A LIMS and Artificial Intelligence (AI) have similar issues in their implementation. It comes down to the issues we have with implementing any software.
Background on This Topic
Earlier this year, I’d been working to take high-level courses and workshops on topics such as AI/ML (artificial intelligence/machine learning). I picked some topics because I just find them interesting. However, there are often enough similar issues between what we do in our industry and these other topics that you’d be surprised what you learn from these topics that might seem unrelated, at first. Taking what seems like unrelated courses can help give perspective on what we’re doing in our industry.
Side note: This is where I took a class on how storytelling helps you with your blog posts. This is where my posts got out-of-control with the stories I included. I had to redo all the posts I made with the new storytelling approach. The other related topic where the storytelling got out-of-hand being the one on CyberSecurity and Ethical Hacking.
In any case, the issues of programming LIMS and Artificial Intelligence share implementation issues. This is true all the way from finding out what the user wants and needs all the way down to ensuring that everything works and that the user finds value in what you’ve provided them. And I will try to tell this without as much of the super-long and incredibly boring story that originally took over this blog post.
First, My Smart and Stupid Home
I don’t have a “smart home.” However, some portions of it do include some automation and AI/ML. Mainly, I’m talking about my entertainment system.
Let me draw a more specific picture. My entertainment system takes up more of my network and has more purchased equipment and software than my office does. Some portions of it run both locally and in the cloud.
Let’s label me – I’m the user, here, and I want yet more than what I already have. This isn’t atypical of the user in any system.
In any case, between all the free and purchased software/apps, and all the equipment running, there are just a few features I’ve wanted to have but that I’m lacking. Everything I have basically works well together and has lots of features and power. I shouldn’t be complaining about the relatively small number of things that it DOESN’T do. After all, some of the apps make truly insightful guesses about what I want. You’d think I’d just appreciate that rather than focusing on its shortcomings.
I hear a lot of snickers, now, from some of you reading this. Because you also know that’s not human nature, for the most part. Even in our LIMS/ELN/LES systems, we bemoan what we don’t have in them, no matter how much we DO have. It’s human nature to pay attention to the gaps.
Still, Why Can’t I Have it All?
There’s one major problem. The features I lack in my system are tasks that are easy for me to do but that I find to be tedious. Doesn’t this sound familiar? Think about your LIMS and the last time you asked for a feature that was easy and tedious for you but where the LIMS still couldn’t take over that task.
Unfortunately, whether in a LIMS or some AI system, that’s NOT the criteria for an easy feature to add.
Explaining Why Things Are Harder Than We’d Expect
Customers occasionally tell me they don’t understand why something is so hard to provide to them. Unfortunately, when we’re implementing software, there are certain elements we can provide more easily than others. In this, it doesn’t matter if it’s an LIS system or an AI/ML app.
When we talk about these “machines” we’re talking about computers. So, it all comes down to some amount of programming, in case you’re not seeing what these all have in common.
In any case, regardless what we’re doing, the tasks that are easy for people to do aren’t necessarily the ones that are easy for machines to do. But, as humans, because certain things are easy for us to do, we think they’re easy to make the “machine” do, as well. That’s just not true.
Illustrations of the Issues
While human beings are fine at calculating things by typing into a spreadsheet, they can’t process thousands of transactions per minute, for example. This is an area where machines clearly have the advantage.
However, where dexterity is required, there isn’t yet a machine that is as good as a human being with fine manipulations.
So, for example, it’s currently too difficult and/or expensive to automate a toothbrush. But what IS available is a toothbrush that you do still have to move around on your own, but that will ensure you’re brushing long-enough, not missing spots and using the right amount of pressure. Now, the timer and pressure sensor don’t seem to be that big a deal over the older electric toothbrushes that basically already do all that, but now as these machines add the feature of also tracking the zones of your mouth now begins to seem a bit complex.
Still, even though every mouth might be different, mouths are very specific things. For a machine (in this case, the toothbrush) to “learn” the landscape of your mouth is a fixed situation. It’s amazing but still quite limited.
Yet, despite this, the machine is still not able to actually do the brushing for you. Now that it can figure out where your teeth and gums are at, you might think actually doing the brushing SHOULD be the next step. Unfortunately, that requires dexterity, which is what you have and the machine does not. And, even if it’s technically possible, it’s probably too expensive to be practical in time for your next brushing session.
My Point in Discussing LIMS and Artificial Intelligence
We need to think differently about the way we build software than we do about our manual processes. This is true for both LIMS and Artificial Intelligence.
For example, for many, many years, we’ve talked about the fact that, in a “paperless lab,”the challenge is that you’re not just replacing a piece of paper. A paperless lab isn’t one that replicates the lab you had with paper – it’s an entirely different system made up of entirely different processes.
Thus, this is the challenge we all have on our projects. It is to come up with new processes that make use of whatever new technology we have while not trying to make them do things as we humans would have done them.
Let me give one more example – have you watched a robot perform on a manufacturing line? If you have, even if it does a repetitive piece of work that a human previously did, it does not do it in the same way. A human being doesn’t rotate their arms or fingers the way the robot does. We don’t move our arms vertically or horizontally the way the robot does to move to a new position.
Likewise, in our software, the way the software presents the screens to us isn’t always the way we approached the work when we were recording it without the software.
The Digital Transformation
Remember that the digital transformation doesn’t replace paper, but gives us new processes. It’s not about a specific tool or even a specific type of tool. That applies whether the tool is a piece of hardware (such as a robot arm) or software (such as a LIMS or AI/ML code).
My point is that software is not the same as the human brain. Our machines are not the same as the human body. Regardless where we apply this, it’s the same. This is true whether it’s for LIMS/ELN/LIS or for AI/ML.