New era is here. Has anyone asked help from ChatGPT when making configurations for Mikrotik routers etc.?

Yes, it can imitate valid mikrotik scripts, but many times it will make mistakes or introduce features that are not actually supported.
It was most likely trained only on a small library of real mikrotik scripts, so it often assumes things.

Not today, but maybe in a few years there will be something truly awesome to help with your configs.

I look at it the same way like Google Translate (and related services). Look how long that already exists.
It can help with broad lines but often it still needs a human to catch the tiny little details which may make or break a translation.
For spelling it is mostly spot on. Grammar and word choice, from time to time rework is needed.

Maybe ChatGPT will speed up development for that service too.

Good example with google translate. Initially it was quite terrible, but sometime in the last 2 years they switched to a new AI/ML based translation engine and suddenly the accuracy is like 99%.

Same with GPT, it continues to evolve, and in a few years, it will become even more amazing. Just have to look at the big picture

I've only been able to get ChatGPT to create ROSv6 configs even though I specifically asked for ROSv7 syntax.

Admittedly, I haven't tried very hard yet. Still, it's a great starting point !

Normunds, is it true Vicktors uses ChatGPT for pickup lines?

ChatGPT is a language processor, not a true AI. It’s more accurate to refer to it as ”narrow AI” or Google with a smarter search interface.

Artificial intelligence doesn't exist at all, someone, somewhere wrote algorithms...

Humans also have no intelligence, only more complex algorithms...
they only have a response in relation to a stimulus or a need.

There are only people who "appear" to have intelligence when compared to others who have less, or none at all.

It is enough as an example to see from a bridge two people walking on the train tracks,
one of them when the train arrives, go away,
the other not and is invested (no one are deaf, etc., etc.).
Who do you think has more intelligence?

Obviously we who watch from the bridge, the other two are idiots walking on the train tracks...

Who do you think has more intelligence?

I was gonna say "the Italian guy" but it seems I'd be wrong

I think you didn't expect that "the solution" was that...

Life is risk reward, clearly some are capable of taking risks and others are pussycats...............

without knowing the gain from the risk... all useless

(say goodbye to the signatures, on forum are disabled everywhere...)

Rex,
Sorry to disagree but you are totally wrong.
People, smart ones, are able to figure out stuff in directions they have never experienced before.
Otherwise we would only have new things in the same context as what already existed.

This is not intelligence, it is imagination.
The machine cannot have imagination (or it can simulate it with complex algorithms, but still algorithms written by real people who have imagination...)

There are deep learning systems that are able to learn and take decisions thereof, tho they are still on a same level of a child compared to human intelligence.

Some state that consciousness will reach to a teenager level within the next decade.

However, who would ever want to reason with or take advice from an AI who behaves like a moody teenager!

Yes, it can imitate valid mikrotik scripts, but many times it will make mistakes or introduce features that are not actually supported.
It was most likely trained only on a small library of real mikrotik scripts, so it often assumes things.

Well, I remember that when I tried to write a quite complex script I encountered lots of situations where I would have thought it would work, but actually it was too complex.
Trying to do too much in one expression (with lots of brackets) would fail, where splitting it in multiple steps would work.
So I can understand that a system that only knows the general syntax will come up with things that do not work. Maybe the scripting engine should be fixed so the ChatGPT inventions actually work...

ChatGPT is a language processor, not a true AI. It’s more accurate to refer to it as ”narrow AI” or Google with a smarter search interface.

It has no access to internet, so you can't compare it to a search engine.

Somehow it has to have access to internet or it would never be able to come up with the info.

I suggest reading what it is and what it does.
It was trained on a large amount of text, including from the internet, up to 2021. It has no internet access and is not a search engine or encyclopedia. It is TEXT engine. It was taught to write convincingly, nothing else. A lot of the info it gives is not correct, and this is normal. It is supposed to be like that.

Now, when Microsoft builds something like Bing+GPT3, then we will have a search engine. But that is in the future.

You are correct. Info stops end 2021.
(tested with some events I knew happened in 2021, which it could answer, and 2022, which it could not).

ChatGPT is based on GPT-3 which it’s a LANGUAGE processor (model) hence the definition “narrow AI”. The comparison with Google was just an analogy but IMO and according to many analyst is the intended business case the next generation of search engines. The rest is correct.

Btw, why not ask ChatGPT itself: https://pastebin.com/k8Yka3xE.

https://en.wikipedia.org/wiki/GPT-3
https://en.wikipedia.org/wiki/ChatGPT

I suggest reading what it is and what it does.
It was trained on a large amount of text, including from the internet, up to 2021. It has no internet access and is not a search engine or encyclopedia. It is TEXT engine. It was taught to write convincingly, nothing else. A lot of the info it gives is not correct, and this is normal. It is supposed to be like that.

Now, when Microsoft builds something like Bing+GPT3, then we will have a search engine. But that is in the future.

I am impressed by your level of knowledge, but I would be even more impressed if you understood the number of MT users running servers that are not IT trained that could take advantage of zero trust cloudflare tunnel if they didnt have to face containers and oh, of course assuming they have an arm device. In other words, less ChatGPT more coding for a ZTCT option package! ;-P

Yes, it can imitate valid mikrotik scripts, but many times it will make mistakes or introduce features that are not actually supported.
It was most likely trained only on a small library of real mikrotik scripts, so it often assumes things.

Not today, but maybe in a few years there will be something truly awesome to help with your configs.

So it would be nice if Mikrotik can offer a "training package" for ChatGPT to learn all details...

(say goodbye to the signatures, on forum are disabled everywhere...)

Great news! No more fancy links to crappy services

So it would be nice if Mikrotik can offer a "training package" for ChatGPT to learn all details...

Certainly nothing stopping anyone from scraping this forum as training data to any AI/ML models.

I'm just not sure ChatGPT has much relevance to Mikrotik configs. Do you want to describe your configuration needs in long form English – to then potentially not be able understand/edit/audit the resulting configuration?

It one thing to say "make me a dual wan configuration in a V7 mikrotik", but it another for it to work on a particular router where things are physically connect specific ports and require certain protocols/design for the specific needs.

That's not a bad training data problem, it's a using the wrong AI/ML model problem.

If that AI use the user's post... @anav @normis, I and @sindy, @Sob, etc. as example, what you expect?

That's not a bad training data problem, it's a using the wrong AI/ML model problem.

Concur. Copilot a much better alternativ when it comes to scripting and programming syntax: https://github.com/features/copilot

If that AI use the user's post... @anav @normis, I and @sindy, @Sob, etc. as example, what you expect?

AI wouldn't use semi-colons at end of line for sure. Likely think the path to salvation involves VLANs and BFD. Learn that Mikrotik Wi-Fi is an existential threat to it's existence?

But mostly resorting to the phrase "I'm afraid I can't do that" a lot.