Paying for AI? Make sure you’re getting AI

Some companies claiming to offer AI are failing to provide even the most basic forms of artificial intelligence.

False claims about the AI capability of their products are rife among analytics providers. According to London-based investment firm MMC, 40% of European AI startups don’t use any AI at all. Writing for MIT Sloan, Michael Wade, Amit Joshi, Mark J. Greeven, Robert Hooijberg and Shlomo Ben-Hur explain how to tell the difference between true AI and inferior technologies.

IS IT REALLY AI?

AI – artificial intelligence – is “any computer-based system that observes, analyses, and learns”. As it gathers and analyses data, true AI becomes more accurate, independent of human intervention. Be aware of these common misconceptions about what counts as AI.

1. 1. Algorithms aren’t necessarily AI technology. Unlike AI, algorithms follow a set of predefined steps or rules to solve a problem. The same inputs consistently produce the same outputs – it’s not learning, so it’s not AI. Algorithms also require the input of structured data, whereas true AI can process data that is not well structured, well defined, or even numeric, and learn from it.
   2. A system’s ability to answer questions doesn’t make it AI. Many technologies can answer questions. Decision support systems, for example, answer questions by either matching each enquiry to a database of predetermined answers, or, by applying an algorithm to the data input. Only AI can place these questions in context and learn from past answers.

WHAT DOES TRUE AI DO?

1. 1. Observes. True AI observes its surroundings, and continuously collects and collates information to augment its knowledge base almost in real time. Tesla cars use their 21 sensors to observe the car’s surroundings and relay real-time information to the vehicle’s autonomous driving system which grows more knowledgeable with use.
   2. Analyses. AI analyses and identifies signals within the data it collects, regardless of how noisy that data is. Tesla cars analyse the data from their surroundings to automate multiple driving decisions simultaneously. Gong.io, an AI-based revenue intelligence platform for business-to-business sales teams, analyses tone and voice sentiment within sales calls. By identifying and analysing the most effective sales techniques, everyone in the sales team has the opportunity to learn from the most successful telemarketers.
   3. Learns. Most importantly, AI continuously tests, learns and improves. By proactively creating and testing hypotheses, and then learning from the results, AI becomes more accurate over time.

Tesla’s self-driving technology analyses data from thousands of vehicles across its entire fleet. Its cars learn from each other, becoming smarter with every mile. Similarly, Netflix’s recommendation system becomes more accurate as it learns users’ viewing histories, enabling more personalised recommendations.

The actual number of cases where AI is significantly better than simple data science is quite low, but if you do decide to invest in machine learning, make sure that AI is what you’re getting.