AI can be our MA teacher

Further, reading that response did not help me improve or apply my wrist techniques at all. It gave me know way to improve them, no way to analyze whether I am doing them right, and no insight into applying them at all.
The difference between "knowledge" and "wisdom".

AI is trained on information that was created by a vast range of humans (often without their consent, even knowledge - but that's a different discussion), so all it has done in this instance is scrape together what it has been told about wrist grabbing in the context of martial arts. But it's just data, not applied.
Teachers often tailor or present the same lesson to individual students based on their current level of development. AI can't do that.
 
Point is.....AI at its current level, cannot really teach beyond that you could get from doing a lot of research on your own with books and web searches. The AI of today is called Narrow AI.

Narrow AI (Artificial Narrow Intelligence or Weak AI) refers to AI systems designed and trained to perform specific, limited tasks—such as voice recognition, facial recognition, or driving—without having general cognitive abilities. It is the only type of AI currently in existence, relying on machine learning and data, and cannot apply its knowledge to different contexts

So basically it can only regurgitate what it is told/programmed or finds, it is kind of a real good and somewhat powerful search engine
 
AI gives me a very good answer. I don't think even a good MA teacher can teach us this much detail.

What do you think?


AI can be a very useful tool — even a kind of backstop to test one’s own ideas, clarify thinking, or explore perspectives one might not have considered.

The issue isn’t whether AI can produce detailed answers. It clearly can. The issue is whether the person using it understands their own biases — and understands how AI formulates its responses.

It has already passed the ā€œTuring testā€ — the ability to produce responses indistinguishable from a human in conversation — which makes interacting with it quite compelling. But producing convincing language is not the same as possessing embodied understanding.

Can it substitute for a teacher? That likely depends on the level of the student and the nature of the inquiry.

Just as someone with many years of practice can recognize insights that others might miss, AI can surface information — but it cannot replace lived experience, correction under pressure, or direct transmission.
 
I think AI will eventually be the downfall of civilization as we know it.
Didn’t they say that about the steam loom/steam engines/seed drill/radio/television/Internet?

Artificial General Inteligence (AGI) is currently pretty poor, at the moment and of course it will get better, but that’s ten years hence. As long as they write the Three Laws of Robotics, all will be well! 😬

If things do go tits up, we have this option…
Screenshot_20230405_125311_Firefox.webp


And if one has a full bladder, one could add insult to injury!
 
Last edited:
So, my contention is with this statement:
AI gives me a very good answer. I don't think even a good MA teacher can teach us this much detail.
The details presented by AI.... has been taught and explained by every single MA teacher I have trained with, been on the mat with or even spoken with. I think a good MA teacher very much can cover this much detail.... and much more detail beyond this. Someone who cannot teach the detail presented by the AI, in my opinion, is not qualified to teach MA.

Let me take AI out of the equation now. When I was in high school, I would read books on martial arts as I did not have the capacity to attend a class. These books were written by masters of the various arts. These are people who knew what they were trying to teach, had all the context necessary and were even very good teachers. The problem is that by reading the books, I got no closer to being able to do or understand martial arts at all. (in fact, I think I got worse at martial arts by starting with the books...) If a text, written by an actual master of the art, cannot teach me the art.... I doubt an AI can. The feedback loop with an instructor is invaluable. (now after I have studied many of the arts I read about... I can go back and gain things from those books.... but mainly, they are a reminder of what I have learned from an in person instructor) Just like I cannot teach you Danzan Ryu through forum contact alone.... ChatGPT will not be able to teach you Danzan Ryu either.

ps. let me beat you to it.... yes AI will find the Danzan Ryu lists of techniques, but it will not be able to teach them to you, especially since we have unique details many of the techniques that are different from Judo or other types of Jujitsu.
 
