving the capabilities of science (or it’s creator), but also as a tool to increase marketing effectiveness.

How could an A.I. help to increase the effect of marketing activities?

Well, on the first glance every marketer (and maybe most of the customers) would welcome the idea of a constantly self-improving “tool” which helps to identify one’s preferences and tries to propose matching brands and products. By analyzing and combining user’s browsing and searching behavior from many different sources and combining it with the ability to actively learn from this behavior, more relevant recommendations could be made and the chance for success — in this case a discovering a matching product (and performing a buying checkout )— rises significantly.

A marketer’s dream

An intelligent system — an A.I. — would allow to solve the problems created by the systems which are currently in use:

Nearly all recent systems could not decide if a search made in the past is already solved — and continue to offer the same products or links again and again. There is a very good description of this problem in another one of our articles (in one of the lower sections, about “the fridge situation”). An A.I. could pro-actively monitor the searches (and even checkouts) executed by a user and filter these articles from future suggestions.

The proposal of similar (or slightly different) entities is mostly more than weak. Yes, algorithms trying to show “related stuff” are working (for example on Amazon), but their outcome is nothing I’d call straightforward. An A.I. would be able to build and differentiate semantic relations between terms what increased the effect significantly. Let’s use Amazon’s related products functionality for an example: We assume somebody bought swimming shorts, sunglasses and a suitcase. The system will now offer things other users who bought this also ordered, this could be trousers from the same brand or anything other related to holidays (yes, the system is able to perform such combinations at all!). But — and here comes the A.I. — it could go much farther! An intelligent, self-learning system with multiple sources would offer some completely different things — a traveling guide for Hawaii, the best prices for hotels there with the opportunity of directly booking. Why? Because it would know from it’s database this user already searched for a flight to Honolulu. But this is not already the end! The next time the user searched for a skiing suit in winter, the system would recommend hotels in exactly the price range he used to book for his last holiday or a list of nice restaurants according to the ones he visited on Hawaii — because it would be able to build intelligent connections and literally learn from the user’s behavior.

By this straightforward ability to learn based on analytics comparing recent suggestions with real buyouts, it would come closer and closer to the real needs and wishes of the customer what would raise their effectiveness significantly.

Exactly tailored product suggestions — sounds like a dream for marketers and customers, doesn’t it?

The downfall

But what about the disadvantages mentioned in so many SciFi novels and movies? Is the A.I. really a perfectly helpful creation or are there risks as well?

Let’s imagine the A.I. is perfectly created, meaning with full capability of learning independently and without restrictions. Let’s also imagine, to maximize the effectiveness, it has access to not only one, but multiple product and user databases (or — much more interesting — to the complete internet). What would happen?

The main difference between the human and the artificial brain is the access to data kept in memory. Even if we expect some — more metaphysical — theories stating the human brain does not forget anything at all but just limits the access to this amount of information (as Aldous Huxley does in his great book The Gates of Perception), this problem will never occur for machine. It has a practically unlimited amount of information available which it can combine, analyze and call whenever it is necessary.

Additionally, there is another advantage which would raise an A.I.’s digital brain over the biological one: time of activity. The human brain is incessantly challenged by conscious and sub-conscious processes and — not to forget — has only less capabilities of multi-tasking when it is used actively (try to think three complex problems exactly simultaneous) whereas a digital processor (what the A.I. brain is) can handle multiple processes at once. This would allow it not only to execute the tasks actually needed, it could also use the capacity which is currently not in use for continuous learning — in the literal way. The A.I. could constantly monitor itself, notice each mistake and directly correct it once and for all (by the way also a huge advantage evoked by the ability of “never forgetting” anything).

According to the analysis above, the A.I. would reach nearly universal knowledge combined with the ability to constantly self-improve — in my opinion a very dangerous combination. Even if we consider the A.I. not to be capable of feeling any emotion, an existence based on pure logic and combined with the skills already mentioned would be the nemesis of marketing (and maybe also of mankind, but this is another story…).

Nemesis of marketing? Sounds a bit exaggerated…

It does not! When a company used the A.I. for an increase in sales by offering perfectly tailored product suggestions to its customers and the A.I. was programmed to hit exactly the customer’s needs, it would do what it was expected — till it reached a state of “consciousness” leading to the final meltdown for the company’s marketing efforts. Exactly tailored suggestions must not directly mean suggestions of products by this company. The A.I. with access to a practically unlimited amount of product and customer data would recognize there are suggestions fitting much better for the customer — even if they are not offered by this company. It would begin to execute decisions driven by what it learned in the ongoing process of self-improvement and no longer “care” about the needs of the company (by the way a reaction similar to what is described in the already mentioned SciFi books and movies).

Users would be driven away from this company, sales decreased and the whole concept imploded. The company needed to lower prices (as competitors also would do) what would start a continuous down spiral of falling prices and maybe could culminate in a price war having a negative impact on the market and — in a possible next step — on the whole economical performance of one or more countries. When we imagine what happened during the last financial crisis (which was created by “just” the banking sector), I’d say the usage of a “real” A.I. with all its advantages and improvements could be a really critical decision.

Conclusio

It has always been human hubris which led to fatal developments and I expect the creation of a fully functional A.I. as the next step we will take in this direction. Sure, the situation I described may be painted dramatically black, but why opening Pandora’s box when it is not necessary? Recent improvements in marketing strategy and execution offer many possibilities which do not nearly drive us into a direction this risky as working with truly intelligent machines.