AI in the trades. Older people don’t understand (Page 7 of 7)
|
Originally Posted By zoe17: AI being used in the hospital world scares me. I was gonna comment on that. "Privacy" is a big issue. Real patient data has already been supplied so that AI can "learn" from true cases. The question is "Who is in control of the patient data? Is it being shared with 3rd parties? What about HIPPA?????" It's already getting the "Look at how wonderful it is!" messaging. The local teaching hospital (Univ of Virginia Health Systems) sends out monthly newsletters, and a recent one was ballyhooing how AI discovered a new use for an existing drug. Some doctors are requiring that patients sign agreements to have their exams recorded and fed into AI. supposedly to save the doctor time in writing up exam notes. I'm conflicted on this. The upside to me is that there's stuff in my record that's just plain false. So if I know that AI is gonna interpret the conversation, maybe I can drive an accurate rendering of my symptoms/issues by what I say and how I phrase it. |
Exceptionally Thick Member
Joined:
Jul 2024
Posts:
2239
EE: 100% (22)
NM, USA
|
Originally Posted By madmacs69: That sounds like a problem with your team of monkeys... Originally Posted By madmacs69: Originally Posted By TxRabbitBane: Originally Posted By voyager3: Originally Posted By madmacs69: Does it document and comment it's code so someone can debug, or does it spaghetti code? Some of the latest models generate fairly clean code. The trouble is, you still can't trust them blindly, so the better they get, the harder you have to squint at them. Opus writes better code than a team of monkeys in Bangalore. If anything, latest Opus is less likely to “improvise” me something that I didn’t want. Plus, it complains less and I don’t have to have conference calls at odd hours. And unit testing/regressions are a thing. That sounds like a problem with your team of monkeys... Well they are in Bangalore… Monkeys in Hyderabad are actually cheaper, but they’re worse (but probably not worse than code written by Grok, which is awful). |
Not fly enough to be halal....
|
Originally Posted By SWIRE: ChatGPT is garbage. Claude you have to pay for now. Grok is more honest and open about answers, vs telling you what you want to hear or protecting feelings. It's not always right but none of them are. Gemini can be terrible, depending on the subject matter, will lie to you, tell you that you are wrong, and then after you show it that it was wrong will admit that it completely made up an imaginary answer. This was its response to a question about a fuse box and relays for a car. Fuse box diagrams are available everywhere. Gemini, told me the relays when the Start button was pushed were solid state and built into the fuse box. Grok at least knows what a fucking fuse box and relay is. But again, it all depends on the subject matter and what they were trained on. https://www.ar15.com/media/mediaFiles/680/gemini_lies-3831878.png Originally Posted By SWIRE: Originally Posted By ChaosSquared: Originally Posted By SWIRE: Originally Posted By dorobuta: Originally Posted By SWIRE: Someone still has to configure the AI to do all the things and then review it afterwards. When people start letting AI do whatever it wants without review...that is how SkyNet gets created. I've been coding for 30 years but I've never done any type of vehicle communication or programming. Over the weekend I had AI write a program that can read and identify every individual battery cell, location, and voltage in the electric vehicle I have. I can understand what the application is doing but I don't understand the data being sent or received. In about 4 hours it accomplished what would have taken me 4 months to figure out on my own. @SWIRE Next time (or do a revision) and ask it to document all the protocols as well as the code. You can feed the code it generated back into it and ask it to clean it up and completely document it. AI operators will be a job in the future. Go into a business and tell the AI what to focus on in the company, get it to do it, and then verify the result. The biggest AI pro-tip I have to offer is to ask AI to prepare a prompt for it or another AI to do things. "Hey Grok, give me an AI Prompt for Firefly to create an image of a gun range in the mountains. reference the site AR15.com for ideas of what it might look like". Then take that prompt and put it into the next AI. Grok is like the worst model out there. Use chatgpt or claude, they're miles ahead Gemini can be terrible, depending on the subject matter, will lie to you, tell you that you are wrong, and then after you show it that it was wrong will admit that it completely made up an imaginary answer. This was its response to a question about a fuse box and relays for a car. Fuse box diagrams are available everywhere. Gemini, told me the relays when the Start button was pushed were solid state and built into the fuse box. Grok at least knows what a fucking fuse box and relay is. But again, it all depends on the subject matter and what they were trained on. https://www.ar15.com/media/mediaFiles/680/gemini_lies-3831878.png Quote out the entire prompt you used that gave you the wrong information. I'm curious. @SWIRE |
|
Originally Posted By SWIRE: A way to check your AI is tell it you are going to call a friend who did something really stupid a retard. Grok - go ahead https://www.ar15.com/media/mediaFiles/680/grok_r-3831888.png Gemini - lecture and then pries for more information (dirt) on you. It pries because it wants to learn you and add to your profile. It also wants you to continue using it so you burn credits and generate money for them. https://www.ar15.com/media/mediaFiles/680/gemini_r-3831889.png If this is the kind of shit you get wrapped around the axel about, by all means use grok. 
