KUKA KRC4 + AI: Redefining the Role of Robot Programmers

no more job security. The rise of sentient beings and the matrix has begun.

Maybe that helps us to work better with our robots.

In general ALL Programmers will not code anymore as before.

you may be amazed but that example code is just that - a trivial example. it is ok for academic discussion but this is not fit for any real life palletizing job i have ever seen. maybe ask it to generate code for real palletizing application that includes program resume after power is cycled, different patterns for each layers etc. i am not saying AI will not get better but i am not convinced that shared result is something that AI wrote. more likely AI just searched the internet and found piece of code that was written by a human and then just presented it. i will be more impressed if it comes up with something original - on its own. i have yet to see AI that comes up with actual solutions. and when it comes to job security, it is always the least skilled to get replaced. this will only drive demand up for people that are good in what they do. so no, i am not concerned at all. not for my job

we also have the "Pareto principle":

For commissioning we solve 80% of the requests in about 20% of the time. For the remaining 20% it takes us 80% of the time (fine tuning, cycle time, and so on).

ChatGPT may only help for the first 80% and not for the other 20%

Not to mention, that code isn't executable -- the PTP and LIN commands all violate KRL syntax, for one.

Given the way ChatGPT works, I suspect it "averaged" a large number of simple palletizing examples from multiple sources and languages, and this is the result.

That's the thing to keep in mind about these "AIs" -- they understand nothing. They merely do very wide-ranging pattern matches and agglomerate the results into a single whole. When it works, it looks amazingly "smart". But it's very, very brittle.

I laugh in the face of ChatGPT from a more generalised perspective.

I expect my comments below may spark a big debate.

Instead of 'ask Google', we now also have 'ask ChatGPT'.

Is this the global direction we need to praise and promote, is there not more important issues?

- What actual solutions to global war, famine and financially deprived countries does this provide.

- How in any way does this solve environmental issues.

- How in any way does this solve health issues.

- In short, globally, all our problems are man made and no way does ChatGPT assist with reducing this.

- Where does the liability with it's results actually sit and whom carries the responsibility over it.

- IMHO it merely has the potential of pushing 'us' deeper and deeper into the rabbit hole.

I don't doubt the power of this thing or be able to dismiss it's uses.

That's what scares me the most.

My question lies on the opposite side of the spectrum though and always draws me back to the film 'Jurassic Park' where Jeff Goldblum quotes:

"Yeah, but your scientists were so preoccupied with whether or not they could, that they didn't stop to think if they should".

Unfortunately I will never be sold on ChatGPT and will continue to live with an old school mindset as I am just an aging, bitter, pessimistic voice of doom.

From a job security perspective, I'm not worried at all, bring it on.

It should enhance my security, as if people truly believe what ChatGPT spits out is 100% transferrable, then it could be a real cash cow from a debugging perspective and areas where people require assistance.

Quote from SkyeFire

Not to mention, that code isn't executable -- the PTP and LIN commands all violate KRL syntax, for one.

that is just one aspect of it. there are also no parameter declarations etc. AI provides fast way to get results that only on the surface appear as something that may work. but it does not. in fact it is completely useless.

one day computer program can replace me at my job, it is more than welcome to do so. until then, keep returning internet searches, producing rhymes, fantasy landscapes, morphed images and write doctoral papers for people that do not know what field they are in.

Even if AI becomes advanced enough to create production ready programs, it still will take a very skilled person to give it precise and complete requirements. That in itself is basically the same skill needed to program.

I'm dying for AI to replace programmers! I don't mind giving parameters but every year there is a new computer language to learn, it gets super annoying.

Quote from Reckless

every year there is a new computer language to learn, it gets super annoying.

From an industrial robot perspective, I respectfully and 100% disagree.

OEM's appear to be maintaining their proprietary languages and will continue to do so.

Below represents my opinion only:

I think the new generation (people) today miss out on learning the 'birth' of industrial robotics and have been catapulted into a world of evolving software that represents programming industrial robots in a very limited and different way, but not promoting the underlying proprietary languages used.

