6 min read
The opinions expressed by the entrepreneur's contributors are their own.
Recently, the entry into AI (Artificial Intelligence) was not manageable for startups and small companies. It required a highly skilled data scientist and machine learning expert to experiment with algorithms. But in a very short time, things have changed. AI that can recognize objects in images, understand documents and texts, and make highly accurate predictions for your user data can now be done in a matter of hours and without coding.
Related Topics: How Small Businesses Can Use AI to Tackle Bigger Competitors
The same thing that happened when building websites. Before, whenever you needed a website, you always needed a developer. A lot of websites these days are built almost via drag and drop using services like Wix and Squarespace. Large or complicated websites are still created by developers, but for small businesses it only takes a few hours to create a website to choose between templates and move sections around.
The same thing happens with AI. Since most artificial intelligence is almost always based on the same few standard algorithms, automating the AI development process was pretty straightforward, which means that you can now essentially drag and drop the AI. Like the websites, the complicated solutions require experts, but most people can find simple solutions.
Then what is easy? For example, let's say you want to teach an AI how to do quality control on the assembly line so you don't ship products that are ultimately returned. This can actually be achieved using a number of tools for AI that are similar to Wix and Squarespace.
Unsurprisingly, the tools are made by big technologies like Google, Microsoft and Amazon. But there are already a good number of startups trying to do the same.
I really believe that there is very little time before you hear this sentence in the office: "This is a boring task. I will train an AI to do it." Maybe I can make a coffee before it's done. "
Related: How To Use Artificial Intelligence To Change The Way …
The step-by-step plan to get you started
If you're like me, your mind may already be racing with ideas and problems that this sudden, easy access to artificial intelligence can solve. Or you might be wondering where to start. Here is a step-by-step plan to get started.
Get your problem right
The first, and perhaps most important, step is to know exactly what problem you want to solve. It sounds like a truism, but this is often the source of the error when problems arise.
To properly resolve the issue, you should at least do the following:
Describe all possible inputs and outputs. Decide on quality goals. Get domain experts involved. It's easy to make a lot of management decisions, but if you don't involve the operational staff, you risk making AI with no real application
Gather your data
Main piece of advice here: data gathering is usually by far the part of AI that needs the most resources. Some projects only require you to collect data at startup, and some projects require you to collect data repeatedly.
Usually, there is no way of knowing how much data you will need in advance. Because of this, it makes sense to move on to the next step early on and train an AI a few times to see if you can get a good result. And don't make the mistake of becoming perfect. An AI, like any other business system, is a way of solving a problem and when it does, there is no need to invest more.
Make sure the following check boxes are selected when collecting data:
When recurring data collection is required, you need to calculate the cost of data collection to keep the business case positive. Make sure that your data cover all possible inputs as much as possible
Train, test and use
Now that you know your problem and you have data so you're ready to build your AI. As I wrote at the beginning, there are several tools that you can use to create your own AI without coding because you already have the data. This is actually the easy part because with most of these tools all you have to do is upload your data and the click of a button will train and deploy the AI. That's it.
The only thing to watch out for here is that the Quality Score you get on your model may not exactly reflect what you get from the actual rollout. So test it out as often as possible.
look at it
The world has a tendency to keep turning and changing. You may have trained an AI that now works seamlessly, but it could deviate from reality as the world changes. Let's say you've created an AI that can detect credit card fraud. As soon as your AI catches the scam, the criminals change their tactics. It is important to monitor changes to the data to identify and prevent this problem.
Related: How Artificial Intelligence Helps Fight The COVID-19 Pandemic
Is it really that simple?
The hard part is actually the problem and the data. It is a little more difficult than other solutions to get to the core of the problem and know whether or not AI is a correct solution as the field is so new and it rarely happens that someone is very experienced.
As mentioned earlier, the data is also usually the most expensive part that is often overlooked. However, this can turn into a competitive advantage. If you can find a way to collect better quality and / or cheaper data than your competitors, you have a good argument for success.
Connecting your AI models to existing software is still a red tape. IT will always be IT, and AI falls into that category. This is associated with uncertainties. When integrating your AI with other systems, problems can arise that make the effort more difficult than you originally thought.