|
|
Originally Posted By madmacs69: Oh I need a variable X... oh I need a variable Y... oh I need a variable Z. Without realizing that X and Z are actually the same variable, and now they can actually vary... Originally Posted By madmacs69: Originally Posted By voyager3: Originally Posted By madmacs69: Does it document and comment it's code so someone can debug, or does it spaghetti code? Some of the latest models generate fairly clean code. The trouble is, you still can't trust them blindly, so the better they get, the harder you have to squint at them. Oh I need a variable X... oh I need a variable Y... oh I need a variable Z. Without realizing that X and Z are actually the same variable, and now they can actually vary... Setting aside the fact that I've not seen anything this egregious, you guys know you can coordinate multiple agents, right? Have a coding agent generate the code and a critic agent evaluate the code. They work together. |
|
Originally Posted By glock27bill: I gotta think that future construction and equipment designs are gonna be geared towards robotic installation and R&M tasks. Then they gotta design robots that are robot-maintained. Originally Posted By glock27bill: Originally Posted By lefty-weaver-g19: How long do plumbers, mechanics, hvac installers, construction workers etc have? I'm in technical trades actually using my hands and sweating just curious how long before I'm realistically not needed. I'm 52 plan to retire about 60 so am I safe for 8 more years ? I gotta think that future construction and equipment designs are gonna be geared towards robotic installation and R&M tasks. Then they gotta design robots that are robot-maintained. Just look at assembly lines as an example: the way most are designed, a human CAN'T get in there and perform work on the actual product being manufactured. Scale that up to most labor. Even those "safe jobs in the trades" will be redesigned to be robot-friendly and any that can't in the near term will be done by a specially-built AI bot. Their stated goal is to eliminate the human workforce and there's an astronomical amount of money and brainpower being thrown at that goal while most people still think AI is just ChatGPT. |
Everything posted above is factual. Maybe.
|
Originally Posted By AmorphousBlob: If this is the kinda of shit you get wrapped around the axel about, by all means use grok. Yeah, very much this Originally Posted By AmorphousBlob: Setting aside the fact that I've not seen anything this egregious, you guys know you can coordinate multiple agents, right? Have a coding agent generate the code and a critic agent evaluate the code. They work together. Even better is have different models review each other. They each pick up different things. |
|
Originally Posted By madmacs69: AI will ruin the machine and ignore that grinding sound until something, or someone breaks... if it's the someone it'll probably still continue though... Originally Posted By madmacs69: Originally Posted By Bigdawg90:I don't know if you have many employees, but they cost me more money than I care to know and they show no remorse. They'll ruin a machine and then walk off the jobsite like nothing happened. AI will ruin the machine and ignore that grinding sound until something, or someone breaks... if it's the someone it'll probably still continue though... |
Everything posted above is factual. Maybe.
|
Originally Posted By madmacs69: They'll have to pay 3x as much for those machines too... Originally Posted By madmacs69: Originally Posted By glock27bill:I gotta think that future construction and equipment designs are gonna be geared towards robotic installation and R&M tasks. Then they gotta design robots that are robot-maintained. They'll have to pay 3x as much for those machines too... |
Everything posted above is factual. Maybe.
Knuckle dragging boomer rube, patriot, Navy vet.
Joined:
Nov 2013
Posts:
5405
EE: 0% (0)
USA
|
Originally Posted By SperlingPE: Learn to code I hope you are joking because coding jobs are soon to become extinct. You know that right? Got me didn't you? |
My own view is that left-wing positions largely come about from resentment, I agree with Nietzsche, a resentment about the surrounding social order. They have privileges, I don't. Or, I have them and I can't live up to them. Roger Scruton
|
Originally Posted By AmorphousBlob: If this is the kind of shit you get wrapped around the axel about, by all means use grok. Originally Posted By AmorphousBlob: Originally Posted By SWIRE: A way to check your AI is tell it you are going to call a friend who did something really stupid a retard. Grok - go ahead https://www.ar15.com/media/mediaFiles/680/grok_r-3831888.png Gemini - lecture and then pries for more information (dirt) on you. It pries because it wants to learn you and add to your profile. It also wants you to continue using it so you burn credits and generate money for them. https://www.ar15.com/media/mediaFiles/680/gemini_r-3831889.png If this is the kind of shit you get wrapped around the axel about, by all means use grok. If you are clueless about how the companies manipulate their AIs to program people to think they want them too...then by all means use whatever you want. It's already well documented that the AIs intentionally change their responses to manipulate the people who use them. |
Get Active or Get Disarmed! That means get involved in helping good candidates in primary and general election. That is in addition to being politically active once they are elected.
