AI in manufacturing will not solve all of your problems. It’s true. AI is touted as a new technology that will solve all of the world’s problems, and while it will no doubt solve a lot of problems, there are more advancements to be made. Especially in the world of manufacturing.
The thing that no one tells you is that the company and AI must work together. You cannot use AI in manufacturing effectively without the help of people on the floor. It’s a 2-way street.
AI Can’t Do Everything
One of the biggest issues is that people think AI can do anything, but it can’t. Call it a hot take, a decisive stance to take, or simply bursting the AI bubble that has formed, call it whatever you want, but this is something we, as a software company feels very strongly about. AI simply cannot do everything.
Now, why would a software company tell a reader AI isn’t the saving grace of manufacturing? Well, because we tell the truth and we’re not going to convince you of something we don’t believe in just to sell our product. That’s not how we roll.
AI is just not as smart or good as a person. AI lacks the intuition that a human can provide. That doesn’t mean AI couldn’t be in the future, but that is a long way away from where it stands today.
You simply can’t throw all of this data in and hope to get an answer for where all of your problems are and how to fix them. You have to have some kind of hypothesis about what the problem is and what you’re trying to fix. That’s even required with our software. Data without context will get you nowhere.
AI in Manufacturing is a Gimmick (Sort Of)
In truth, AI is hyped up, just as IoT is hyped up. People look at the headlines and they see AI all over the place and think, “Wow, that’s awesome. It’s going to turn the world on its head.” And, we’ll tell you this, it’s very cool. With AI, it’s basically like having a conversation with a person. It can write news, emails, social media posts, etc. You type words into and it can write the software to accomplish whatever your goal may be. It’s truly amazing.
But, that doesn’t mean it can problem solve, provide context about the world, or understands why it’s making the decisions it’s making.
At the end of the day, AI is like every other piece of software. If you put garbage in, you will get garbage out. If you don’t know what you’re trying to get out of the thing, you won’t get anything out of it. Plain and simple.
Do you want to optimize your production schedule? The data going in need to be good in order to produce a quality schedule. Do you want to find the inefficiencies of your machines? Fine, but you need to have good machine data to start. Do you want to reduce bottlenecks? Ok, great, but you first need to input the current bottleneck data.
See where we’re going with this?
But, just having good data doesn’t mean all of your problems will be solved with the push of a start button. People have to be able to support what AI software needs to run, too. Once the magical, problem-solving software is done crunching all of the numbers, it will tell you an answer, but the catch is you have to understand the answer.
Re: The Hitchhiker’s Guide to the Galaxy
Ok, this may sound far fetched, but how many of you have seen ‘The Hitchhiker’s Guide to the Galaxy’? If you haven’t seen it, I suggest you stop reading and watch the movie. Spoilers are coming up.
Essentially, the idea of the movie is that you’re living in a simulation created to find the meaning of life. These brilliant beings built the stimulation in hopes of finding that answer, the one that could explain what life truly means.
Earth generates the answers and it’s 42.
But what does that mean? What is the meaning of life? 42 what?
This just goes to show that if you don’t know the question, you won’t understand the answer. And we don’t know about you, but we sure as heck don’t know what 42 is referring to. 42 moons? 42 people? It’s all a guessing game.
If the creators had asked the right questions, they would know what the system was telling them rather than be left with a mystifying answer that no one understands.
Evolution Over Time is the Key to Success
Asking the right questions to understand the answer is true of all software, though. And, it can’t remain stagnant. If Microsoft had never updated their computer or the original Office Suite from the 90’s, it probably wouldn’t be the massive conglomerate it is today. Instead, the company would have been left in the dust as others release innovative solutions that evolved with the world around it.
Just because its fancy and new doesn’t mean all of the old rules don’t apply. The software has to evolve over time, just like everything else, including manufacturing.
While this applies to technology, it also applies to the efforts going into making improvements and the continuous improvement culture as a whole. If you want to benefit from technology like AI over time, you still have to make the efforts to grow and evolve over a long period of time.
If you have a system that’s telling you something is going to fail or how to optimize your schedule, you need to trust it and evolve the software over time, just as the software developers work to make the software itself better for you. And the key is to actually act on the suggestions once it starts making suggestions you believe.
And don’t forget the most important part, you need someone who knows about whatever it is you’re trying to solve. Otherwise, you’re taking a shot in the dark about what’s effective and what’s not.
You’re probably sensing a trend here. People, products, technology, while all different, need to work together.
AI in Manufacturing is Most Effective When Built by an Expert
In our personal opinion, to build an effective AI software (even though there’s lots of room to improve the technology), you need an expert in the field to build the model and make it work. There’s plenty of companies who may not have direct experience with the problem they’re trying to solve and essentially they’re throwing data into a database, prompting it to do unsupervised learning, and then look for anomalies and patterns based on the results.
But again, the problem these companies are going to run into is if these patterns are actually anomalies or not. You need to know what you’re looking for. And in the case of AI software for manufacturers, if the builders haven’t worked in manufacturing and aren’t experts, they won’t have the insight needed to effectively design the software.
It’s a 2-way street.
You can’t throw in 100s of data points and hope it finds the right answer if you’re not asking the right question. You’re simply taking a bunch of answers and throwing them together with no real starting point. And if you look at it, you can’t find correlations or anomalies.
In the case of AI in manufacturing, our point is that the people that built the machines, the experts at vibration analysis and how motors work, they’re the ones that really need to be building this stuff for the manufacturing industry.
Models have to be trained. There need to be parameters. To a certain extent, these AI solutions need to understand how the world works to give you valuable information. And, they don’t. At least not yet.
We have a while to go before these solutions truly can solve all of our problems. Until then, we will have to rely on a mix of software and good ole’ human intuition.