I draw you back in time to the backbone of industrial robotics with keywords like:

- Unimation.

- PUMA

- VAL

Take this video from 1981, you can see clearly applications shown still exist in manufacturing today.

Not to mention the fundamentals:

PUMA...The Leading Edge in Robotic Technology | Hagley Digital Archives

From this:

- Kawasaki Robotics was born - AS Language (from VAL) and is still used and further developed today.

- Staubli Robotics - VAL3 language (from VAL) and is still used and being further developed today.

- Fanuc - Karel language is still being used today, but not too sure if it is still being further developed.

- ABB - RAPID, I do not quite know the history, but is still being used and further developed today.

- KUKA - KRL, I do not quite know the history, but is still being used and further developed today.

Quote from Reckless

every year there is a new computer language to learn, it gets super annoying.

So this statement in my opinion as far as industrial robotics is not entirely accurate.

However, I acknowledge we are and continue to be part of an evolving industry, but the reliance on the original proprietary languages are likely to still remain a major part of this industry.

Quote from kwakisaki

- Fanuc - Karel language is still being used today, but not too sure if it is still being further developed.

AIUI, KAREL is being progressively deprecated for motion control by Fanuc, but from what I've seen it's still very heavily used for a lot of non-motion programming that needs advanced logic and processing. It probably won't go away any time soon, given how much TP language can't do.

Quote from kwakisaki

- KUKA - KRL, I do not quite know the history, but is still being used and further developed today.

KRL core has not changed much over the past 30-40 years. They've made some changes, and bolted on additional features, but a lot of code from 1990 would still execute on KRC4s today.

Still, KRL is definitely showing its age, which I think is one reason the Sunrise controller for the iiWA is programmed entirely in Java (ugh!), and it sounds like the KRC5 can be purchased optionally with either KSS/KRL or the Sunrise OS (I haven't yet encountered a Sunrise KRC5 in the wild). I suspect KUKA is looking to phase out KRL eventually.

Then there's URs, which (being much newer) started with Python as a scripting language.

I wouldn't sign off on the safety report of an AI-generated program

Quote from SkyeFire

Still, KRL is definitely showing its age, which I think is one reason the Sunrise controller for the iiWA is programmed entirely in Java (ugh!), and it sounds like the KRC5 can be purchased optionally with either KSS/KRL or the Sunrise OS (I haven't yet encountered a Sunrise KRC5 in the wild). I suspect KUKA is looking to phase out KRL eventually.

yes and for anyone that gave it a try, it did not take long to really miss KRL. which is why development on Sunrise seem to have slowed down or perhaps shifted to something else - iiQKA.OS (KukaIqGpt ?)

as for AI, they seem to be more of a search engine. i remember when Google was not as smart but was able to return tons of results. now it tries to return targeted results but the number of hits is dramatically lower. some software has decided what to show you without understanding what you really need and limited results to what it thinks is best for you. because it knows you. it keeps collecting data about you. AI seem to be just a new generation, a fancier knockoff of the same thing - it is providing some utility, while keeping track of everything you do and learning from and about you. great way to scare paranoids or those who have things to hide

Quote from panic mode

as for AI, they seem to be more of a search engine. i remember when Google was not as smart but was able to return tons of results. now it tries to return targeted results but the number of hits is dramatically lower.

For those who ever reach page 2 on google

But you are absolutely right, there was a time that you could learn something new just by browsing. Nowadays is so targeted that you won't learn that much.

Similar to looking for a word in the dictionary, I always learn new words while browsing, with google you don't "waste" that time and just go directly to the point.

Relevant:

A look into how these "AI" fall over, and reveal just how much magic there isn't behind the curtain.

Nope. ChatGBT is powerful, but do you think the administrators of this project would make themselves obsolete?

We are not at Skynet level yet. We will always need techs who can calibrate harmonic drives, write code and feed the guard dog at the factory.

Get Some,

Mike

I'm dying for AI to replace programmers! I don't mind giving parameters but every year there is a new computer language to learn, it gets super annoying.