ゲストの紹介
Hello, here is FA Radio, and welcome to hear our podcast. So, this time, I'm Chris. Okay, let me introduce myself again.
I'm Chris Chung. I come from Hong Kong, but I'm working in Japan and for factory automations. And then, you see here is Takahashi-san.
Takahashi-san is also a factory automation engineer. He's working for factory automation. Now he's focused on the software PLC. He's a very great engineer in Japan.
Okay, so, maybe can you introduce yourself, Alex?
Thank you for having me and inviting me to your podcast. It's an honor to be here, so thank you.
But what I would like to... how I introduce... I'm an engineer. I'm an automation engineer, actually electrical engineer.
So, I worked as a controls engineer, maybe around 15 years altogether. Maybe about 12 years I was fully in controls.
Oh, really?
So, yeah, before this, I was doing a little bit of electrical design, putting things together.
So, building substation starters with a lot of electrical variable speed drives and spent some time on the control system.
So, building PLC programs and HMIs and starters.
So, yeah, altogether, when I was working, I think the majority, maybe 60% of the time of my experience, I worked with coal mines.
So, in Australia, the region where I'm from, it's Newcastle, which is, in general, New South Wales, this state where we have a lot of coal mines.
In fact, we have the largest coal export port just next door. I think it's the largest, not in the southern hemisphere, just in the world, as far as I know.
Could be wrong. You should ask AI.
So, I know that you have so much experience from being an electrical engineer, starting from supplying, to the hand working, and then to design the PLC program, and then maybe also good commissioning to the customer side.
So, you did all this stuff in your career, is that correct?
Yeah, that's exactly. All of this.
So, design system, program, test it, and go inside and spend a few days inside, and maybe a week, sometimes a month, commissioning.
Just to get some of the large projects we just have to spend.
Not because of the control system, it's usually so much mechanical things need to be done also.
Yeah, all of this. Design, test, program, test, commission.
Infinitive loops, is that?
It is, exactly.
And tell me, where was your country? Where was your company's country? In Australia?
Australia, yeah.
So, yeah, I moved here, migrated about maybe 17 or 16 years ago, I think. From Russia, originally.
And I live here now, and I have kids, and I have my family here.
Yeah, but I worked a lot in control system automation here also.
But at some point, I left the company and I opened my own, which is the Contral-X I have.
AI技術の活用
So, what we actually do in Contral-X as a control, we mostly do the PLCs, some scatter projects.
We do a lot of software also development, where we combine PLCs.
We combine PLCs and other IT softwares. So, kind of OT IT connections.
So, we build bridges sometimes to get some data out and communicate to other systems like MES or ERP.
But not actually ERP much. We had only one little thing.
But there are other little software packages you have to talk to.
Okay. So, I think the main reason why I would like to invite you to this podcast is,
I think it's sometimes the lenient that you send an email to me that,
Hey, Chris, do you like to try our AI-based runtime? Is that correct?
And I try this. Actually, I think I'm being smart.
So, could you introduce me and Takashi-san, what kind of magic that your company made?
Yeah, that's a good question. So, at some point, we were doing a lot of software.
And a few years ago, right? How many years? Two or three years ago, when TouchPT came out, right?
We saw, I still see the power, how amazing AI generates tokens, right?
And some of them put together in a code actually works.
So, what we did by making any other projects and working a little bit on the structured text in PLCs,
You go in Gemini. I mean, now Gemini, but ChargePT before was the first one.
Or Cloud, what we use also a lot.
You put your question and produce something and you copy and paste it to…
So, we thought it is a lot of work to copy and paste. And AI does know a lot of things.
So, what if you combine them together?
Combine together? What do you mean combine together?
So, what if you have a program which is actually PLC program-like, which can execute logic,
But also have AI built there. So, you don't need to go another window. You stay in one place.
You mean that we can integrate, I would say, we can just ask ChargePT,
Hey, let's do something for us, then you directly integrate in the project.
Exactly. That's how Jasper Node was born. That's how we decided to build one.
I see.
Jasper Node is actually a decision to put this together and make it easier and simpler for us and other engineers.
To develop control systems.
And I'm saying control system, not only split pieces of logic because of AI doing this very quickly,
But architect to create big systems.
So, yeah, that is the first step.
A few years ago, when we started putting together, that's when Jasper Node was born.
Jasper Nodeの紹介
So, it means that this JavaScript is still very young. It's just 2, 3 or 4 years old still.
It's only one year old. Yeah, exactly.
Yeah, we are very young. We're just getting started.
In fact, what we have is Jasper Node is in the stage when it's production ready.
And we're looking for engineers and other testers who can join us.
We're carefully selecting people who can, like yourself, for example, who can try out and give us valuable feedback.
So, we want to introduce that to the people who want to try.
We know that there are problems, right?
So, we know things that a lot of people are scared of AI.
How do you put AI in the control system? Are you crazy?
Not just in Austria, same in Japan or other places.
Yeah, exactly.
And I understand everybody should be afraid of this because of AI generates so much code that you just sometimes cannot comprehend.
If you put it in a system and just run it, it can have big problems.
Okay, I get it. So, tell me a little bit about your JavaScript.
So, what kind of platform do you need? Do I need to choose some special hardware to run your application?
Or I can just take a small paper like this to run that?
How about your other feature of your software?
Yeah, that's a good question because that is usually confusing.
Like, what is a controller? Because people used to see a hardware, a piece of something.
So, yeah, the goal of JavaScript is to become one software which can include any Ethernet protocols.
I mean, all Ethernet protocols.
But have any ability to communicate to any Ethernet protocols, industrial protocols.
And to be used on any hardware.
When I say any hardware, I mean a PC.
So, we can run on Mac, on Windows, on Linux, on anything you have available for you, you can run it.