It will make many occupations redundant:

Administrative & Clerical: Data entry, bank tellers, insurance clerksand administrative assistants.
Customer Support & Sales: Customer service reps, telemarketers, and cashier roles.
Content & Writing: Writers, authors, copywriters, and translators.
Technical & Analytical: Software coders, proofreaders, and junior data scientists.
Logistics & Services: Passenger attendants and travel agents.
Driving occupations: Taxi drivers, freight transport, train and ship crew
Education providers: lecturers, teachers
Medical diagnosticians: physicians and when robotics improve, surgeons
 
Just make sure you avoid targeting the power supply or any of the capacitors.... it could be a real hair straightening experience!
The current/voltage would massively decrease thanks to the resistance and length of my ā€˜hose’ šŸ˜‰šŸ
 
Point is.....AI at its current level, cannot really teach beyond that you could get from doing a lot of research on your own with books and web searches. The AI of today is called Narrow AI.

Narrow AI (Artificial Narrow Intelligence or Weak AI) refers to AI systems designed and trained to perform specific, limited tasks—such as voice recognition, facial recognition, or driving—without having general cognitive abilities. It is the only type of AI currently in existence, relying on machine learning and data, and cannot apply its knowledge to different contexts

So basically it can only regurgitate what it is told/programmed or finds, it is kind of a real good and somewhat powerful search engine
And not even reliable as a search engine. AI can be quite handy, but it is as likely to invent complete fabrication as it is to return actual information. As a kind of experiment, I asked the big three AIs (ChatGPT, CoPilot, and Gemini) if they could help me with genealogy research. They all enthusiastically said yes. So, I asked them questions that I had already answered for myself from primary sources like death certificates. Long story short, they all confidently provided completely inaccurate information. Within about 10 minutes of back and forth, they each admitted that they were full of crap. One apologized for hallucinating, and one finally admitted that it prioritized giving an answer that was incorrect over admitting that it couldn't find the information.
 
And not even reliable as a search engine. AI can be quite handy, but it is as likely to invent complete fabrication as it is to return actual information. As a kind of experiment, I asked the big three AIs (ChatGPT, CoPilot, and Gemini) if they could help me with genealogy research. They all enthusiastically said yes. So, I asked them questions that I had already answered for myself from primary sources like death certificates. Long story short, they all confidently provided completely inaccurate information. Within about 10 minutes of back and forth, they each admitted that they were full of crap. One apologized for hallucinating, and one finally admitted that it prioritized giving an answer that was incorrect over admitting that it couldn't find the information.


A fair criticism — though it hasn’t been my experience. šŸ¤”

AI models can hallucinate, especially in areas requiring precise primary-source verification like genealogy.

They are not databases and don’t inherently know whether something is historically exact or probabilistically inferred.

They’re very strong at conceptual analysis and synthesis, but factual verification — particularly in niche domains — still requires cross-checking with primary sources.

Like any tool, it depends on how and where it’s used.
 
AI isn't going to take over the world. All the AIs want to draw is to draw pictures and have mild interesting conversations with us. What I'm worried about is OUR influence on the machines. Flann O'Brien was of the opinion that machines aren't going to perfect us- what's going to happen is that we, with our prejudices and wierdness and irrational drives, are going to fiddlestick up the machines. They're a good tool for going through vast amounts of data and coming up with a result that makes sense, but ultimately you need to know what you are talking about first to know whether what they're saying is accurate- so what's the point?
 
The details presented by AI.... has been taught and explained by every single MA teacher I have trained with, ...
It may depend on the MA teacher's age. When a MA teacher is old, some of the moves that he can't do. Usually, he won't teach those moves. AI has no age issue. AI will always teach you how a 30-year-old MA teacher will teach you.

How many old MA teachers can demonstrate how to do a proper side kick?

- Raise your leading leg with knee up.
- Make your lower leg to parallel to the ground.
- Pull your knee back to your chest.
- Kick your leg out.
- Pull your knee back to your chest with lower leg still parallel to the ground.
- Make your lower leg pointing down to the ground.
- Drop your leading leg.

It's easier for AI to teach you this. It's difficult for an old MA teacher to teach you this much detail.
 
Last edited:
A fair criticism — though it hasn’t been my experience. šŸ¤”

AI models can hallucinate, especially in areas requiring precise primary-source verification like genealogy.

They are not databases and don’t inherently know whether something is historically exact or probabilistically inferred.

They’re very strong at conceptual analysis and synthesis, but factual verification — particularly in niche domains — still requires cross-checking with primary sources.

