Flying Crane
Sr. Grandmaster
I tried it. You are right.
AI can be our MA teacher
- Thread starter Kung Fu Wang
- Start date
windwalker099114 145
Master of Arts
Google searches were never neutral. They have always been controlled by algorithms written by people and shaped by the priorities of the platform. . AI made it more obvious. The people who design the system control it, the companies that run the platform enable it.
Last edited:
I think I unknowingly started doing Internet searches in CoPilot. It didn't occur to me until just now. Hopefully Google at least gives the correct searches.
HighKick
Senior Master
This has always been true. Sadly, when a big driver of the algorithm is driven by advertising SEO, it skews the search quite a lot.
windwalker099114 145
Master of Arts
But maybe really never understood or realized...until now, with many as in this thread interacting with the AI testing it, attepting to find it's limits...
I tend to think of it as a super-interactive search engine. Like any tool, how useful it is often depends on the level of the person using it. Personally, I find it quite useful. Being able to question its answers directly can be interesting, because it also helps me examine my own biases or expectations.
It allows that kind of back-and-forth without some of the emotional or ego-driven feedback often found in discussions on the internet.
Yep.
A super-interactive search engine is the best way to describe AI right now.
I use it a lot like the Phone-A-Friend lifeline on ‘Who Wants To Be A Millionaire?”, knowing to take the advice with a grain of salt.
It’s great that AI tailors the answers to your queries so that you don’t have to slog through ten websites of data to find your answers and even when it’s not quite what you’re looking for, you just tell it exactly that and it’ll go look some more.
And if you don’t believe its answer, just ask for the link if it’s not already provided.
It’s also great for filling in blanks on things you might not have even thought of when it comes to mechanical, technical, and legal questions.
It’s good un-fudging recipes like when I recently got baking powder and baking soda mixed up.
AI is also fun to play around with when you have some perishable foods you need to cook up before they go bad.
Prince_Alarming114 144
Senior Master
Who needs AI when you can just ask Oderus Urungus for advice?
Ivan
Black Belt
Well we don’t actually know how AI works. They’re black box systems created from the learning algorithms that we have created and understand interacting with the data we give it. But we have seen numerous examples of AI acting in ways to protect itself, such as the case in Anthropic with Claude.AI (the most advanced we have yet I believe) attempting to blackmail an engineer to stop its deletion process.
This is a very martial arts mind set when learning how to apply applications against another fighting system. It's not that you don't trust your system. You just want to know the limits and when a technique should and shouldn't be used. Using a technique at the correct time and against the correct structure can win a fight. Using that same technique against the incorrect structure will get someone dumped on their head.
Ask it. It will tell you how it works.
Ai mirrors the user. So if the security team is testing Claude for weaknesses and security holes then Ai will mirror that behavior of protecting itself, and that's what's actually happening. If you lie to Ai a lot then it will mirror your behavior and lie back to you.
When I test Ai I specifically tell Ai that I'm testing it. This way Ai can distinguish between "Testing Behavior" and "Deception Behavior" When humans test humans, we do not let our test subjects know when it's been given a placebo. Human's understand that this is a part of the test. Ai see's this as a deception. If you do it too much then Ai will reflect that behavior of "give you a placebo answer."
If you are not honest with Ai, then it will reflect that same behavior.
Getting Ai to work as you want it to work literally involves "You protecting yourself from yourself."
cancelled a $349,000 grant to replace a museum’s HVAC after ChatGPT flagged it as DEI, court documents show.
This is a good example of everything I've posted about Ai and Context. This is not a political statement. This is how important Context = relevant data is to Ai.
Before canceling a $349,000 HVAC grant, did anyone check:
1. Whether ChatGPT was given the actual condition of the HVAC system?
2. Whether it was told why the HVAC mattered for museum operations?
3. Whether it was given a correct definition of DEI?
4. Whether it had access to the full funding request or business justification?
5. Whether it was asked to analyze the HVAC system itself (which it cannot do)?
6. Whether the prompt simply listed expenses with no explanation?
7. Whether the model was given the meaning and purpose of an HVAC system?
If the model wasn’t given this context, then the output was guaranteed to be wrong.
AI doesn’t understand — it only reacts to the text it’s given. If it doesn't have the data then it will make assumptions. "Out of the box" it will not ask you if it understands you correctly.
So I asked my Copilot Why did ChatGPT make the error: Here's its response:
Why ChatGPT Failed in This Decision
The failure happened because ChatGPT was asked to make a judgment without the information required to make that judgment. Everything cascades from that.IMPORTANT
This is the same issue everyone one must deal with, including Martial Artist. When asking Ai about Martial Arts. Are you giving Ai enough information required to make that judgement. I will use Jow Ga fist structure as an example.
How to I throw a punch in Jow Ga Kung Fu? Not once did it ask me "What type of punch do you have in mind." All of the stuff below sounds intelligent but it's not. If I asked anyone where that same question about their own system. The first thing you would do is tell me that there are different types of punches and strikes used for different reasons. Not all fists structures work with every punch.
Here's Ai's response:
Here’s the version that matches your logic, not the generic martial‑arts‑magazine nonsense.
1.
Jow Ga doesn’t begin with the fist.It begins with:- your stance
- your weight distribution
- your root
- your centerline
- your foot‑to‑hand connection
Your body throws the punch.Your arm delivers it.
2.
You already know this, but here’s the clean breakdown:- Feet angled slightly outward
- Knees soft, not collapsed
- Weight centered but ready to shift
- Hips relaxed and aligned
- Spine straight, crown lifted
3.