And the whole idea is to create a system which can communicate to everything and be ready to execute your control logic you need.
But we know that in reality, you will never actually have a Mac in industrial settings.
But what you want to do is, you have your own Mac, you install a new program, you create the systems, you create a backup.
Or save file.
After you bring over to your little PC, you showed me like a Seed or Raspberry Pi or Revolution Pi.
Any of the ARM-based or you can get little NUX.
It's from, I think they are absolutely ready.
So, there are little industrial PCs you can buy with Windows.
You install Jasper Node there.
And from your developing machine, you bring over this package, your backup and restore that.
So, you basically develop anywhere on any machine.
You create little save file and bring to your industrial PC and restore it and it works.
So, it means it's cross-platform.
So, I don't need to care what kind of platform that I've created the project by Jasper's NUX.
And then I just save it and I can download it everywhere, every runtime.
Exactly.
Okay.
So, I think it is very great that I think now, I mean not just in Japan and also European or USA also thinking.
The Indians don't like to, I would say, choose the hardware.
They would like to choose the hardware that they want.
So, maybe in the old school style, I also used Siemens before.
I used Siemens, Rockwell or Mitsubishi before that.
The Chinese don't know how to have PLC.
But in your application, you prefer to choose the hardware that the end user, they can want to install, play with your runtime.
So, is that correct?
Yeah, exactly.
So, they can play with the runtime on any computer.
But when you decided to already use it in your automation system, you bring a little industrial PC, which most of the time now runs Linux.
Okay.
You run one command, only one command.
I think you have tried before.
So, it just magically installs everything for you.
Yeah, sure. And auto start, no need to care everything.
Exactly. And it's just there.
And it's ready to go as a service.
It creates all necessary processes at the back end to reliably run it.
And do you need any tubes to config your runtime?
Because, for example, Siemens, you use TI Porter.
And in Rockwell, you need to use, I think it's Logic 5000.
So, what kind of engineer tubes do you need to configure your JSPENODE runtime?
Yeah, so that's a good question.
So, one of the goals of JSPENODE is to make it as easy as possible for engineers to use.
You don't have to have anything installed.
The runtime already has user interface.
The runtime?
So, whenever you go to this link, that is your user interface.
This is your engineering environment.
So, that way you can see the connectors, where you can see the tags and create tags, delete tags.
Also, you open AI agent and you can chat with AI.
And it is all on that runtime.
So, it's all inside there.
So, you don't have to have a special software for this.
You can become your tablet.
Like if you have a Samsung tablet, it becomes your programming computer.
Your Mac becomes a programming computer.
Your Windows is a programming computer.
You just only open the link to your industrial PC and you start programming.
You mean that your runtime has an integrated web server that we don't need to…
Exactly right.
I think that's a very great thing.
For example, I also played with some big European companies before.
Their engineer tube is more than, I think, more than seven gears for just install.
You take one or two hours.
And even it's just for you to control with two or three pumps or ten bottoms or something.
Actually, sometimes I also don't know what I'm doing.
I know exactly what you mean.
Because of the same thing.
I had a project where I had a new hardware.
It was a Siemens and I had to install TIA, the different version.
It's actually hard to spend the time to wait till it's installed.
Either the time, I was actually inside sitting in front of this cabinet.
It was a super hot day.
And I was just hiding there and trying to program.
And it probably took me a couple of hours to just install things.
So I don't think it's a good approach.
I think it should be fixed.
And we're fixing it by embedding the web interface inside the runtime.
Okay, so maybe this should be to your runtime.
AIシステムの紹介
You told me that you have an AI system.
So you told me what is your AI system in your runtime?
And how does it work?
And how do they help the engineer to make their life more happy?
Yeah, no, that is a great question.
I think this is probably maybe the first with AI-based and AI-first DLC or controller, software controller.
So we build everything about that is AI has more power than the engineer.
What I'm saying more power, it means to modify and to figure out things it needs to.
So it has access to little things.
So many of them.
Every engineer can also look into this what they want to in configuration files.
But they can be long.
But what we do that is AI has access to everything.
But only acts when you ask.
So it is very.
So we also when I started using this, we have a project, a few projects.
Around maybe 10 different where we have Jasper now running.
And we were also cautious and worried about what's AI going to do.
We didn't want to.
I mean, I don't want to wreck my machine, my equipment.
It's the client's machines.
And if something goes wrong, I have to pay.
So I didn't like this.
So we did the system when AI is really, really much guardrailed.
So it has the system in place where it designed the way that it only can implement one task at a time.
And when you want to, it can confirm every single task when it does.
Every single task.
You mean that it's not running cyclic?
So this question I'll ask you a little bit later.
Let's focus on the AI assistant first.
So how does it work?
For example, how is it?
The AI assistant I think is integrated into your runtime, is that correct?
Is that correct?
Okay, so how does it work?
How can I play with that AI assistant?
Yeah, that is one of the misunderstandings.
People think that AI does run something.
But what do we do?
We have a very similar architecture to the PLC.
We have a logic.
We have inputs coming in.
Logic executed.
Outputs comes out.
Inputs come in.
Logic executed.
Come out.
And it's cycle.
Typical cycle as a PLC.
With the two, maybe three different things we brought into to make AI more reliable.
First, this logic in the middle.
All this logic can be modified by AI or by you.
But it has a very careful thing we implemented.
It's called atomic tag control.
So atomic, that means only one tag can be changed by their own script.
It's a bit difficult to understand.
I might show you a little bit later how to...
How this actually is running.
No, how it actually looks similar to Siemens or to Alan Bradley or even Node-RED.
I have an example.
Give an example what script and tag is in Node-RED or in TIA.
But anyway, so we decided to have this.
So only one tag can be executed by script and script can be changed by AI.