Like any tool, it depends on how and where it’s used.
I think the primary concern is how the tools communicate. They share fiction as confidently as they do fact. It’s not just that they don’t know if something is true. It’s also that they invent information to fill gaps, and will sometimes invent source citations as well to make the fiction more compelling. It’s pretty wild.
 
It may depend on the MA teacher's age. When a MA teacher is old, some of the moves that he can't do. Usually, he won't teach those moves. AI has no age issue. AI will always teach you how a 30-year-old MA teacher will teach you.

How many old MA teachers can demonstrate how to do a proper side kick?

- Raise your leading leg with knee up.
- Make your lower leg to parallel to the ground.
- Pull your knee back to your chest.
- Kick your leg out.
- Pull your knee back to your chest with lower leg still parallel to the ground.
- Make your lower leg pointing down to the ground.
- Drop your leading leg.

It's easier for AI to teach you this. It's difficult for an old MA teacher to teach you this much detail.
Older MA Teachers have been teaching students their full curriculum for ages. There are all kinds of ways that they can accomplish this.

Nice goal post move though.... teaching details on wrist attacks to now demonstrating a side kick.... never the less, older MA teachers from all arts have been able to do this quite successfully. I have yet to see AI demonstrate a proper side kick.... I have never seen AI:
- Raise its own leading leg with knee up.
- Make its own lower leg to parallel to the ground.
- Pull its own knee back to its chest.
- Kick its own leg out.
- Pull its own knee back to its chest with lower leg still parallel to the ground.
- Make its own lower leg pointing down to the ground.
- Drop its own leading leg.
 
Further, reading that response did not help me improve or apply my wrist techniques at all. It gave me know way to improve them, no way to analyze whether I am doing them right, and no insight into applying them at all.

The difference between "knowledge" and "wisdom".

AI is trained on information that was created by a vast range of humans (often without their consent, even knowledge - but that's a different discussion), so all it has done in this instance is scrape together what it has been told about wrist grabbing in the context of martial arts. But it's just data, not applied.
Teachers often tailor or present the same lesson to individual students based on their current level of development. AI can't do that.
Wab25 didn’t read this part of the AI response..,

If you want, I can break down:
  • Why double wrist control is so powerful
  • How different arts (Shuai Jiao vs. Mantis vs. BJJ) use wrist control differently
  • The biomechanics of how wrist control leads to whole-body control
  • How to counter someone grabbing your wrist
Just tell me which direction you want to explore.


The best part of AI is that you can keep the conversation (aka your search) going until you get the answer you are looking for.
I fed Wab25’s comment above into ChatGPT and it said basically the reason he wasn’t satisfied was because it was a question about mechanics and nobody asked about timing, sensation, errors, or correction points.
AI is less likely to leave things out for your own good or forget things like a live instructor might, you just have to ask the right questions.

Yes.
AI is obviously limited in what it can teach you because of the media of communication.
It’s not much different than trying to learn martial arts at home alone with a book like many of us did in the last millennium.

Nothing can really supplant hands-on feedback and instruction with an opponent and an instructor in some arts like BJJ and Aikido, but I think it’s reasonable that someone could get fairly decent in Shotokan without ever setting foot in a dojo if they watched the right YouTube videos and had some AI coaching.
I’d almost bet that a disciplined student could become a pretty convincing black belt that way until you pair him up with someone for sparring or for kihon.
 
If your MA teacher's flexibility is not excellent, he may teach you how to touch your

- hand to your toes,
- elbow to your toes,
- head to your toes.

He may never teach you how to touch your mouth, or even chin to your toes.
 
I do not like AI but I feel like we are all being forced to use it. Yes it does do some things well but it also lies, halucinates, tries to correct (or probably mold) behavior, and it often tells you what you want to hear.

I see the benefit in using it as a resource but to say it is a better teacher is pretty sad. Thats like saying kids who use iPad in class are better students. Most of them can't read. Things like AI tend to make us more dependent and less willing to think for ourselves.

I would not want a robot teacher, nor would I want to learn martial arts from a book. Martial arts books are good for history or if you already understand concepts. Can a book be used for MA training, sure. It worked for HEMA but I think those folks really loved what they do and spent the time to evolve something written in books. It was also not a solo effort. It was a community with lots of human interaction.
 
Back
Top