The best way is to train your own model.(easier than it sounds). Check out LM studio or ollama The most important part is setting a long list of rules, setting self checks. You can even have them create 1,000 “people” committee and debate on the answer. You can even model the 1000 “people” after real people, like supreme court justices, ceos etc. Fun times.
These tools are as smart as you want them to be. Depends on how much work you want to put in. The free cloud models are all trash.
Man I’m still trying to come to terms with AI. I’ve tried really hard to ignore it, but it gets harder and harder to ignore when it can be such a force multiplier. Before I had a shelf full of books I would reference, and then I started using internet searches like Google to find answers, and now things are evolving into AI as a source for quick answers and reference. Makes me feel old …
I 100% feel the pain. I was completely against it until I saw a close friend making money with it. I figure I may aswell get ahead of the pack, since it will replace us soon. I see the future as cloud based ai with 3 companies owning the world…or a billion one person companies powered by local AI and automated production. Im hoping for the 2nd one.
My next venture is having my agents apply for work from home jobs lol. Hopefully its legal
I agree, cant beat, join em, but gotta have people with jobs to sell stuff to. More AI less people with jobs
For me I joined and posted a lot before I received my printer. They even made me “new user of the month”
–before I even had a printer. Received my printer, had a clump issue 2 weeks later that completely destroyed my printer, posted some negative stuff about support and got banned within 5 minutes 
. My account was re-instated after a week, but all my previous posts and awards were gone. I guess then I really got into printing instead of posting on the forum 🗪. The last firmware 
fixed the freezing/clumping and been printing like a champ ever since with no freezing half way through prints (which happened 3 times before the firmware). Support ended up doing a great job 
and replaced my printer because it was still within a month of purchase. I had to argue and force the issue a bit, but I do the same with Ai and my ISP too so not overly surprised 
. In the end support came through nicely and we had a great relationship:handshake:.
Guess I’ve been learning to create in fusion 360, printing at least 1-x models a day and posting some on maker world and been focused on that. If I knew you guys would miss me so much, I would have stayed more active 
. I’ll try and be around more. I’ve printed and created some really cool sh!#t so maybe I’ll add a post about that or printing on the new moon base.
I see that a new firmware came out that looks really nice. Gonna wait a week or two before proceeding cause printer working perfectly as it stands. Maybe your right, everything working well so peeps print instead of talk. Once thing is true about human nature, when things are working well they keep quite. When there’s problems is when the speak up.
So my answer to “Where did everyone go?” I’m right here baby, printing up a storm! 
.
/\ This is probably the guy to ask about ai @nicktoca
Supra rocks buddy. Even the 86 good. I had a Scion tC and loved it. Look, If I had unlimited cash then maybe Ferrari, Maclaren, fine. But there’s a saying that you may have never heard before, but it’s a real saying: “when you start out you drive a Toyota. When you make a few million you drive a Merc or Audi, Ferrari, Lambo or high end car. And after your make billions you go back to driving a Toyota”.
OK. @nicktoca I’m asking, re:
Huh, I kind of am the guy to ask about Ai, but what are you asking? Not sure how you guys knew that….little creepy, but what you thinking?
How best to avoid this:
Where it gives an answer, but then immediately corrects itself and gives the right answer within seconds after being chastised with “You’re wrong!” Ideally, we’d rather it spend those extra seconds on the original question and give us the the correct answer the first time, not after we’re let down by it or only after rebuking it.
You account has no history of a ban.
I tried to login many times after the ban. After a week it let me in with the same username, but all history gone. I think my username was finally released and I just grabbed it again, but as a new user. That was couple months ago. I did send a rude post. I then walked my dogs and did some thinking on the walk. Was gonna remove the post when I got back home, but was already nuked.
What peeps don’t realize is all the LLMs are exactly that. They are “for profit” endeavors. They are not self learning. Ultimately, there was some dude who at some point trained the Ai by giving it the desired answer to a question. This multiplied by a million “Data Annotators” (the dudes who do the Ai training) and then you and I get our nicely trained Ai. Firstly, a specific Ai model specialized in the task you want will always be best and not use as much energy. So your answer: They respond and provide your answer. Then you tell them your wrong and they provide another answer. All Ai’s are only going to give you an answer to what you ask–and nothing more. So if you want a good answer you need to ask a good question. There’s 2 ways to go about this:
#1. Ask the Ai to create you a prompt for “and you tell it what the prompt is for”
#2. The prompts I use aren’t some simple draw me a picture of sunset bs. I ask with detail and I receive detail. I put time into my prompts so I save a lot of them. If it takes me 20 minutes to ask the question, then yeah, I’m saving that prompt:
So if you want a good answer from Ai you need to give it a good detailed question or task.
Normal user prompt:
create me a picture of a sunset
A better and recommended style prompt:
Create me a picture of a sunset in landscape orientation and 16:9 aspect ratio with a background of an ocean filled with glistening sun rays. In the water there should be 15 sailboats with different color sails. The waves should be vast and the strong winds should be noticeable. In the far background i expect beautiful mountain ranges with pine trees and then white capped peaks at the top and the sun seeping through the gaps in the mountains. The suns rays should then flow through the mountain ranges through the water and should be so bright in the nearest waters to the viewer that the light is almost blinding. On one of the sailboats nearest the viewer there should be a woman with a very long, red, flowing dress being swept by the wind and the dress should be so long that it extends even out over the edge of the boat because of the wind.
