My name is Dan Rose Johansen and I’ve got the pleasure of guest blogging an article for Acubiz about artificial intelligence. I’m CEO and founder of the company TodAi. I’ve established TodAi to make artificial intelligence down to earth and accessible to as many companies in Denmark as possible. I’ll try to make the concept tangible and applicable in this article. Happy reading!
Artificial intelligence, or AI as I call it, is a diffuse topic for many. But it doesn’t have to be. It’s possible to make the concept both tangible and applicable and in that way make it much more appealing for companies and its employees.
One of the big challenges with broadening AI is the natural resistance towards change. Whichever way you look at it, AI is a subject of a paradigm shift within many functions in many industries. The opportunities offered by AI, and for that matter also Robotics and Machine Learning, will in the future lead to a different workforce. A workforce that will change and adapt. Employees will not lose their job, but job functions and tasks will change.
Therefore, it’s important that companies try to embrace AI’s many opportunities to remain competitive in the future – and it’s not at all as dangerous as many people think.
We don’t agree on what AI is all about
The biggest hurdle I experience with the aforementioned “resistance” to AI is that we’re all a little confused when AI is the topic of conversation. We don’t have a common language yet. The news feed can be a part of the hurdle. If we take a closer look at the news about AI, 12% of the news are about Elon Musk and self-driving cars. But AI is so much more than that. I’ll get to that in a moment.
We don’t yet have a common language, or understanding, of AI. Therefore, many companies are saying that they use AI, but in reality they use a whole different technology. Many people automatically believe that AI is in play if a process is automated. That’s just not always the case. In fact, it turns out that only 40 % of Danish startups make use of AI. I’m convinced that this is due to a misunderstanding of the concept. The reason for the misunderstanding will most often be because AI is confused with robotics. I understand the confusion, and it’s therefore important to emphasize that the big difference between the two concepts is that robotics can’t solve a task that it hasn’t seen before. But AI can! In other words, AI is experimental and basically mimics human actions. Robotics, on the other hand, can only act on the instructions received.
Loss of control is the last important factor when talking about reticence with using AI. If a company choose to use AI, the feeling of loss of control will inevitably occur but it’s not as extreme as it sounds. I’ll get to that in a moment.
What’s the difficult part of working with AI?
When I started working with AI, I thought that the easy part was to find a problem suitable to be solved with AI. This turned out to be the difficult part. Actually, what I thought was the difficult parts has proven to be the easy parts of AI. The easy parts is to build AI models, make prototypes and get ideas for what can be solved using AI.
The difficult part is to understand the essence of the problem you’re trying to solve and then solve it using all the necessary data. Because it’s usually both expensive and cumbersome to collect data when an AI task has to be solved – even if the task, as a starting point, seems simple.
Don’t be scared – this doesn’t mean that AI is so difficult that it’s impossible to get started with. Unlike a few years ago, it’s possible to build an AI model without any kind of code work. And as more and more AI models are developed to solve an ever-widening range of problems, more of the models will become so-called off-the-shelf products that you can just peel down, buy and use. Easy and convenient – plug and play.
What are the future with AI going to look like?
I know that many people are discussing how the future will look like with a more mature market for AI. When AI was introduced, there were a focus on whether the robots would take over our jobs – it naturally created part of the resistance that I initially touched upon. Neither robotics nor AI will take our jobs. I’m absolutely convinced of that. But our jobs are changing.
Let’s use an accounting department as an example. If an employee today spends 80% of his/her time on doing entries in an ERP system and only 20% on controlling, the distribution will reverse itself in the future. In the long run, it will be possible to get AI to handle tasks that employees do manually today. It may be via an off-the-shelf product, but even an off-the-shelf product must be developed over time in order to function optimally. Here, AI is significant different from robotics. AI is making its choices based on experience. In other words, you need to “train” your AI so that it can solve the tasks you want it to solve.
You train your AI by showing it examples of what a “real” employee in a specific job situation would do. The AI will then register all the factors and variables that were present in the example and then learn from it.
An employee in a finance department can for example train an AI by showing what an invoice looks like. The AI will then use its preconception to assess whether it’s an invoice or not. If an off-the-shelf product is used, it will already have a pre-defined understanding and it’s possible to further develop the understanding by training the AI. The employee thus tries to fill in knowledge in the AI, which may be relevant to whether the given task or problem can be solved.
When the employee is training the AI it’s important to give it tasks of varying difficulty. It can be a curly receipt from which the information is to be read. If the AI that is used to read the receipt reports back, that it’s 90% sure what’s written on the receipt, then the employee can try to train the AI further by putting data in that can help the AI to become even more certain next time it has to read a receipt. If the AI can’t be trained further, then this might be one of the tasks that the AI passes on to a “real” employee who can perform the necessary controlling.
How to get started with AI
If you think that AI is exciting but aren’t quite comfortable with letting a piece of self-thinking software handle work tasks in your company, then my best advice is that you start with an emailbot. An emailbot is an AI that can learn to understand, sort, forward and reply to emails. It allows you to test the technology and see if it works in your organization.
AI will be a competitive parameter in the future. Therefore, you might as well start considering what work tasks AI can solve for you. And remember that it’s not difficult to find ideas to where AI or robotics can solve some of the “boring” work, but the difficulty lies in figuring out what the boring work actually consists of. One thing is for sure – if you spend the time it takes to integrate AI into your business, you’ll be able to benefit from it for many years to come.