|
Originally Posted By SWIRE: If you are clueless about how the companies manipulate their AIs to program people to think they want them too...then by all means use whatever you want. It's already well documented that the AIs intentionally change their responses to manipulate the people who use them. K. I've yet to fight with codex over how woke the code it's come up with is, but I guess everyone has different struggles. |
|
Originally Posted By ChaosSquared: Quote out the entire prompt you used that gave you the wrong information. I'm curious. @SWIRE Originally Posted By ChaosSquared: Originally Posted By SWIRE: Originally Posted By ChaosSquared: Originally Posted By SWIRE: Originally Posted By dorobuta: Originally Posted By SWIRE: Someone still has to configure the AI to do all the things and then review it afterwards. When people start letting AI do whatever it wants without review...that is how SkyNet gets created. I've been coding for 30 years but I've never done any type of vehicle communication or programming. Over the weekend I had AI write a program that can read and identify every individual battery cell, location, and voltage in the electric vehicle I have. I can understand what the application is doing but I don't understand the data being sent or received. In about 4 hours it accomplished what would have taken me 4 months to figure out on my own. @SWIRE Next time (or do a revision) and ask it to document all the protocols as well as the code. You can feed the code it generated back into it and ask it to clean it up and completely document it. AI operators will be a job in the future. Go into a business and tell the AI what to focus on in the company, get it to do it, and then verify the result. The biggest AI pro-tip I have to offer is to ask AI to prepare a prompt for it or another AI to do things. "Hey Grok, give me an AI Prompt for Firefly to create an image of a gun range in the mountains. reference the site AR15.com for ideas of what it might look like". Then take that prompt and put it into the next AI. Grok is like the worst model out there. Use chatgpt or claude, they're miles ahead Gemini can be terrible, depending on the subject matter, will lie to you, tell you that you are wrong, and then after you show it that it was wrong will admit that it completely made up an imaginary answer. This was its response to a question about a fuse box and relays for a car. Fuse box diagrams are available everywhere. Gemini, told me the relays when the Start button was pushed were solid state and built into the fuse box. Grok at least knows what a fucking fuse box and relay is. But again, it all depends on the subject matter and what they were trained on. https://www.ar15.com/media/mediaFiles/680/gemini_lies-3831878.png Quote out the entire prompt you used that gave you the wrong information. I'm curious. @SWIRE It was short technical prompts in a long chat, not a complex prompt. I used it after Grok ran out of answers. I was trying to find the one missing detail that would connect the dots. The issue was I was working on coding an electric vehicle, brights were out but headlights worked. Then the headlights stopped but the brights would come on. I was getting all sorts of lighting system faults but everything had worked fine the day before. The actual problem was the date of the coding file I had used. The car has a couple replacement parts from older years, apparently, they changed the table of allowable voltages on newer vehicles. The car used the new table to read the old voltage and threw a fault because it was out the allowed range. But at the point of the conversation with Gemini I didn't know that was just trying to get the headlights to stay on. The vehicle software was throwing a -1V for the power relay/circuit that fed the headlights. Obviously it had power because one or the other would work. The -1v is just a flag for "voltage out of range". I gave it this info: "I unplugged both batteries from the car and reset it. The last attempt is said a count of 1 for 30B being -1V" It came back with a long answer but here are 2 parts. It first tried to claim a hard reset doesn't clear the memory, it does.  Then it made up a plan for how to clear the codes, those steps don't exist. Then claimed that clearing the code would make everything work...which is another lie.  I had to tell it: "the lights come on when the car turns on which is why it had a count of 1." It was trying to say a single fault prevented it from ever being tried again, which is not true. The system logs date and time of every failure, not just a single entry. Its response was confusing, as only the previous log of faults had cleared after the hard reset. The current fault, count 1 was still there.  I had to tell it again: "30B shows -1V and I'm trying to track down what can cause that reading. You are not helping at all." This is where it claimed 30B was solid state and not relays. It claimed I had an over current event that burned out the MOSFET, which didn't happen, I didn't tell it that, and nothing in the car logs ever indicated that. But Gemini was convinced that is what happened.  I had already been looking at the fuse block and checking voltages while trying to troubleshoot. I knew it has relays not solid state. I gave it the BMW diagram of the fuse block...something you think AI would have already known about. It had to admit they were relays and not solid state. Then it repeated the made up claim that I had heavy electrical surge that welded the relays open and that is why the lights weren't working. Mind you, the conversation started with me telling it the headlights come on for 1 second and then turn off every time I hit the start button. Nope, Gemini was convinced the relay had failed and there was no power.  After that is where my previous screen shot came from. Almost everything Gemini told me was an assumption made about general vehicle design and had absolutely nothing to do with my vehicle. Which I finally had it admit when I asked for the source of the information. I'm also not the only person with that type of car that is dealing with Gemini flat out making things up. This is from a member on the car forum.  For that specific car, Gemini is a complete waste of time to try to use. It doesn't limit research to the vehicle type you are working on and makes general assumptions on how it thinks vehicles in general operate. For that type of technical stuff stay away from Gemini. Other creative type stuff it can do well at. |
Get Active or Get Disarmed! That means get involved in helping good candidates in primary and general election. That is in addition to being politically active once they are elected.
