Low-code solutions for AI are on the rise
As the capabilities of artificial intelligence (AI) continue to grow, engineers are increasingly adopting the technology into their daily lives to help them improve their existing systems.
However, as more people are becoming aware of the appeal of AI, not everyone has the coding experience, or the capability, to understand how to incorporate it into their existing systems. Step-in low-code solutions.
Electronic Specifier’s Sheryl Miles speaks to Johanna Pingel, Product Marketing Manager at MathWorks, about low-code solutions for AI and what this means for engineers, interdisciplinary teamwork and future trends.
What is a low-code solution for AI?
Every industry and application that AI is relevant, low-code can also be relevant.
[If you want] to figure something out without getting caught up in the nuances of the code, that is where low-code shines because you can copy what someone has done, and then use it for your application. It can take you from the very beginning to the middle of the pack, and [allow you to] get the ball rolling from there.
Low-code is trying to level the playing field for everyone, regardless of their experience. The point and click aspect helps guide you to the right code that you can learn from, and then bring into your applications.
How does low-code facilitate effective work in an interdisciplinary team?
Interdisciplinary work is an extra benefit of low-code because anytime you have two distinct groups of people, you have to find a way to connect them.
You could have data scientists working on the model, and then engineers who need to use it. Low-code can be the bridge taking the model from all of the code, [and converting it to] something that engineers are more comfortable with, [such as] a point and click or drag and drop solution.
What's coming out as a new trend is that data scientists are publishing their models on GitHub [for example], and engineers can pull [the model] down from any language and incorporate it in a low-code way.
In what areas of the AI workflow can low-code solutions be utilised?
The AI workflow:
You've got four parts of the AI workflow. You start with data, then you move on to modelling. Then you've got testing. And finally, you have deployment. Within every one of those steps, low-code can help.
Data preparation is probably the most obvious … people spend a significant amount of time on data preparation, and low-code can help increase productivity. For example, you may have to crop your images or shorten your signals, all of which can be done within a [low-code] point and click environment.
Then you've got modelling … there are great drag and drop tools out there, [and they] allow you to create your model, connect all the layers, test the model and see whether it's going to work for your application. All within a low-code environment.
How can low-code help engineers?
[Low-code can] free up engineers to focus on the more challenging aspects of their work. If they're not focused on the coding aspect of it, what more could they do? The benefit for engineers [is that low-code can allow them] to problem solve more.
MathWorks has been in the low-code game since before low-code has been a thing. We [strive to] help engineers and scientists do their best work, and that means freeing them up from coding. MathWorks and MATLAB are always trying to abstract the complexity of the language into something that's very easy to understand and very intuitive for engineers and scientists.
MathWorks also recognises that software, or code in general, is still essential to business. We’re not trying to get rid of all coding. It's about being able to free up the engineers to do their best work.
We don’t use the term no code because there is going to be code, and the final solution will be heavily code intensive. We're not creating a [no code] world, we’re just making the process easier.
Where do MathWorks see the future of low-code AI?
I think the future, for me, is when everyone who wants to incorporate AI is able to do so, regardless of their coding skills. And in some ways, we’re already there. There are tools for every part of the AI workflow that can help from start to finish without having to worry about coding.
That means every engineer can be given the tools to best help him or her succeed with the level of coding that they feel more comfortable with. And it allows people who perhaps used to code [but no longer have the time] to stay involved. [Low-code] helps everyone stay in the game.
Not all low code tools are created equal. You need to have a tool that best fits what you are trying to do ... If it doesn't allow you to incorporate or explore all the different options, then you're going to be limited. You want to make sure that you're creating or using tools that allow you to expand your universe, rather than keep you in a box.
That is where MathWorks comes into play because we speak engineer. We speak with signal processing experts, and we have signal processing experts. So, we understand and know what they're looking for in a solution.
The work is real, and we have customers [such as DRASS and PathPartner] who are using our platform. Low-code is not the headliner, but it is the supporting cast.