The second thing is that AI also has access to modify these connectors.
Where the input is coming from and where it comes to.
So for example, when you go to Siemens, TIA or any other development environment.
You can create additional hardware.
So in our case, AI can create the configuration for this.
You mean that, for example, I would like to create a network with this 3 or 4 nodes.
Then you just chat with your AI system.
The AI system automatically create all the conversion for you.
Exactly. So in the back-end, it creates this all configuration for you.
Like if you go to PodSys and you go to EtherCAT.
You know it has so many little settings.
So our AI does it for you.
So we don't even have that thing.
The AI does it for you.
You just ask that and it creates it for you.
You still have an option to go and have a look.
It's usually in text now.
We don't have the user interface yet built properly for this.
I mean you have user interface but very limited.
You can see in weekly things.
But the whole idea is to be AI first.
So you just ask what you want.
And after you can polish different places by yourself if you like to.
Okay, I get it.
AIによるプログラミングの革新
So it's very powerful tools to maybe change what we do for the internet works before.
Because I also worked with a VLC more than 10 years.
We need to open this drawing.
And then to check out what kind of module that we need to install in our project.
And then we need to install it one by one.
To define the tasks one by one.
And then we can start the programming.
So it means that your AI system can not just create the logic.
It can also create the conversations.
What kind of stuff they would like to connect.
Is that correct?
That's exactly right.
Oh, that's great.
And I would like to have a question.
You say that your AI system can create the logic.
So what is the language that your JPEG stock system can create?
What kind of language will your...
Yeah, you just jumped into the third part of what design decision we had to take.
Is that we decided to use JavaScript.
JavaScript?
JavaScript, yeah, exactly.
The reason is because of...
We didn't aim to create firm real-time deterministic systems.
We wanted to build the controller for general automation.
Where 5 or 10 milliseconds is not a critical delay.
So now we have a question.
What language is most popular in the world?
And two languages come up.
It's the JavaScript and Python.
So two languages, that means the LLM knows really, really, really well.
So what our goal is to ask LLM to create the logic for you.
And we want LLM to be right.
So we want LLM to create logic which works.
So if LLM knows very well JavaScript, so we should use JavaScript.
With Python, we didn't want to go ahead with Python.
Only one reason, because it's too slow.
Because if we actually...
Our runtime is JavaScript runtime.
The Python is much slower.
Actually, it's pure JavaScript your runtime is.
Correct, yeah, exactly right.
Actually, the logic you created from AI assistant is JavaScript.
That's exactly right.
Everything underneath is JavaScript runtime.
Have you heard about Dino?
I think it's more fast.
It's more fast than Nototo JS, right?
Yeah, exactly, exactly.
It's two alternatives now to Node.js.
It is Dino and Bun.
So we use Dino.
So they're all program written in this.
It's just because it helps us to create a very quick, great program.
We have a roadmap.
We have an idea.
We want to move further to deterministic systems later.
We're already putting a few little bricks of the Rust.
So our EtherCAT connector is written in Rust.
You mean you have EtherCAT connectors now, your runtime?
Yeah, exactly.
So our runtime can speak to EtherCAT devices like Backoff behind me.
We wrote this in Rust.
We actually used somebody else's library,
but we had to write additional functionalities to cover what's needed.
So when the EtherCAT is done by your team, it's stuck?
Yeah, exactly, exactly.
I'm very surprised that when I first time when I meet Alex,
I think it's the last year around October or something, right?
October or November or something, right?
I think September, October.
At that time, he told me that you are still delivering EtherCAT.
Maybe it is sometime in January, this year January.
He told me, hey, we did the EtherCAT PC.
And then it works very well.
I'm very surprised.
How can you do that?
I don't know.
You're not a very big company.
I don't know how you can do this task very fast to implement the EtherCAT.
And I see your roadmap also EtherCAT.
You are now trying to implement EtherNet IP also.
Is that correct?
Yeah, exactly.
EtherNet IP is going to be our next.
We have actually three more coming.
It's this EtherNet IP, OPC UA, and BACnet.
BACnet.
So the runtime also focus on other field, not just factory automation.
We go to the field building automation also.
Exactly right.
We're building a software controller, which will be talking to any EtherNet protocols.
Eventually, we'll have all EtherNet protocols there.
It just takes time.
Eventually, you'll ask AI, connect to this device.
And you take a photo of this device saying, this is a device.
I don't know what it is.
And AI will figure out everything for you.
It will know what a device, it will try things.
We have tools like a ping.
Our AI can ping if it needs to.
Our AI can check the traffic on the network.
Our AI inside Jasper node can do a lot of troubleshooting on network to find out what's available, how it works, if it's connected, if it's not connected.
So that's the system.
Do you mean your AI is not just a tool for the engineer to create a project or to create a machine?
Also, after the machine is done, it can also do the dynamic state and for the maintenance, for the troubleshooting.
Is that correct?
That's exactly right.
So this is one of the reasons what we use with other clients, Jasper node.
When you have a system and something breaks down, the person who has no idea about the control system can ask, what's happening with my system?
You can, in plain language, come over.
Like the operator.
AIアシスタントの機能
We really want to have a maintenance team to be able to solve the problems by themselves by just coming over to this machine and saying, this motor is not running.
Tell me why.
And AI can do troubleshooting for you.
It's not going to change anything.
It's not the tool for change.
So I think your version still doesn't have it.
But the newer version in AI box, it's going to have ask or change.
So you can switch to ask.
And AI will have no rights to change anything.
It will only look through, ask you questions, look through more and will help you to troubleshoot.
Okay, so thank you.
So now I know that your system can, in your Jasper node runtime, the core thing is you can use AI assistant to create the logic for you, to create the conditions for you and to have the troubleshooting.