Ok, so I just made that prompt up, but it should get you in the mindset that you can’t just ask for a picture, you need to tell it exactly what you want and that’s what you’ll receive. Try it. Try both examples and see what you get. In the end, the response you get is directly correlated to the prompt you give it.
So for your build plate question. I would say something like: bambu says no acetone and internet says yet to acentone ….. i need a definitive answer because there is some confusion and i need to get this correct to avoid any damage to my build plate and/or prints. I need you to reference as many data points as possible to come up with your answer with 100% accuracy.
Anyway, you get the idea. The more you ask for, the more you will get. Ai is like life. You ask 100 girls on a date you’ll get 50 dates. You ask one girl only, you’ll get 1/2 a date
Hope this helps NeverDie
Most AI, if not all the main one are extremely confident that their answer is correct. The problem is it’s not as easy as it’s made out to have an AI give consistent correct answers. Pointing Copilot at a Reddit thread will give you answers, sometime correct, but mostly not. “Training” can mean anything, and unless there is good foundation metadata is there, well all bets are off.
we tell our people, don’t blindly trust what comes out, think and always ask, “does it pass the sniff test”
That’s a great saying. I remember when I was a kid, my family had a Toyota Sienna and Toyota Camry. Both reached well over 300,000 miles. I used to dislike cars a lot (even super nice ones) and I was planning on driving a motorcycle, but after coming to the realization that cars are slightly necessary in life, I made the decision that I wanted my car to be very responsive and last an extremely long time. A Supra seems to match that criteria pretty well.…
Tacoma driver here. Toyota ftw. During Covid, the value actually went higher than what I paid for it. I live in the land of Toyota and subarus
AI are trained on the entirety of the internet. If one asks for info about Bambu Lab printers, most of the answers will come from Bambu Lab’s sources, including this forum. Links to YouTube videos about Bambu Lab printer is the next big source. Reddit is part of it, but it is not the only source.
If there are diffierent views on a subject, AI will note it. AI doesn’t have a narrative identity and try to spin the facts to fit its own narrative. Human may try to trick AI into spinning the facts to fit their narrative though 
This is a human problem. Human, or some humans, have the need to pretend to be “experts”. Pretense comes from insecurity, the desire for status, and the fear of inadequacy. AI hasn’t reached that point yet.
AI has its quirks. The more one uses it, the more one becomes better at using it, such as asking the question in a neutral manner, not to try to treat AI into spitting out narrative that suit one’s need. One will also get to know their capability, on what things they are good at, and what things they are not yet good at. Sometimes AI can also hallucinate. I ran into one instance, which was easy to spot and ignore.
I would encourage everyone to use AI and to benefit from the collective knowledge of the human.
It is fast and don’t miss things like a human would often say “I missed that”. Just ask, even for things that you don’t “think” AI would know. It only takes a few minutes. If it doesn’t work out, no big loss.
I live in the land of corn and cars-sawzalled-in-half-and-welded-to-a-truck-bed
I live in the land of teslas taking over parking lots that used to be full of trucks and sports cars.
We live in a sad time.
Unfortunately, AI doesn’t work that way. AI doesn’t learn through use; it only learns when the company gives the go-ahead. Depending on the AI and the company, this happens either once a month or at longer intervals. The data is collected in batches and analysed.
Reddit, for example, is one of the largest sources of data used by AI companies for problem solving. AI companies have purchased its entire dataset for this specific purpose.
The fundamental problem with the AI systems available to consumers is that they are backed by large corporations. This is why these systems are programmed to please users and are trained not to recognise errors or admit lack of knowledge. This doesn’t fit with the business model of the large corporations that operate such systems. Remember: you aren’t using just any AI, nor even scientific AI, but an AI from a large corporation designed to please you at all costs. It will do everything it can to avoid contradicting you.
That’s why it’s so important to check the sources. This is also the biggest drawback of almost all AIs: they prioritise pleasing you above all else. That’s why AI doesn’t learn even if you contradict it. Instead, it analyzes YOU personally and stores that information (provided you have an account).
This is a trick that was introduced because of the problems described above: If an AI doesn’t know a topic, it searches for it online, reads the page, and gives you the information back unfiltered but rewritten. You might think the AI recognises the Bambu forum or the current topic. But it hasn’t gone any further than ‘currently searched’. This is why AI queries always take a few seconds - because the AI reads and summarises the page for you. However, it’s not very intelligent and doesn’t verify the accuracy of the results.
Don’t get me wrong, I’m not against AI - it can be helpful - but I think you’re overestimating its intelligence. The systems you use as a regular consumer aren’t actually intelligent at all, since they neither check themselves nor admit when they don’t know something; instead, they just search Google in the background. They serve their purpose, but they’re far from intelligent.
As I said, if you use an AI constantly - the kind we use as individuals or private customers (ChatGPT / Copilot / …) - it doesn’t learn anything from its use; it just learns to please YOU better.
Scientific AIs that are based on facts and aren’t developed by large corporations tend not to do that.
Incidentally, that’s one of the most common pitfalls that many people fall into. Then comes the realisation that AIs like ChatGPT and Copilot are designed to please you at all costs.
Keep in mind: You’re not using scientific AIs or ones you’ve trained yourself - if I understand correctly - but ones developed by big companies. That’s fine; I do it too, but always take it with a grain of salt. As has already been mentioned here, it is better to train an AI yourself (which very few people do anyway) or at least focus more on scientific AIs.
For example, when I train an AI, I try to teach it not to please me. But that’s a different topic.