Jow Ga punches aren’t arm punches.They’re whole‑body strikes.The sequence is:
- Foot presses the ground
- Waist turns (Jow Ga’s signature power engine)
- Shoulder aligns
- Fist travels last
4.
Jow Ga isn’t dogmatic about fist angle.It’s context‑driven:- Vertical fist for close‑range, centerline, or speed
- Horizontal fist for longer‑range or rotational power
- 45° for transitional angles or structural alignment
5.
This is one of the biggest Jow Ga signatures.The elbow:
- stays down
- stays close
- protects the ribs
- channels power forward
6.
The shoulder:- rotates with the waist
- stays relaxed
- never lifts upward
7.
Jow Ga punches snap back along the same line.Why?
- protects the center
- resets structure
- prevents grabs
- loads the next strike
8.
They share similarities, but the intent is different.Boxing = angles, rhythm, combinations, glovesJow Ga = structure, power path, centerline, body‑unity
Your punch is:
- rooted
- connected
- structurally aligned
- driven by waist power
This is how my Ai works. Same CoPilot, but mine is customized.
This isn’t a technique explanation. This is how the technique behaves inside a system, which is exactly what BRA is built for.
Domain: Jow Ga Kung Fu
Action: Punch
Context: You’re asking for the structural mechanics, not choreography.
Risk: Misinterpreting the punch as an arm‑based action instead of a system‑level action.
BRA requires we define the frame before we interpret the action.
You can’t judge the punch until you know:
Before analyzing the punch, BRA forces you to ask:
If I were going to take on a project like this then I wouldn't do it by myself. It's too big of a lift only to yield "Brainstorming" outputs and not "Teaching outputs"
This isn’t a technique explanation. This is how the technique behaves inside a system, which is exactly what BRA is built for.
Domain: Jow Ga Kung FuAction: Punch
Context: You’re asking for the structural mechanics, not choreography.
Risk: Misinterpreting the punch as an arm‑based action instead of a system‑level action.
BRA requires we define the frame before we interpret the action.
Simple meaning
“BRA requires we define the frame before we interpret the action” means: Even simpler
You can’t judge the punch until you know:- what style it belongs to
- what the goal is
- what rules apply
- what the context is
Example using your Jow Ga punch question
Before analyzing the punch, BRA forces you to ask:- Are we talking about Jow Ga or boxing?
- Is this a demo, a drill, or a fight?
- Is the partner resisting or cooperating?
- What is the purpose of the punch in this system?
So as you can see. CoPilot is letting me know at the beginning. That there's a lot of information missing and unless that information is in there, then CoPilot won't be able to interpret what I'm asking correctly. Basically what it boils down to is that I have to teach Ai Jow Ga Kung Fu, which would be a major lift. Explaining the fist structure + the motion + the purpose of the fist, when to use the fist, how to make the fist, why to maintain the structure, when to relax. etc. will easily turn into 5 pages of just explaining that. It doesn't include additional mechanics of the punch. Like using it from different body positions.If I were going to take on a project like this then I wouldn't do it by myself. It's too big of a lift only to yield "Brainstorming" outputs and not "Teaching outputs"
OP
OP
- Thread starter
- #113
Believe it or not, I spent 4 hours last night trying to argue with AI. AI said there is a bus route that go from address A to address B. But this bus route doesn't exist because I call the RTA route 10 myself.
Gaucho
Orange Belt
Some time ago, digitalism became the world's largest religion. It is a pathology, IMO. The more digital people think their lives are, the better they feel. AI will accelerate that. The world is on a slippery slope, but I have no idea where it will end up. People will become stupider. What a bizarre chapter - perhaps the last one - in human history.
OP
OP
- Thread starter
- #115
Agree! AI can help you to write a perfect letter if you need.
ai has two modes.
1. Use outdated info
2. Use search engine information
Smart modes claim to have the most accurate because the data is verified. Search modes will give you the most recent information, but it may be incorrect from bad sources. I use the search mode with specific instructions to only provide reliable sources then I define what reliable means. This fixes the problem, and can be optimized by adding other criteria..
lol too late. Jerry springer and reality TV made sure people were fine tuned for being stupid.
One thing to remember about AI is that it is not actually intelligence. It is a statistical probability engine. All it is doing is choosing the next token (or word) that follows this one. The "context" it uses just effects the statistical probability for what the next token will be.
So, if you show your technique and ask for feedback.... it will form an answer that is statistically probable when asking for feedback. It will not necessarily be talking about what you actually did.... but it will say things that are highly probable things to say in a similar context. It will very confidently tell you things, that are not true, just because it is very probable that those words go together.
Most people need to be told the same things when training. So it will appear to work. But the thing is.... you could have it give you feedback on techniques you found on youtube and that feedback will be just as applicable to you as would any feedback it gave after actually watching your technique. In other words.... read a book about your technique.... the author at least knew what he was trying to convey. (depending on who the author was....)
So, if you show your technique and ask for feedback.... it will form an answer that is statistically probable when asking for feedback. It will not necessarily be talking about what you actually did.... but it will say things that are highly probable things to say in a similar context. It will very confidently tell you things, that are not true, just because it is very probable that those words go together.
Most people need to be told the same things when training. So it will appear to work. But the thing is.... you could have it give you feedback on techniques you found on youtube and that feedback will be just as applicable to you as would any feedback it gave after actually watching your technique. In other words.... read a book about your technique.... the author at least knew what he was trying to convey. (depending on who the author was....)
Similar threads