And also it's running in Java, actually, not just a runtime and all the logic is generated by JavaScript.
And because it's the most famous language in the world now.
So I think I summarized now.
So the next question is.
You did well, you did well.
Yeah.
So yeah, I'm very surprised that my first time I hear this type of architecture.
I didn't imagine it.
It's Jesus.
It's very clever, this guy.
How they think, how they convict this type of runtime.
Okay, so my next question is your AI assistant.
So your AI assistant is in the runtime.
So actually, in that time, we need to connect the internet to your server, to LM server, is that correct?
Yeah, exactly.
So you'll need to have internet to program at the runtime.
So how it works.
So this is the cycle, always works in execute logic.
The actual logic you have, it's in runtime, saved on the machine.
On this computer you're using.
So if you used before an industrial computer, it still sits there.
All the values will all be saved always.
So everything saved on the disk.
Every time, every single run, everything's happening.
But when you need to program this, you have two choices.
One choice is optional, which is you go and open and manually look for the code and change it.
So you'll be able to see the actual logic, see the actual JavaScript language code.
So you can modify this a little bit if you need to.
Or you can connect the system to the internet.
And then ask AI, hey, so I have this problem.
Or let's say another problem.
Maybe let's say we're bringing you five inputs.
Or we have extra motor here.
Let's get this new motor commissioned and tested.
So AI will do all logic for you and it'll help you to commission and test you.
But for that, you will have to have internet.
You will have to have internet to communicate with AI.
So when you're finished, so that's it.
You're done, you commissioned, your system running.
You look at this, press buttons, looks good.
You take and disconnect the internet and that's it.
So the system will keep running.
So the same logic was there.
If you turn it off, turn on again, turn off, turn on.
It'll be just everything the same here, there and keep running.
Yeah, this is a very interesting stuff.
Yesterday we have a podcast, me and Takasa.
We're talking about the Cloud PLC.
It's made by Toshiba.
Toshiba, a very big Japanese company.
They will put their own whole controller and logic on the cloud.
And then inside they will use some special technology
to make the very high-speed communications.
Only 20 milliseconds something to control the IOs.
But their logic will be on the cloud.
But for your solution that your logic is still in the local side.
And then for some of the control logic to turn on the motor,
to get the data from the sensor,
all this logic or all this control medicine
will be still in a local machine, is that correct?
Exactly, exactly.
Yeah, we just want to keep machine running.
クラウドとローカルのロジック
So we don't want to have, if internet's gone or anything,
we just want machines running.
So it needs to keep running.
Okay, I get it.
And yeah, if you want to manage this,
because I know people, the reason why we do this,
because sometimes management is very difficult.
So when I have hundreds of different machines,
we have JASPX.
JASPX, it's a platform in the cloud,
which communicates to nodes.
And if you change something on a node,
if you still have internet,
you can create a snapshot.
It creates a snapshot there.
So if you have a JASP node,
hardware broken or whatever happened to it,
you take it away and chuck it in a bin,
bring a new PC and go to JASPX and deploy that.
So it deploys the whole of what it was before.
They have exactly the same configuration,
exactly the same logic, and it just runs again.
So with all this small runtime you have,
nodes that is run on each local machine,
but you have other platforms,
like JASPX X to control,
to manage all your nodes.
That's why I call it X the nodes, I get it.
Yeah, exactly, exactly.
X is like a central place and nodes are everywhere.
It's the nodes, I get it.
So it's very great.
Actually, I have other questions.
ビジネスモデルの考察
I know that most of the companies
that would like to use the AI assistant,
they need to pay.
If you think that's okay,
could you tell me a little bit about your business model?
Finally, people cannot use the AI, it's free, right?
Could you tell me what is your thinking of your business model?
Yeah, so with business model,
we want to be always fair to people.
I mean, I don't like too many subscriptions.
Sometimes you don't use things
and subscriptions you have to pay.
So we decided to not use subscriptions for AI here.
We basically created a tool with a connection to AI,
and when AI is used,
how many tokens you have to pay for that.
So you pay as you go.
There are obviously benefits and disadvantages for that model.
But this is how we started.
We were already talking with one of our potential clients
who wants to have an actual subscription.
So I think it will be flexible.
We'll probably have two different tiers for the whole time.
One of them will be you pay as you go.
Like let's say you have two projects,
or one, three projects a year.
You bring in, you program your system,
you install the machine, disconnect internet,
and it's running.
You don't have to pay.
You don't have to do anything.
So it's just running.
Whenever you have to come over troubleshoot,
you connect, you pay for that use,
and disconnect.
For some of the clients who have,
like you just said,
like Toyota or any other you used to work with,
they probably don't like this idea.
They want to have the same cost every single month.
So for this kind of organizations,
we're going to have an enterprise version
where we just introduce a subscription,
and it will be easier.
And there will be probably some particular limit on the tokens.
Because if they're using millions and millions and millions of tokens,
it still can be very costly.
So it depends on how they would like to use your,
how much they need to use your AI.
If they need to use more and more,
at that time, they also have a subscription plan for them.
But if they just use for troubleshooting or something,
you can just pay for what you do.
Exactly right.
So even that I don't want to use your AI,
I can just use the runtime to try to program by myself,
is that correct?
Exactly, you can do everything manually.
Everyone can do AI.
You can do it yourself.
I feel very surprised that you didn't charge on your runtime.
You didn't charge on your tubes.
You charged on your AI assistant.
And then you told me that it's charged if you use it.
I think it's very surprising.
I didn't see this kind of business model before.
Takasa, I'm very surprised.
Actually, I introduced your platform to him a couple of times.
Every time he said he was very surprised
why they can make a business model like this.
We think it's going to work.
I think it will be the future.
So you want to use AI and you want to pay when you use it.
