The recent news that Facebook temporarily suspended its Artificial Intelligence (AI) program when the bots within this program started chatting to each other in a strange language only they understood, has reignited the debate on the potential dangers of AI. Several media outlets have jumped on this news with predictions that artificial intelligence will become independent and pose a threat to mankind.
In reality, however, the development was more of a human error rather than any kind of AI breakthrough. Furthermore, this event has several implications for many facets of business. To begin with, what actually transpired was that the AI bots had been programmed to learn negotiation techniques, and were being rewarded on how quickly they achieved the best solution for themselves. However, the programmers forgot to tell the bots to restrict their negotiations to human language (specifically, English!). Due to the incentives in the program itself, the bots quickly realized that using colloquial spoken English was wasteful, and (since their incentives were designed in that manner), started to use shorthand to communicate. Indeed, Facebook stopped the experiment only because they realized that they had neglected to instruct the bots to communicate in proper English.
For businesses, this development is indeed scary, but not for the reasons portrayed by media. Once Facebook (or Google, or Amazon, or some lesser known AI or Machine Learning (ML) start-up, such as MindMeld) cracks the art of bot negotiation, any function within business that relies in interpersonal arbitration could potentially be replaced with an AI bot. For instance, consider pricing for standard industrial products. It is not too hard to imagine innovative firms implementing bot-based negotiations for purchase orders and even for pricing of B2B products.
Or think about the allocation and buying of advertising media. With the explosion of online and offline channel options, media purchase and negotiation is no longer feasibly handled by product managers or Chief Marketing Officers. The best firms already use fairly advanced data analytics to decide on media spends, and its allocation across channels, using sophisticated bots to create optimal media plans, allocate resources and even bid for the lowest rates with providers. While current bots are very good at managing the bidding process, ML-enabled bots can also handle unstructured interpersonal negotiations, such as personal selling. Thus, the entire sales process may get revamped (or at least augmented) with such technologies.
At a consumer level, one can envision mobile phone-based bots that bargain with businesses on our behalf (based on preferences that we set) to obtain the best deals on consumer products, manage the use of our private data (and ensure that we get some tangible return for sharing data), and perhaps even negotiate with us (“Why don’t you walk back home today and then maybe you can pick up the ice-cream that you really want?”).
While most firms have made at least some investments in these areas, AI and ML are still in a very early stage of the adoption cycle. The five factors suggested in Roger’s theory of Innovation Diffusion, provide a mixed message on the adoption rate and speed for these technologies – while the Relative Advantages can be extremely significant, and Compatibility with existing systems may be achieved through a phased implementation, the high levels of Complexity combined with low levels of Trialability and Observability provide significant challenges.
For these reasons, we may see a slow acceptance of these technologies among firms. Quite possibly, large firms may develop such systems, but delay implementation until the slew of concerns have been addressed. Nonetheless, the emergence of such systems feels inevitable. The overwhelming competitive advantage offered may ultimately overshadow all other fears and anxieties associated with AI and ML.
Amit Joshi is Professor of Digital Marketing and Strategy at IMD business school.