AI-Myths that are holding you back from creating an AI project
There are a million and one myths about artificial intelligence that make people believe that the AI world is like the Iron Throne, unattainable.
Maybe it’s about not being able to understand everything you need in order to succeed. Maybe you think your business idea is not sophisticated enough to use an AI tech. In this post, we’ll unravel some of the most commonly heard AI-myths, it’s time to face them…
1. AI is too new
Even if AI is not always easy peasy, we have to accept it, it has been around for decades. John McCarthy, recognized as the father of AI, used the term “Artificial Intelligence” for the very first time in 1956, which he would define as “the science and engineering of making intelligent machines”. The concept is not new in our world; humanity has made great progress on it, such as the image and voice recognition evolution or high performance searching. This doesn’t mean that everything that can be done with AI has been done, the hard problems have not been solved yet and big challenges are in front of us; remember, a computer that can pass the Turing test and prove that a machine can imitate the sentient behavior of a human doesn’t exist yet.
2. AI is only about maths and algorithms
AI applications can be powerful and very simple at the same time. We tend to think AI is about heavy maths and algorithms but in fact, AI can be defined as a data play. Indeed, sometimes it is better to use a simple predictive model for a machine learning process. Why? Because using an extremely complex model based only on one group of data can decrease the predict performance for the rest of it (overfitting). For instance, if you want to predict the behavior of a customer group using their credit card bill, you need to use a model that can work for any group and not only for one fraction of the clients. Remember, a data expert spends a small fraction of his time doing modelling and optimisation. But they need to focus on the value of data instead - this is the most important task.
3. Artificial Intelligence is only a nerdy word
Okay, so maybe pop culture is not helping that much, but we can’t think that every AI project team is going to look like a classic Big Bang Theory scene. Even if some people may be apprehensive about whether they would fit in this technological world, they need to know that in general, tech people don’t look like Silicon Valley actors. Besides, an AI project is not only about tech, it’s also about creativity, analysis and learning how to solve problems. For instance, Stanhope says companies will need marketers to input the necessary data and monitor the projects in order to meet the right KPIs, a content team to produce enough variants to test and another bunch of people to get involved in this very symbiotic human-computer relationship.
4. AI is like magic: it can solve everything by itself
AI is not magic; different situations demand different AI solutions. Therefore, if your data and situation are not well thought out, the solution proposed and results can be an absolute failure. Without a real analysis of your original data, you cannot expect the solution to figure out itself what you really wish. Even if you are using a machine learning process, you need a minimal understanding of the problem you want to solve. Moreover, you can’t pretend that the solution will clean a prejudiced source data for you. As a result, you need to work on your data pool before beginning the project. Last but not least, AI solutions are created by humans. So, human intervention is mandatory. You can eliminate human involvement with some steps, but you can’t pretend that AI will work without human help.
5.Only big companies can implement an AI project
This is a blatant lie. No one should be blocked by the fear, “Is my business big enough to introduce AI project ?” Maybe your enterprise is little, but that doesn’t mean you don’t have an important data pool. Indeed, machine learning is used in cases where projects have big data sets, even if the enterprise has 5 employees. The data sets are not directly related to your turnover. If your data project is not big enough, you can find a solution by trying to augment it with public or purchased data or by creating an MVP which could generate the data you need.