Actually, I thought about this a lot.
And I think AI is going to be very similar to electricity.
That's what I think.
It's going to be another commodity.
It will be just like you can drink the water from your home.
You can turn on the light.
The lamp will be turned on.
If it's close to you, then you don't feel it's AI anymore.
Something like this.
No, I think in terms of the payment.
How do you pay for electricity, right?
You turn the switch, you use it.
And at the end of the month, you pay a bill.
Is it like this in Japan?
I don't know.
Same.
You just pay how you use.
I think it will be the same with AI in the future.
It doesn't matter who produces tokens for you.
It will be all of them.
Open AI, XAI, Anthropic or Cloud or Google.
Whoever produces for you.
You probably will have one supplier
who will switch between them.
It depends on your task.
And you probably will have just paid to the supplier.
So you don't really know what models they're using.
They will choose depending on the task.
I think so.
I think that's how it's going to happen.
Because of this technology which we'll live with all the time.
It will be here all the time.
And I think how to make money is all the companies.
They will just produce like electricity.
They will keep producing these tokens for you.
And you're going to consume them and pay for these tokens.
And you use more this month, less next month.
I think so.
Yeah, because I know that for example,
if you're having in EU for some Siemens or even Lockwell or Backoff.
They also try to do their assistant.
But I think they now or still in 2026,
they still have a very clear business model.
How they charge to the market.
I don't want to say too much in here.
Maybe they would let you to use free at the first few months.
And then one time they say,
Hey, you start from charging this month or something.
I don't know what is their business model.
But I've talked with others, talked with you,
JasperXの技術とデモ
that are very clear what is your target.
And what is thinking about AI.
And how you can use, how you can save our lives.
So yeah, I feel like your business model.
And I'm very surprised that there's something like this.
Yeah.
No, it is going to change a lot about what we do.
And it will help controls engineers, automation engineers a lot.
It's regardless who is going to bring it.
Us or any other companies.
It does great work.
But what we have to do as a tool provider.
I am making a tool.
My job is to very carefully look into where AI make mistakes.
And remove them.
So you, when you use this system,
You feel comfortable with this.
You trust it.
And it produces great results every time.
And if it doesn't, it has to alert you.
Or ask questions.
Or follow up with the system it doesn't understand.
So it's a lot of problems every single LM has.
And I know that Cloud and Google and other work on this.
Or OpenAI work on this to make it better.
And our job also now is just to know.
I don't want to damage equipment or injure anybody.
It's not something I ever wanted.
So my job is to be very careful with this.
And that's why we are looking for the testers who can test with us carefully.
We can find these old bugs and problems.
We went far.
We have a really good system now.
It just has so good.
This chance of making a mistake is so little.
Okay, I get it.
And so we're making tools to test this.
So you can have trust in it.
Okay.
So I think now we talk with 40 minutes.
And I think it's now the time to give some real demo to our listeners, correct?
Yeah, exactly.
I think probably it's a good idea to give a little bit of an idea how it's...
Yeah, yeah.
Because just talking, I think the people will say,
Hey, is that true?
They say it's AI-based.
Is that true or not?
Do you know how to share your screen?
So yeah.
Do I share just here in that?
Yeah, just share.
Okay.
You can share the screen now.
We're now using Google.
You can just directly share the screen.
I will try to speak through what I'm doing
because some people are listening, right?
Yeah.
Don't worry.
We'll upload as a video.
So don't worry.
Just...
Okay, we see.
Okay, so could you show...
There's something different what I see before.
Did you update it?
Hey.
Is it a new version?
Yeah, a little bit different.
Yes, exactly.
So what you see here is just the home screen.
Okay.
They usually...
What we want to show the connectors.
We have two different groups of connectors
where we call some of them as a hardware connectors
on the left side.
And you can create extra connectors
or you see already existing ones here.
Okay.
And a connector is basically
where the data comes from to Jasper node.
Okay.
And we have IT connectors
which are where data goes out
or also can come in
where also another side.
We're showing just two different sides to make sure.
Because the hardware connectors,
you want to have a fast and reliable and high priority
to execute anything.
And software connectors, it's a less priority.
And it's data usually for monitoring
or for reporting, basically.
Okay.
So you're split to two types of connectors.
For the hardware, for the view control,
the control that I use for the pumps
or to get a signal from the sensor
and make a high priority.
And the right-hand side is the software controller
to connect the IT system database or some stuff.
Okay.
Exactly.
Let me just switch to the light mode
because if I think dark mode,
it will be not really easy to see for some people.
I'll just switch to light.
It's straight up right to us, right?
So anyway, so this is the connectors on the left side
and the other connectors on the right side.
We have a little bit of here as a dashboard
shows you a little bit of information
what you want to show here.
That's all of this can be adjusted
so you can bring anything from your program.
So it's quickly to see some values.
You can toggle things, you can run things.
You can see the value changes,
but this is only for just little HDMI dashboard.
Just very quick one.
Oh, my God.
Just very quick.
I didn't hear these features two weeks before.
I didn't.
No, no.
Last week, I still didn't hear these features.
Yeah, no.
It's a lot of things coming.
It's really good.
It's great.
So this dashboard is directed to link to your talk.
You can directly turn on, turn off,
or you can directly see your talks.
Yeah, yeah.
Here, this is basically...
I can jump in quickly, but I don't want to...
No, no, I don't want to jump.
Sorry.
I will continue just in the flow.
Just take your time.
Now is your show time.
Just take it.
So yeah, there are two ways you can create connectors.
There is a manual here.
You can click on the one you want to create.
You have a Modbus device connected to your hardware.
JasperXの導入
You press the plus button here.
Okay.
And you fill in the information as usual.
New config, put IP address, and other details like registers here and addresses.
