This article is inspired by a talk “The AI Manifesto” by Szymon Łukasik, presented during the Business Leaders meetup.
How to get rich in 7 steps
- Invent a startup name that includes “Artificial Intelligence”. Just the name, not a product or service.
- Spread the word about it among the top AI influencers in the world without telling them what the startup does exactly.
- Create some hype in social media.
- Throw a launch party in a rented Airbnb apartment, total cost: $496 (excluding the fine for disturbing the peace in the usually quiet area).
- Wait until major investors from Silicon Valley start throwing money at you.
- Tell them the whole idea was just a joke.
Sound too good to be true? The bad news is that someone has already done it – a startup called RocketAI. Just like that, they took advantage of the most hyped technology in recent years. They created a completely fake company that made investors see nothing but dollar signs.
You could say that the lesson from this story is to stay away from technological buzzwords, but before you do that, let’s explore the topic from a different perspective.
Artificial intelligence – dream big, start small
Before Artificial Intelligence became a startup goldmine, there was a time when thinking machines were romantic – or catastrophic – visions in literary and cinema fiction. This motif has been present in fiction for centuries, waiting until people caught up with their technological capabilities. The time for the first boom came in the 1950’s.
Meet the Farrington Automatic Address Reading Machine, built in 1953. Source: Smithsonian Institution (Flickr)
The term Artificial Intelligence was invented in 1956 by professor John McCarthy at the University of Dartmouth. It was included in the name of his summer college about cybernetics and so-called thinking machines: Dartmouth Summer Research Project on Artificial Intelligence.
In the times of John McCarthy, this area of expertise was called Machine Intelligence. When the term Artificial Intelligence emerged, the topic raised great curiosity and marked the first time that AI took over the popular imagination.
Five o’clock tea with a robot
To fully understand the concept of Artificial Intelligence, we need to go back a few years. In 1950, Alan Turing, an outstanding computer scientist, developed a test that has become a symbol of AI even for non-techies.
The test was supposed to determine whether a machine could be identified as intelligent. It consisted of a text conversation between a human “judge”, a couple of other human participants and a machine. During the test, the judge engaged in a conversation with all the participants, evaluated their responses and tried to determine which one was the so-called intelligent machine. If the judge wasn’t able to tell, the machine passed.
Of course, the Turing test needed significant improvements. The simple fact that the judges knew about the machine taking part in the conversation made them seek the symptoms of non-human communication more carefully. It was extremely difficult to perform the test in an isolated environment, without significant limitations, like the topic of the conversation.
There’s a long road ahead of us
In the 1950s, when AI started attracting significant interest, it seemed we would be able to develop machines that talked just like humans within the next 20 years. Unfortunately, investors would have to watch as much more time passed before such goals were achieved. At that time, the computer performance necessary to create such machines was out of reach.
The number of backers for AI research was reduced to a small group of enthusiasts. They believed that the goals put before AI were achievable, but not at the current level of technological development.
“What the lever was to the arm, the computer is to the brain.”
(Charles Philip Lecht)
Alan Turing truly believed in the potential of intelligent machines. He predicted that by the year 2000, a machine of an “astonishing” 100 MB memory would be able to fool more than 30% of judges in the test. It turns out that he wasn’t that far from the truth. Such a result was noted in a test that took place in 2014 in Reading, England. This event has affected the perception of the “intellectual” potential of machines forever.
Walking in a Winter Wonderland
Public interest, investments and development in Artificial Intelligence as an area of research has fluctuated, just like other areas of tech. After the boom in the 1950s’, Artificial Intelligence had waned by the end of the 1970s. However, not all faith was lost.
When it turned out intelligent machines required much more advanced technological conditions than were available at that time, the primary concept of AI moved towards computational (or collective) intelligence. It was still far from the dreams and ambitions that had risen in the previous decades but presented many possibilities for supporting human activity by cooperation with people.
Computational intelligence was used for solving simple problems based on the mechanisms of prediction or approximation, such as diagnosing diseases or suggesting products to purchase. Machines of such capabilities were supposed to automate processes and deliver results as accurate as those based on human work.
“Our ability to collect and store data is increasing at a faster rate than our ability to analyze it.”
(Keim et al., 2007)
One problem with Artificial Intelligence and its progress was that the scientists’ overly ambitious concepts were impossible to bring to reality. From the1970s to the 1990s, the development of AI technologies slowed down significantly. It was only after this long break that we started developing real intellectual capabilities in machines.
Using basic algorithms, e.g. network learning, we entered the era of Artificial Intelligence that had been anticipated for years. Together with AI, new areas to discover have emerged: Deep Learning, Big Data, etc. All of them work together and help each other grow.
Data – the oil of the 21st century
In today’s world, we use Artificial Intelligence for processing huge amounts of information. The fuel for these activities is… data. As a global resource, you may compare data to oil. In a way, it’s even more powerful than oil because you can’t run out of it. The amount of data we’re producing is enormous and growing exponentially.
But if you don’t have an engine to burn all these resources and use the energy they hold, what’s the point in gathering them? To create a working AI structure that turns data into desired results, you need to:
- Gather data. Most algorithms need a learning source. A recent report by McKinsey implies that effective implementation of AI requires a solid foundation. This includes access to large databases, depending on the digitization of data inputs.
- Discover the potential behind your data. Storing zettabytes of data without putting some thought into the way you want to use it is pointless. For example, listing store locations doesn’t give you much in-depth knowledge. But when you analyze the number of visitors for each store, the combination of insights can help you make better decisions.
- Build algorithms. Recently, algorithms have developed significantly due to the renaissance we’re observing in deep learning and data science. Deep learning uses neural networks with multiple layers supposed to imitate the human ones. Thanks to advanced algorithms that enable these systems to learn, we are able to tackle problems which until now were unachievable for machines, such as voice and image recognition or semantic analysis.
- Collect funds. Developing AI structure demands a high budget. That’s because it’s a relatively new professional path in the IT industry and the costs of hiring a data scientist should be treated as an investment. Without it, Artificial Intelligence will remain out of reach.
It’s happening in front of our eyes
To give you an example of the capabilities of modern Artificial Intelligence, take a look at the illustrations below. The neural network behind it can change the horse in the picture into a zebra or apples into oranges, and vice versa. Imagine a Gauguin painting reprocessed using open-source algorithms so that it looks exactly like a Van Gogh! And this is just one narrow field of research regarding visual processing…
Source: Zhu et al., Unpaired Image-to-Image Translation Using Cycle-Consistent Adversarial Networks, 2017
I realize the overview of Artificial Intelligence’s progress in history may fool you into thinking that the process may still take some time. However, the recent business data reveals an unbelievable speed of change. According to the Vanson Bourne research from October 2017, supported by interviews with VPs of $50mln annual revenue, AI is already transforming the business landscape:
- 80% of companies have implemented some form of AI.
- 30% of companies are planning to increase their investments in solutions based on AI within the next 36 months.
- Improving customer satisfaction with delivered services was named as the basic motivation to use AI solutions.
- The main barriers to implementation of AI are the lack of appropriate IT infrastructure and experts who can help with the implementation. The least-mentioned reason was bad experience with AI-based solutions or not being convinced of an application to the business.
- The leading industries in implementing AI are telecom, finance and online businesses.
There is no other choice but pack your bags and get on that AI train!
Don’t let the case of RocketAI scare you away – Artificial Intelligence has never been a joke. These days it’s more powerful than ever. In the business reality, AI helps to read between the lines of customer decisions and motivations. It lets marketers create more immersive and delightful customer experiences. Brace yourselves, AI is coming.