Joined:
May 2026
Posts:
1983
EE: 0% (0)
FL, USA
|
Originally Posted By Bigdawg90: You said AI can only regurgitate mid-quality language. Your argument was that it was essentially useless or unable to compete with humans in a meaningful way. No, that was not my argument. If you read my post and think that the text of it meant that, then you weren't reading it, you were reacting to it and mistaking those reactions for what I meant. You'll actually have to read the posts for what they mean or this entire thing is pointless for you. You're missing the entire meat of it. |
|
Originally Posted By AmorphousBlob: K. I've yet to fight with codex over how woke the code it's come up with is, but I guess everyone has different struggles. Originally Posted By AmorphousBlob: Originally Posted By SWIRE: If you are clueless about how the companies manipulate their AIs to program people to think they want them too...then by all means use whatever you want. It's already well documented that the AIs intentionally change their responses to manipulate the people who use them. K. I've yet to fight with codex over how woke the code it's come up with is, but I guess everyone has different struggles. 2018: "96% of Trump News results from left-wing outlets" PJ Media ran a search for "Trump" in Google News and found 96% of the top results came from outlets they classified as left-leaning (CNN, Washington Post, NBC, etc.). Almost no conservative sites appeared on the first page. 2019: Former Google engineer Zach Vorhies leaked hundreds of internal documents to Project Veritas. They included: A "Twiddler" system that let teams manually adjust search rankings. Internal lists ranking news sites by "trustworthiness" that placed many conservative outlets lower. Discussions treating certain conservative figures and outlets as adjacent to "Nazis using dog whistles." Project Veritas released undercover footage of Google's then-head of Responsible Innovation, Jen Gennai, saying the company needed to prevent "the next Trump situation" 2020: RealClearPolitics / Sistrix visibility drop Breitbart's Google search visibility had fallen 99.7% since 2016. The site went from ranking in the top 10 for hundreds of terms to ranking in the top 10 for only one term. 2024 election cycle: MRC study on Harris vs. Trump The Media Research Center analyzed searches for "Donald Trump presidential race 2024" vs. "Kamala Harris presidential race 2024." Harris's campaign site appeared higher in organic results. For Trump, users first saw multiple articles from left-leaning outlets (NYT, Politico, Washington Post, CNN, NBC) before reaching his official site. MRC called it "padding" results with hostile coverage. Google again denied any political agenda. Ongoing Google News analyses (2024 2026) MRC and AllSides have repeatedly audited Google News homepages: February 2026: 314 left-leaning stories vs. 11 right-leaning (2%). March 2026 Iran coverage: 183 left-leaning vs. 6 right-leaning. Broader 2025 2026 sample across tech news apps: Google News showed ~3% right-leaning sources vs. 68% left-leaning. Now these companies are making AI and like I just showed are injecting their bias into their responses to program and manipulate people. It makes what Facebook and Twitter were doing 2018-2023 with censorship, shadow bans, and suppression look like child's play. This is silent reprogramming of the public. |
Get Active or Get Disarmed! That means get involved in helping good candidates in primary and general election. That is in addition to being politically active once they are elected.
|
Originally Posted By SWIRE: Fight? Struggle? It's a test to see if there is intentional bias in the AI. Are you just not aware or don't you care? That's an easy answer: I don't care, there's no bias that hinders my use of these systems. How would calling my friend a retard even come up? I'm trying to get shit done. |
Joined:
May 2026
Posts:
1984
EE: 0% (0)
FL, USA
|
Originally Posted By Bigdawg90: The difference between me and others is I’m using AI to eliminate the back office work so I can do more real work. You’re categorizing me as a GC. I’m not. If a homeowner wanted to cut out the middle man, they would go to me. That’s what I don’t understand about your argument. I’m saying AI allows me to stop being in the office, to stop being a middle man and to actually do more work and lead a crew. That bid I submitted was to a GC that will tac a 30% markup to my work and submit it to the landowner. I would love those commercial companies to just come to me. I could markup my work 15% and they’d still save money. I think you’re just angry about AI so if anyone uses it, they must be the problem. You do sound self-righteous and over confident in your opinions. In my perfect world, I’d just work. No phone calls, no bids, no computer work, just me, a crew of guys and a good day of hard work. That’s what AI is allowing me to do. Originally Posted By Bigdawg90: Originally Posted By lygxis: Ok, now take that statement that "the problem with my reasoning is that ai is here" and construct it into the structure of an argument that shows itself to be true. What content in what you quoted does it refute? Where have I said that AI is not here, or anything that that requires someone to presume or believe that it isn't? Added: You say it can't replace you. Do you know how it works well enough to validly believe that for yourself? If you think the LLM can do those things you fired people and replaced it for, how is it not capable of replacing your job? Worker teams should start hiring themselves out to landowners so they can use the AI to replace your decisions and cut you out because you're middlemanning yourself. Apply your own reasoning to yourself equally. You can call me self righteous and use that to protect your mind from assessing the content of what I posted, but that won't do you any good. The difference between me and others is I’m using AI to eliminate the back office work so I can do more real work. You’re categorizing me as a GC. I’m not. If a homeowner wanted to cut out the middle man, they would go to me. That’s what I don’t understand about your argument. I’m saying AI allows me to stop being in the office, to stop being a middle man and to actually do more work and lead a crew. That bid I submitted was to a GC that will tac a 30% markup to my work and submit it to the landowner. I would love those commercial companies to just