Do you usually use artificial intelligence (AI)? If you answer no to this question, what follows here may surprise you. You probably use AI daily, even if this happens without your knowledge. Every time you use a fingerprint or iris scan to unlock your cellphone you use AI. Every time you use Facebook or Instagram, the advertising that you see is adapted using AI. Every time you use Google Translate, the text is translated with the help of AI. Every time you look at books online, the recommendations are based on AI, and when you press the order button your credit card is checked with AI to avoid potential fraud. And so on.
Digitalization has been a common topic of discussion in recent years. What is going to dominate discussion in the coming years is AI. It is changing everything. Right now, we are probably witnessing the most dramatic technological shift that we have seen since the rise of the Internet.
AI is the most dramatic thing that has happened in computer science in 40 years. | Stefan Carlsson, Professor in computer science, KTH Royal Institute of Technology
To answer the question, “What is AI” we must start by defining intelligence. The AI researcher Max Tegmark prefers a broad definition: “the ability to achieve complex goals.”1 If we accept this definition, AI is simply the ability to achieve complex goals in an artificial way. Unlike our biological intelligence, it is usually “computers” that have this intelligence.
One way of looking at it can be that AI received input (for example data or information from its surroundings) and then interprets and learns from it, to achieve specific goals.
Mayor components of the technology behind today’s AI solutions are not new. The world’s first neutral network was actually launched in 1959, to eliminate the echo from telephone lines. In the following decades experimentation continued but no great breakthroughs were achieved – until now. Unlike earlier systems that were rule and statistics based, modern AI solutions build upon machine learning. This concept means that computers can learn and, after the fact, improve their abilities. It is about letting computers train, so that they get better and better.
The purpose of machine learning is, explained simply, to create systems that are more human-like in their ability to identify and understand the world. It involves pictures, text and other things that historically have been easy for people to manage but difficult for computers. Computers have been better than people at certain things for a long time: particularly calculations. Multiplying 2,539,342 by 4,393,245 is ridiculously simple for a computer but complicated for a human. Conversely, a human finds it easy to recognize, for example, a dog in a picture, while it has been traditionally very difficult to perform this type of analysis with a computer.
Currently, the development of AI is going very fast in some industries, where actors want to position themselves and benefit from the new technology. In sales and marketing, we are already seeing many concrete tools based on AI and machine learning. Chat bots, virtual assistants and many other sales tools utilize AI to create powerful solutions. Over time, the performance of these solutions can also improve as a result of algorithms getting better or when more data to train with becomes available. One of these areas is presented in other chapters in this book, but let us look at a couple of examples now.
Phrasee is an AI solution that composes powerful headings for email campaigns. This can sound trivial, but if one can get a couple more percentages of recipients to open an email this can mean big money. Or why not write an entire sales text with AI? Using AI, the Chinese e-commerce company Alibaba can create 20,000 lines of advertising text in one second, something that several Chinese garment companies have already begun to use.2 Another ingenious solution comes from Cogito, which can analyze a sales conversation in real time and provide feedback to the sales rep. Is the sales rep talking too fast? Does the sales rep interrupt the customer? And how are the levels of energy and empathy in the conversation? Things like this are rated so that the sales rep can make immediate adjustments.
One can even use AI to manage leads and sort through incoming email. The telephone operator CenturyLink in the USA, which has both small and large companies as customers, recently invested in such a system. They call their virtual sales assistant Angie. This AI-based software not only sends out about 30,000 email messages every month, but also receives and manages the responses.3 Angie identifies the leads that have the greatest potential and can also determine which sales rep is best equipped to contact the customer. In their first pilot test, Angie was able to interpret 99 percent of the incoming emails. The last percentage was forwarded for manual analysis. Angie shakes out 40 qualified leads per week and the investment has been returned many times – every invested dollar has returned 20.
With the introduction of new technology, business logic and relationships between actors often change. Barney Pell, who is an American AI researcher and entrepreneur, has observed the dynamics that occur when an actor begins to utilize AI, which he summarizes as “Pell’s law of AI lock-in.”4 The reasoning goes like this: Just about every industry will experiment with new AI-based products and services. Those that become commercial successes will be highly motivated to invest more in AI development. This, in turn, means that more AI-based products and services will be on the market, which forces competitors to act. This creates a lock-in effect, where no actor in the long run will be able to resist.
This has already happened in many industries. For example, in computer games and search engines one cannot compete without world-class AI solutions. And this is spreading fast to other industries like finance, healthcare, logistics, the automotive industry, and education. As a result, companies need to improve their understanding of the central purpose of AI. It is about learning early on what the possibilities and limitations are, daring to experiment and test, and later implement and launch the solutions. Those who are early adopters can get ahead of the pack.
References
1 Tegmark, M. (2017). Life 3.0: being human in the age of artificial intelligence. New York: Alfred A. Knopf.
2 Low, A. (2018, 4 July). Alibaba’s new AI can generate 20,000 lines of copy in a second. [blog post]. Downloaded 2018-10-22 from https://www.cnet.com/news/alibabas-new-ai-can-generated-20000-lines-of-copy-in-a-second/
3 Power, B. (2017, 12 June). How AI is streamlining marketing and sales. Harvard Business Review. Available: https://hbr.org/2017/06/how-ai-is-streamlining-marketing-and-sales
4 Azhar, A. (2015, 17 June). The six accelerants of the AI boom. [blog post]. Downloaded 2018-10-22 from: https://medium.com/the-exponential-digest/the-six-accelerants-of-the-ai-boom-903e5b5a0719
|
Henrik Larsson-Broman
|
Share this article:
