We tried asking ChatGPT about debugging Mikrotik scripts, and his response was:

"Use the script editor in the Mikrotik RouterOS web interface to write and edit your scripts. This editor includes syntax highlighting and error checking features, which can help you catch mistakes and make your scripts more readable and maintainable."

Is there really a script editor with syntax highlighting and error checking in web interface of Mikrotik or is it another one of ChatGPT's lies? I looked for it but couldn't find it.

ChatGPT is not an encyclopedia, it is a conversation bot. It has no knowledge, it invents stuff, for conversation purposes.

We tried asking ChatGPT about debugging Mikrotik scripts...

I do not believe ... really?
It took years to persudae people that search engines are only additional tools for finding information and the answers should be verified. I'm still not sure if it succeded.
Now we face the same problem ... how to teach people thatChatGPT is not a "swiss-army knife" for solving every problem as it "knows better".

Edit:

According to ChatGPT:viewtopic.php?t=194143

ChatGPT “lies” not more than average citizen in real life. It’s not a “lie” by itself but something like a cognitive distortion that humans have. But some people treat it like a strict computer program, like an expert system. No, it’s a “black box” machine transforming input from really big database. Like you sit down to read all books in whole library and then will try to make adequate and verified answers. Like talking about countries where you never been by only looking Instagram of titty girls. You should go to a professional rather than to a book worm who read a lot but understand very little. I hope, I answered this and future questions about ChatGPT.

Is why:
viewtopic.php?t=194143


I finally understood...

They are creating "a machine" to replace one of the most hated jobs in the world: a politician...

On the other hand, he speaks well, promises well, it seems that his speeches contain something ... but in the end he doesn't do s–t ...
THE SAME...

I can understand why ChatGPT would make this mistake. It sees patterns like "the 3 configuration interfaces of RouterOS present the same data in 3 different ways" and there is a commandline script editor/printer with syntax highlighting and error checking, so why would it not be the same in webfig and winbox?
It seems like a "things to do" item for MikroTik to fix that, so that ChatGPT again is speaking the truth...
(not that I value this higher than achieving feature parity between v7 and v6, but that is another matter)

So thank goodness that there is also ChatGPT that "tells" what MikroTik must do, in addition to all the other experts on the forum who know how to run the company?
How lucky the MikroTik Staff is, it's full of people who tell them what to do, otherwise they had already closed since 2003...
What does MikroTik do with administrators, just read the forum, or ask ChatGPT.
All money saved, so he hires another dozen programmers, in order to stand still for months to realign all programmers with the existing hardware & software...

I asked ChatGPT how to best setup a server on MT, do you know what the response was?

Not possible until Mikrotik adds zerotrust cloudflare tunnel as an options package for all MT routers, and then Mikrotik will be the safest server option on the market!

Knowing that source info from ChatGPT ends around somewhere 2021, I highly doubt it responded as such.

But kudos for your persistence

@anav, @holvoetn,
if both you want, I ask ChatGPT to make me a script, I have ZeroTrust with that...

Kind of oxymoron: to trust ZeroTrust.

I can understand why ChatGPT would make this mistake.

Well, it sees a large number of older feature requests, and assumes a vendor might have added them after X years... not unreasonable

We tried asking ChatGPT about debugging Mikrotik scripts

On the "debugging" part, it does avoid that discussing that. And it would certainly miss the esoteric ":import verbose=yes" syntax to at least get line numbers to debug a script, the "verbose=" part isn't documented.

I asked ChatGPT how to best setup a server on MT, do you know what the response was?

Not possible until Mikrotik adds zerotrust cloudflare tunnel as an options package for all MT routers, and then Mikrotik will be the safest server option on the market!

I look forward to the day when Normis and Viktors release the "Anav Tunnel Package"

ATP

Hmmm.... Would be a nice name.

That sounds like broader offering that also include ZeroTier for TILE/etc as well.

Well, I watched a video about ChatGPT generating code for an Arduino setup and I was truly impressed. I had never expected that from the vague description of what it had to do, it would generate code that was close to correct. It made a couple of beginner mistakes, but it did a lot of things right the first time, and when pointed at the mistakes it had made it came back with corrections.
当然Arduino代码就像C,所以其syntax is clear and well-defined. RouterOS scripting is a completely different animal. Maybe it makes more mistakes with that.
(和我一样,although I am an experienced programmer. with RouterOS scripting I often exceed syntax limitations that one would not expect to be there)

There is a syntax checker within RouterOS, open up Terminal, "/system script edit scriptname source", you will have to press f5 to redraw the syntax highlights though.

That is what I wrote above. ChatGPT probably assumed that it would work in webfig and winbox as well (as one rightly can assume in the general RouterOS world)...

In fact they work both on winbox and on webfig, just open the terminal...

ChatGTP can give you some idea on how to do stuff, and I have for sure solved some problem with the answer I has got.
But as other points out, take care, it makes lots of stuff up.

@Rextended I will travel to Italia in may to see your beautiful country

@Rextended I will travel to Italia in may to see your beautiful country

Thank you
I hope you enjoy