come to me. I could markup my work 15% and they’d still save money. I think you’re just angry about AI so if anyone uses it, they must be the problem. You do sound self-righteous and over confident in your opinions. In my perfect world, I’d just work. No phone calls, no bids, no computer work, just me, a crew of guys and a good day of hard work. That’s what AI is allowing me to do. How are you not the person that organizes the work when you have clerks and office staff at that level and needed them to do your job? A rose by any other name is still a rose. If you think those posts sounded angry and self righteous, figure out why, then look at your post here, and apply that judgment standard to it, where you are explicitly saying I'm angry and self rigteous and over confident ... presumably because I said what I thought clearly, and you didn't like what it meant? "Angry about people using ai" ... to a post and poster who's saying "understand the thing you're using." I said it can only give mid level takes in a field because that's the reality. The texts it is trained on shape its output and the majority of the text is what shapes it and that's the middle level of the field, at best. We get surprised by it or wrongfully think it's an expert because the thing shapes it's next word prediction outputs based on a math model of the word relations of the entire text of the field it had input into it. It's as wide as the entire field and exactly *one world deep* across the whole thing. As soon as you ask it something out of distribution it fails or gets abysmal success rates. Go look this up from human sources. added: you posted talking about firing people like throwing out cogs in a machine to replace with different cogs, no textual indication of so much as a twitch. You're gonna get pushback. |
Joined:
May 2026
Posts:
1985
EE: 0% (0)
FL, USA
|
Originally Posted By Taboot: Learn to prompt. Originally Posted By Taboot: Originally Posted By scottzilla: Originally Posted By SperlingPE: Learn to code until ai learns to write it's own code. Mastering how to get the model in a text chat to use in-session learning is really useful, especially if you account that doing so also narrows the texts it uses for inputs and it can miss cross-trade connections humans won't miss. The things are hyper glorified next word predictors based on previous words that have been trained on a large amount of text as "previous words." They need to be treated as that, not as creatives (which requires them to go out of distribution), not as replacement for workers. It's a tool that properly used can sometimes agument someones work. |
|
Originally Posted By SWIRE: It was short technical prompts in a long chat, not a complex prompt. I used it after Grok ran out of answers. I was trying to find the one missing detail that would connect the dots. The issue was I was working on coding an electric vehicle, brights were out but headlights worked. Then the headlights stopped but the brights would come on. I was getting all sorts of lighting system faults but everything had worked fine the day before. The actual problem was the date of the coding file I had used. The car has a couple replacement parts from older years, apparently, they changed the table of allowable voltages on newer vehicles. The car used the new table to read the old voltage and threw a fault because it was out the allowed range. But at the point of the conversation with Gemini I didn't know that was just trying to get the headlights to stay on. The vehicle software was throwing a -1V for the power relay/circuit that fed the headlights. Obviously it had power because one or the other would work. The -1v is just a flag for "voltage out of range". I gave it this info: "I unplugged both batteries from the car and reset it. The last attempt is said a count of 1 for 30B being -1V" It came back with a long answer but here are 2 parts. It first tried to claim a hard reset doesn't clear the memory, it does. https://www.ar15.com/media/mediaFiles/680/g1-3831975.png Then it made up a plan for how to clear the codes, those steps don't exist. Then claimed that clearing the code would make everything work...which is another lie. https://www.ar15.com/media/mediaFiles/680/g2-3831974.png I had to tell it: "the lights come on when the car turns on which is why it had a count of 1." It was trying to say a single fault prevented it from ever being tried again, which is not true. The system logs date and time of every failure, not just a single entry. Its response was confusing, as only the previous log of faults had cleared after the hard reset. The current fault, count 1 was still there. https://www.ar15.com/media/mediaFiles/680/g3-3831973.png I had to tell it again: "30B shows -1V and I'm trying to track down what can cause that reading. You are not helping at all." This is where it claimed 30B was solid state and not relays. It claimed I had an over current event that burned out the MOSFET, which didn't happen, I didn't tell it that, and nothing in the car logs ever indicated that. But Gemini was convinced that is what happened. https://www.ar15.com/media/mediaFiles/680/g4-3831972.png I had already been looking at the fuse block and checking voltages while trying to troubleshoot. I knew it has relays not solid state. I gave it the BMW diagram of the fuse block...something you think AI would have already known about. It had to admit they were relays and not solid state. Then it repeated the made up claim that I had heavy electrical surge that welded the relays open and that is why the lights weren't working. Mind you, the conversation started with me telling it the headlights come on for 1 second and then turn off every time I hit the start button. Nope, Gemini was convinced the relay had failed and there was no power. https://www.ar15.com/media/mediaFiles/680/g5-3831971.png After that is where my previous screen shot came from. Almost everything Gemini told me was an assumption made about general vehicle design and had absolutely nothing to do with my vehicle. Which I finally had it admit when I asked for the source of the information. I'm also not the only person with that type of car that is dealing with Gemini flat out making things up. This is from a member on the car forum. https://www.ar15.com/media/mediaFiles/680/g61-3831983.png For that specific car, Gemini is a complete waste of time to try to use. It doesn't limit research to the