So you can set everything manually here.
Okay.
But also from the homepage, when you press plus,
it actually opens to AI.
Oh, this one. Let's see.
Yeah.
And here, you probably can just simply say,
create Modbus connector.
And you probably want to do maybe some IP address and so on.
So you put the information right here.
Let's say this address.
And it will run and it will start creating connector for you.
Okay.
But only if you have auto run.
There is also agent with approval.
So when you click run with approval, it will ask every stage.
So it will not do change unless you confirm.
You can put in ask and just ask a question.
How many Modbus connectors I have and what does it do?
Let's say.
And what it does, it just goes in trying to figure out,
how many connectors you have, how they look like.
Sometimes it has errors when it's trying to run two of them.
And it goes through all of that by trying to describe everything for you.
In the language you'd like to hear.
So we have a Modbus TCP connector client.
It tells you how it's connected, what's IP addresses, cycle time.
Okay. And what it does.
And we have a server.
Okay.
Which is also and how many registered has.
It says server is serving everything from HMI page.
And you have temperatures, you have discrete inputs and so on.
You can ask about your system in just one quick sentence
instead of jumping through everywhere.
But you could do it yourself.
You could close it and go and look and read and see everything here.
Okay.
And you can see, okay, I have a server.
You have path, you have port.
You have one variable, another variable and so on.
You can look at them and see.
But you could ask AI.
So you could do it everything yourself.
So you mean for the ask feature.
So is that possible to let AI to analyze what the project that I'm doing?
Yeah, exactly.
So you can, when you knew, for example, you had five engineers before work on some systems, right?
Some particular system where have a Jasper node installed.
When you come in, you have no idea what's going on.
Yeah, sure.
You can go to ask mode and say, what does this application do?
Yeah, because it's very painful.
Every time when I go to site, I need to upload the program and I need to check the logic.
And that time I need to imagine what is the program is running.
It's very painful.
It's exactly.
But now it tells you.
Sorry, just keep scrolling for me.
Yeah, okay.
You also analyze the field bus.
So the application is PIC, Reactor Temperature Control.
So it's the main purpose.
Control a heated mixing vessel.
Automated temperature regulation and safety interlocks.
So it describes to you the process.
Describe what hardware you have connected.
We have Logo PLC connected.
We have Modbus connected.
And I have EtherCAT, which is the back of EL2008 unit.
It describes the control logic, what it does.
And system how it works.
And what's something special about this.
It's actually talks a little bit about Jasper node.
I don't know why I decided to do, but I decided to add this.
But it's very helpful.
And you as a new engineer.
You just quickly understand what's going on.
And it can go drill down into particular things you want to know.
Okay.
So now we know the ask.
And could you show me how to create a logic?
How can I create a logic by using your AI assistant?
So here in automation design, that's where you go and see.
This is usually that all the systems.
We usually ask AI to create summaries.
So this is as a summary of your system.
You can always see what system looks like.
So you can ask AI a program.
You say create a tag with explanation.
Oh.
Or here, this is my readme file.
You can see what actually system does without asking AI.
You can always have it inside your program.
Just describe what it is.
So we always like to do it.
Okay.
But if you want to add logic, for example, let's see.
We have a HMI right here.
HMI tags.
Let's say I want to add one more here.
Let's say, which one do you want to?
We have controls.
We have over-temperature alarms here.
Heater interlock.
So from this control, we have a heater interlock.
I can go here and say.
I would go without approval and say add.
I've just copied this tag.
Tag to the HMI folder.
So what it's going to do is going to check this exists.
Next, it's going to find out what it needs to do.
So what it does is duplicating the tag.
Now, if I go back to my HMI, it's already here and ready to go.
It's disabled to run.
We can enable here or just keep it disabled.
So we keep it enabled.
But what it does, it's duplicated.
It usually has exactly the same logic.
But you probably want to see the controls.
It's a very basic way.
AIによるプログラム生成の仕組み
I think the listener is also very interested.
What kind of logic or what kind of stuff did your system create for them?
You told me that your system created a program in JavaScript.
Yeah, exactly.
This is system enable, right?
If you click on that button, you can modify the script.
So this is script how it looks.
This is what AI generated and you can do it yourself.
So it depends on if you want to do yourself or not.
Yeah, exactly.
But when you start using it, you'll be doing zeros.
I used to do it a little bit by myself.
I know JavaScript very well.
But I don't do it anymore.
I only rely on AI.
I only read what AI tells me.
Because sometimes it goes like, oh no, you're doing something wrong.
No, no, no, stop, stop, stop.
And because sometimes I'm too ambiguous with my question.
Because my prompt sometimes creates a connector.
I didn't tell anything about what device.
I didn't tell anything about IP address.
I didn't tell about what registers we have.
I just say I'm so ambiguous.
That's why I start guessing.
But I usually stop and say, okay, let's add this information to you.
And AI gets really good at this.
I get it.
I think we have maybe more complicated.
Do you have more complicated?
No, this is just scaling.
So it takes raw value and creates for you.
I get it.
That's great.
I see.
We access by just using Chrome or age to the web browser.
You don't need any special application as you mentioned before.
Exactly.
I have a little thing here I want to show you.
Sometimes people don't understand when I'm explaining.
This is a tag.
This is my digital output.
Let's say I have...
Which one will be more complicated?
Probably I'll try to do this.
At the back of my background, I have this tower.
If I change it to true, you can see I'm overriding this output.
I'm actually writing this output like a force in this output.
Now I see who is actually triggering this output.
If I want to troubleshoot something, I come here.
If my output is not working, because of only one place where you can change this tag, it's only this script.
This is called atomic control.
It's only one place.
No way in a program, in any other script, I can change this tag.
Only this tag.
I go here and I look.