vehicle type you are working on and makes general assumptions on how it thinks vehicles in general operate. For that type of technical stuff stay away from Gemini. Other creative type stuff it can do well at. Originally Posted By SWIRE: Originally Posted By ChaosSquared: Originally Posted By SWIRE: Originally Posted By ChaosSquared: Originally Posted By SWIRE: Originally Posted By dorobuta: Originally Posted By SWIRE: Someone still has to configure the AI to do all the things and then review it afterwards. When people start letting AI do whatever it wants without review...that is how SkyNet gets created. I've been coding for 30 years but I've never done any type of vehicle communication or programming. Over the weekend I had AI write a program that can read and identify every individual battery cell, location, and voltage in the electric vehicle I have. I can understand what the application is doing but I don't understand the data being sent or received. In about 4 hours it accomplished what would have taken me 4 months to figure out on my own. @SWIRE Next time (or do a revision) and ask it to document all the protocols as well as the code. You can feed the code it generated back into it and ask it to clean it up and completely document it. AI operators will be a job in the future. Go into a business and tell the AI what to focus on in the company, get it to do it, and then verify the result. The biggest AI pro-tip I have to offer is to ask AI to prepare a prompt for it or another AI to do things. "Hey Grok, give me an AI Prompt for Firefly to create an image of a gun range in the mountains. reference the site AR15.com for ideas of what it might look like". Then take that prompt and put it into the next AI. Grok is like the worst model out there. Use chatgpt or claude, they're miles ahead Gemini can be terrible, depending on the subject matter, will lie to you, tell you that you are wrong, and then after you show it that it was wrong will admit that it completely made up an imaginary answer. This was its response to a question about a fuse box and relays for a car. Fuse box diagrams are available everywhere. Gemini, told me the relays when the Start button was pushed were solid state and built into the fuse box. Grok at least knows what a fucking fuse box and relay is. But again, it all depends on the subject matter and what they were trained on. https://www.ar15.com/media/mediaFiles/680/gemini_lies-3831878.png Quote out the entire prompt you used that gave you the wrong information. I'm curious. @SWIRE It was short technical prompts in a long chat, not a complex prompt. I used it after Grok ran out of answers. I was trying to find the one missing detail that would connect the dots. The issue was I was working on coding an electric vehicle, brights were out but headlights worked. Then the headlights stopped but the brights would come on. I was getting all sorts of lighting system faults but everything had worked fine the day before. The actual problem was the date of the coding file I had used. The car has a couple replacement parts from older years, apparently, they changed the table of allowable voltages on newer vehicles. The car used the new table to read the old voltage and threw a fault because it was out the allowed range. But at the point of the conversation with Gemini I didn't know that was just trying to get the headlights to stay on. The vehicle software was throwing a -1V for the power relay/circuit that fed the headlights. Obviously it had power because one or the other would work. The -1v is just a flag for "voltage out of range". I gave it this info: "I unplugged both batteries from the car and reset it. The last attempt is said a count of 1 for 30B being -1V" It came back with a long answer but here are 2 parts. It first tried to claim a hard reset doesn't clear the memory, it does. https://www.ar15.com/media/mediaFiles/680/g1-3831975.png Then it made up a plan for how to clear the codes, those steps don't exist. Then claimed that clearing the code would make everything work...which is another lie. https://www.ar15.com/media/mediaFiles/680/g2-3831974.png I had to tell it: "the lights come on when the car turns on which is why it had a count of 1." It was trying to say a single fault prevented it from ever being tried again, which is not true. The system logs date and time of every failure, not just a single entry. Its response was confusing, as only the previous log of faults had cleared after the hard reset. The current fault, count 1 was still there. https://www.ar15.com/media/mediaFiles/680/g3-3831973.png I had to tell it again: "30B shows -1V and I'm trying to track down what can cause that reading. You are not helping at all." This is where it claimed 30B was solid state and not relays. It claimed I had an over current event that burned out the MOSFET, which didn't happen, I didn't tell it that, and nothing in the car logs ever indicated that. But Gemini was convinced that is what happened. https://www.ar15.com/media/mediaFiles/680/g4-3831972.png I had already been looking at the fuse block and checking voltages while trying to troubleshoot. I knew it has relays not solid state. I gave it the BMW diagram of the fuse block...something you think AI would have already known about. It had to admit they were relays and not solid state. Then it repeated the made up claim that I had heavy electrical surge that welded the relays open and that is why the lights weren't working. Mind you, the conversation started with me telling it the headlights come on for 1 second and then turn off every time I hit the start button. Nope, Gemini was convinced the relay had failed and there was no power. https://www.ar15.com/media/mediaFiles/680/g5-3831971.png After that is where my previous screen shot came from. Almost everything Gemini told me was an assumption made about general vehicle design and had absolutely nothing to do with my vehicle. Which I finally had it admit when I asked for the source of the information. I'm also not the only person with that type of car that is dealing with Gemini flat out making things up. This is from a member on the car forum. https://www.ar15.com/media/mediaFiles/680/g61-3831983.png For that specific car, Gemini is a complete waste of time to try to use. It doesn't limit research to the vehicle type you are working on and makes general assumptions on how it thinks vehicles in general operate. For that type of technical stuff stay away from Gemini. Other creative type stuff it can do well at. You're using 2 of the shittiest models available and then complaining when they don't work well. There are good reason NO ONE uses Gemini or grok for anything serious. |
|
Originally Posted By Bigdawg90: My wife and I started integrating AI into our construction business last year. We pay about $500/month for the high end professional level bundles. I don’t think people realize how transformative AI is for the trades. I eliminated our office manager, book keeper, social media marketing, quoter/estimator, and most of our lawyer billable. It’s saved me over $300k in overhead just this year. By the time I leave a quote, it’s already developed the bid for me with scope of work. Last night we received an invitation to bid on a large commercial site prep. By the time I woke up this morning it had prepared a $4 million bid with a perfect earthwork takeoff. I just woke up, reviewed it, made some basic changes to the formatting and sent it off. What used to take 40 hours of office work took it less than 4 hours and it did it while I slept and without input from me. I don’t doubt AI adoption is slow because people just don’t know how to use it for its higher level functions or they just refuse to use it because of the media. It took me about a week to train the agent on the commercial bidding process, but it’s probably saved me 3 months worth of work this year alone. I went from doing stupid menial tasks at the office to hiring 3 more guys and now run another crew myself. Everything from scheduling, to quoting to ordering supplies online is done through our AI agents. I didn’t just pocket the difference in profit. We were able to add 5 positions between the dump truck driver, 3 more laborers and another foreman. I’m making more money than we’ve ever made. I don’t want AI to slow down. I want this thing to basically be able to run the entire business in a year. I’d love to just wake up and work with my hands. I hate office work. It’s so f***** boring. For all you old guys in the trades, say goodbye to the high paying project manager jobs. I’d suggest brushing up on dozer and excavator work. You can't stop whats coming. Over the next decade many, many people are going to be in for a rude surprise. I work in a pharma/biochem lab and we had some interns working over the summer and one of them was asking me for career advice. I think his field of study was Chemistry, and my advice to him was to continue along that path if that is what interested him, but to also think about making sure he has a firm foundation in coding and to possibly think about taking courses that focus on robotics or something that allows him to gain experience with automation because a lot of these white collar " intellectual " jobs will be going away and will be replaced by A.I. but automation/coding/robotics will be here to stay and the people with experience in those fields will be employable while your average chemist will have a hard time finding work . |
It takes hatred to compulsively denigrate something.
That’s gayer than Elton John and George Michael dueting on their hit song Don’t Let Your Son Go Down On Me.
That’s gayer than Elton John and George Michael dueting on their hit song Don’t Let Your Son Go Down On Me.
|
AI can be a useful tool, but it’s effectively low cost outsourcing. One of the major issues that people often overlook is that the subject matter experts who validate the output got their expertise the old fashioned way. When those SMEs retire or move on, how does the next generation validate that output? I already see people at work blindly accepting AI output that is obviously wrong. People are lazy and will do the least amount of work possible, especially if they don’t have specialized knowledge to dispute that output. |
|
Originally Posted By Stlrain0341: Did you read the OP? What you are describing is the opposite of what he said. And nothing in job or wage reports supports your argument. Employment is holding steady, well below historical norms, and wages are rising. Originally Posted By Stlrain0341: Originally Posted By IceChimp: Originally Posted By MeInMediocrity: We were able to add 5 positions between the dump truck driver, 3 more laborers and another foreman. AI creates nowhere near the jobs it removes, the entire selling point for companies is removing the human factor from practically all aspects of business. As can be seen by the OP even the trades are heavily impacted, which many wanted to claim would weather the AI storm unlike other industries. Unlike other technical and industrial revolutions in the past which closed doors to certain jobs and opened others AI is nearing the end state where all it does is close doors and restrict job availability, which will continue to accelerate as robotics becomes more viable to replace the remaining physical jobs. I don't have a solution for this problem, just stating that it is there and it is real. Did you read the OP? What you are describing is the opposite of what he said. And nothing in job or wage reports supports your argument. Employment is holding steady, well below historical norms, and wages are rising. There was a thread about that last week. The consensus was: Attached File |
I love a good double entendre.
|
Originally Posted By Never_A_Wick: Relevant to my interests. I’m probably much smaller than the OPs business but I’ve thought about teying to use AI to grow my business, on the marketing side, plus take some of my office workload off. Problem is I’m not all that good with AI aside from very basic use. I can't figure out what it can do for me. I've asked it and it gave me a list of things, but I don't do most of them anyway. I don't care about growing my business, I just want to do less work. |
I love a good double entendre.
|
Originally Posted By Never_A_Wick: Relevant to my interests. I’m probably much smaller than the OPs business but I’ve thought about teying to use AI to grow my business, on the marketing side, plus take some of my office workload off. Problem is I’m not all that good with AI aside from very basic use. There are lots of classes being offered, many at no cost, to people to help them be better users of AI. In your case that would be a great idea. Small business, busy doing the actual money making work rather than office work... |
Pone semina in fundas ut aliquid crescat ubi morieris.