Mixer run is only one thing which triggers this output.
In a whole project, one digital output can only trigger by one script.
Exactly.
Now I go to this tag.
Mixer, that one.
Now I see logic.
I can see logic.
I can also see who is triggering this logic.
Now I'm like, oh, that's a system enable or system alarm.
What does overtemp alarm does?
I go and look on control overtemp.
I go to control overtemp and I click here.
Now I can see what's triggered this.
Can you see the logic?
Because one output is only in one place, you can have these links.
We have even little diagram.
When you click on this folder, you can create relationship.
You can see how they all interacted with each other.
Is it new features?
No?
It was before, but we're still working on it because that is very helpful.
Eventually you get to this heater.
This is digital output one.
You can see who is actually, what particular tags.
The relationship.
It's actually started with the temperature.
The raw temperature and after temperature and end up in digital output one.
Hold on.
Did I talk about one or two or zero?
I think I talked about zero.
Sorry, I actually talked about zero.
Yes.
We talked about this mixture.
Now the mixture is overtemp and system enable.
This type of flow, your system flow is only focused on the output.
The output is turned on from where and then to go to find out where the trigger point.
This is very important.
That's actually, I've said this so many times, but this is very important because that helps AI to make less mistakes.
When you have a program where everything in one place, one mistake can be crucial and big problem.
トラブルシューティングとシステムの重要性
Yeah.
Some people just don't work.
But the problem will be less happened.
The mistakes less happened because the AI knows the structure.
If you break somewhere in the middle, something will just not work.
But nothing going to be crucial happening like a bad happening.
I feel like your question is that don't let AI to make the mistake.
Yeah.
We don't let them by creating very tight and specific system for the AI.
You cannot do by another way.
You cannot be super free.
You just focus on one tag.
That's what you're asking.
So why you generate the system with the output can turn on just by one point.
If your program can turn on one output in different subroutine that maybe you make the mistake.
Where is the point to trigger the output?
It helps AI to troubleshoot and AI making less mistakes by doing this.
Okay. Thank you.
Now I see you're connected.
Now you still have Voodoo Bus and EtherCat and S7 connectors and Lego.
So I think in the future, you also increase more and more, right?
Yeah, exactly.
It will be hundreds of them here.
So we'll add all of them.
Okay.
JasperXの機能と利点
Could you show me what kind of information can we see in your system?
Diagnostics, you say?
Yeah.
This is the very basic one.
It's a big log of information that's happening.
Thank you.
It's the same thing.
It's the same thing.
If you're here, you can quickly see that log what's happening.
If you have errors, error will come up here and tell you it has an error.
Okay.
And then all your logs can be controlled in your whole platform called JSPack X, is it correct?
Yes, exactly right.
So the whole goal is to manage.
We have a few projects and we want to know if some of them are faulty or something happened.
We have one place to look at this.
So JSPack X gives you just visibility to all of them.
You can do it.
Yeah, you can do everything.
You can have a look.
Okay.
Quickly here, I want to mention one thing for engineers.
If anybody is still looking and still here.
This is actually a representation of what script, triggers and tag is.
So if you go back here, if you go to any of these tags again.
Let me just quickly go and let's say, which one are we looking at?
That one, right?
Yeah.
We have triggers, we have a script and we have that tag.
So this is the fourth.
Okay.
So if you look at the Node-RED, for example.
Okay.
Then at the end, you can say this is a tag.
Anything in between, this is your script.
And this is your triggers.
So something triggered to get your flow working.
And you actually got a tag result.
This is a very similar approach.
If you are using TIA portal, look here.
Yeah.
These are your triggers.
Okay.
This is a very basic script.
So that's what we have.
Sometimes scripts actually getting super long.
Yeah.
But most of the time, they're simple like this.
And this is your result of your tag.
Oh, they're very clear.
They're very clear pictures.
Yeah.
And if you are Brad Watson, the same thing.
Look at this.
This is one, two, three, four.
Four triggers.
So one of them will start triggering the entire process here.
Okay.
Entire calculation.
Not calculation.
It's logic, basically.
Logic.
This is your script.
And eventually, you have only one tag out is that result.
Oh, I get it.
So that's what we have all in JasperNode is hundreds of these networks.
Okay.
Around of flows.
Separate to trigger, script, and result.
Exactly.
So now that's secondary running.
It can be a trigger in other logic, right?
Yeah.
In another script.
So this result may be other results trigger.
Oh, I get it.
So actually, I think the concept is very similar to what we've done before.
Exactly.
And you do it in a very similar way.
Similar to the PLC.
It's very similar to PLC.
But we are trying to solve that the way that AI makes less mistakes.
I get it.
Okay.
Thank you.
Oh, I still have one hour.
We talked about one hour.
Yeah, don't worry.
I hope your audience is still with us.
I don't know.
I didn't say that for the last time.
Last week, I talked with Alex more than two hours.
Yeah, exactly.
Just for talk.
We just had a little conversation.
We decided to speak for 30 minutes.
Yes, yes.
So, okay.
Now, I would like to know a little bit more about what is your roadmap?
What is your next step?
What kind of stuff, what kind of magic would you like to implement in the next two months or next years or in these years?
Yeah, that's a good question.
So, our one and important mission is to make automation easier.
So, we want to have all the processes which are tedious and boring, give it to AI.
Okay.
So, what are we trying to achieve now is to all these configurations to connection to something should be done by AI.
Okay.
Because this is sometimes boring and annoying.
Sometimes you don't even understand why it's not working and you have to spend so much time just to weaken things.
We want to AI do this.
In terms of logic, all logic, if you look into any PLC logic, run separately.
They are very basic.
So, that can be also done by AI.
But what we want to do by humans is this is what I'm now enjoying doing a lot, is to architecting the system.
Okay.
Just designing what sensors and everything I want to use, how it's connected in real life, and the rest of these tasks are given to AI to solve.
And now, if you're talking about sensors, we talked about video the other day.
It can be hundreds of different protocols.
Yes.
Different configurations.
Yes.
So, our goal is to achieve that as it becomes so seamless and easy.
So, you only tell what device is connected.
And actually, in fact, with EtherCAT, we love EtherCAT so much.
I mean, I love this protocol because it actually tells you what's connected.
So, you don't even need to know the IP address.
JasperXの進化と機能
So, you just go and you can discover on your network what exists and it tells you.
Okay.
So, that's our roadmap.
Basically, implement protocols, improve our AI, and make management easier.
So, you have 20 different nodes, different places.
You go to JASPX.
You can see all of them there.
You can easily connect to different of them.
See what's the status of them, if they are running well or any issues with any of them.
So, all management is becoming easier.
Yes.
And the good way I ask these questions is even at this scene, all this now in this web server is totally different.
Not the same as I've seen last week.
And they're doing very, very fast.
And he told me, hey, Chris, last year, hey, we will release EtherCAT on next year.
And then in 2026, in January, they released it.
So, and I see their website.
I'm always working.
I'm always browsing their website to see what kind of stuff they release.
I see there.
And for the recent NIP, and for the backlog, they're still in testing phase.
And they will also try to implement so many different protocols, like OBC UA or something in this year.
So, I think they will let us to be more easy to collect the data from the fieldbus, not just fieldbus.
I also have IT side software connectors and hardware connectors.
And actually, I feel like this solution really.
And I also would like to try to tell more people in Japan that what kind, what is, what, we love PLC.
I love PLC also.
I would like to play PLC.
I would like to play server.
And after I talk with Alex, then I know, hey, I think there's so many different ways to go to finish our tasks.
And Alex told me that, hey, you can use AI assistant.
For some of you, you only have three or four problems with a tank switch.
Actually, you don't need to use something like TI, a big software to buy PLC to do that.
You can just scan the network and then ask AI assistant for what you want to do.
And then check it.
And then finish.
So, I really like this approach.
And maybe this solution will be not focused.
Can focus on the very high end, maybe less than one millisecond application.
But I think for the 10 millisecond application or even some application is running in 100 millisecond for the building automation.
Maybe in this time, Alex, Japan's NUC is a very good solution for this type of application.
Yes, they're correct.
Absolutely.
エンジニアへのメッセージ
No, exactly.
Yeah.
If you have a very time critical applications, not recommended.
But if you have something process, pumps, general industrial networks, so great.
You should use it.
OK, thank you.
And don't worry, I will focus.
I also will see, talk with Alex.
I think I have talked with Alex one month, two or three times to communicate.
So if they have any great magic that they released, I will let all the Japanese internet to know.
Don't worry.
OK, so I think we speak more than one hour.
So Alex, do you have anything that you would like to talk?
What would you like to say to the Japanese engineer?
Yeah, I actually would like to add that is if you're an engineer in Japan or any other country listening to this, stay open.
Try things, be open to AI tools.
I know that gets sometimes annoying, but be open for this.
It's some of them very helpful.
If you'd like to join our program of early access, please contact me.
Just connect with me on LinkedIn and we can work together on some projects you have.
So we'll be happy to help you.
And this way you probably can help us.
Yeah, if you feel shy, you can just ask me.
I can ask Alex, don't worry.
Exactly.
Yeah, very great guys.
And he just paid me.
He really tried it and sent me a command that I tried.
And then I tried.
It's a great solution.
That's why we have this podcast.
That's great.
Thank you for the podcast too.
It's great.
Yeah, good.
Now I think this podcast is a collaboration.
It's my first time to do this collaboration.
But I really feel very like Alex.
He's a very great guy.
I don't really learn so many different things.
And last time he just asked me, hey, Chris, could you teach me something about IO-Link?
I said, yes, okay.
Just book a meeting.
We can have a two hours meeting.
Just talk about IO-Link.
Exactly.
Yeah.
Okay.
So I think you talk about one hours.
I need to let our listeners know that actually Alex now is around 11pm, right?
Yeah, it's 11.20.
Yeah.
So I don't want to...
I have to go to bed soon for sure.
Yeah.
I don't want to take you too much.
I think you should drive to your home, right?
Yeah, exactly.
Just drive.
It's okay.
Okay.
So I think our podcast will be ending here.
So I think Takada-san, do you have anything to write to us with Alex or some stuff?
そうですね。
英語は聞いてたんで、内容はわかってるんですけど、ちょっと翻訳してもらっていいですか?
はい。
アレックスさん、ありがとうございます。
そうですね。
JasperXの可能性
製品はもともと知ってたんですけど、今回アレックスさんの思いだとか、製品だとか、いろんなものを聞かせていただいて、本当に将来の楽しみな製品だということがより強くわかりました。
これからも期待してますんで、続きの開発待ってますんで頑張ってください。
はい。
Okay.
So Takada-san said that actually he learned your solution before.
You have to talk to him.
I talked your solution to him so many times.
And he think it may be also a future of what factory automation can control the machine line.
And he is very interesting.
He would like you to go forward and forward.
He would like to see what is your next magic can show to us.
Perfect.
Yeah, exactly.
We will be.
We will go forward.
And with the factory automation, at some time we will get this Auditorium Mystic version released.
And you can use it in any applications.
Okay.
Okay.
So thank you for Alex, and thank you for Takada-san, and thank you for me.
Chris, thank you so much for this.
Okay.
So please have a good night.
Thank you.
And I think I will book a meeting to you very soon.
No problem.
Good night.
See you.
Bye-bye.
Thank you.
See you.
Bye-bye.