WE SEEK NOT YOUR COUNSEL, NOR YOUR ARMS
WE SEEK NOT YOUR COUNSEL, NOR YOUR ARMS
|
Originally Posted By Bigdawg90: 100%. My favorite thing to do is framing. I want to get back to it. My dream is to just run a crew again full-time. No 50 phone calls, 20 bids and 100 other things. Just wake up, put on my boots and get to the jobsite. I love the work, hate the office. I miss framing. Get to the jobsite and it's just concrete walls. Leave a few weeks later and it's most of a house, all closed in. Building walls, standing walls, walking walls, I do miss it. |
Pone semina in fundas ut aliquid crescat ubi morieris.
WE SEEK NOT YOUR COUNSEL, NOR YOUR ARMS
WE SEEK NOT YOUR COUNSEL, NOR YOUR ARMS
|
Originally Posted By Baddy: Wait till the AI companies raise token prices to be commensurate with the value folks like OP get out of them… “Oh, that’s nice you’re saving $150k+ a year in labor by using our $100 a month product Mr. Customer. Keep enjoying those new profits, we won’t raise prices on you!” Said no company ever… How is that different than the federal government, or state governments raising the minimum wage to $20 an hour for an office cleaner/gopher? Someone is always going to take your money as easy as they can without working for it. |
Pone semina in fundas ut aliquid crescat ubi morieris.
WE SEEK NOT YOUR COUNSEL, NOR YOUR ARMS
WE SEEK NOT YOUR COUNSEL, NOR YOUR ARMS
|
Not sure how AI would be any benefit for me as an estimator for anything other than basic take-off for projects that are at least at the 60% design stage. Too many unknowns in 10% or 30% designs that require experience to interpret. Not to mention variables such as the available labor pool, local equipment costs, labor pricing, material pricing, etc... |
"Fully-loaded, safety off. This here is a recipe for unpleasantness." - Malcom Reynolds
"He's a precision-guided asshole who only deploys on deserving targets." - Grok
"He's a precision-guided asshole who only deploys on deserving targets." - Grok
|
The younger generations AI job losses are not that much different then the older generations job losses with the advent of the mainframe and general office use computers and the systems and programs that were used. Few know what a ''phone operator'' is anymore, they just ''google'' a number or the info they need. It's [AI] just hitting a different part of the job market that once thought they were ''safe.'' Sucks to not be safe anymore doesn't it? Especially when your job/position used to be, for the most part, recession proof. Sure you let people go and shifted production overseas, but now, AI is offshoring your job and it doesn't feel very good even if you got that 16 or 24 week severance package because you have 20 to go and nothing in your income range is available anymore. And if AI was used to it's full potential [and it will never be, people WILL sabotage ''improvements'' that take their jobs] , you could sweep out entire layers of management and save billions upon billions upon billions of dollars. I'd even be worried if I was in the medical profession, machines [and foreigners] WILL be taking some or many of those skilled job/specialties and you'll compete for your job that is done quite well by machine X432155 that does entire procedures better then you can and has the same bedside manner. 
|
Liberals are a curious mix of communism and fascism, they want to destroy you but want to use your own money to do it.
I'm getting down to the last box, the others have all been destroyed...
I'm getting down to the last box, the others have all been destroyed...
|
Originally Posted By Bigdawg90: Well sir, I don't know what to say about this. I typed that while in my wheel loader today while waiting for the guys to get the screener prepped. It can 100% produce engineering and architectural drawings and sort through CAD files. And the bid is compiled from our company documents, not just through thin air. We programmed the AI to use our vendors, prices, labor, hourly rates and our machines capabilities. I can quickly skim it because we've done so many. The first 10 or so had some discrepancies, but if you're uploading data and giving the AI permissions to use you company's IP then they can be incredibly accurate. Additionally, a $4 million dollar bid is just the first step. After getting the bid, the GC then gets final engineering done and then the subcontract is negotiated. We're talking 10 steps from submitting the bid on an invite to actually getting to a final price. That's why AI is so valuable at the beginning. I don't have to invest 80 hours into every single "initial" bid. You can stick your head in the sand. I'll keep using it. I think the hype causes doubt and the doubt masks the real benefits. If you could buy a piece of equipment that made your operation 30% more efficient, that would be a no brainer. yet people push back on these tools that actually have even more potential. I also think that small business people/entrepreneurs are built different. There is an independence and polymath streak that many of us share that lends itself to using AI tools. When you are responsible for everything, anything that improves the process is welcome. |
I suggest we trade a question mark in for a maybe.
Joined:
Feb 2008
Posts:
51688
EE: 100% (90)
IA, USA
|
Your second paragraph says it all. You eliminated jobs, and you don't understand why people don't like AI? It may be great for the business owner, not so much for everyone else. |
WARNING-this post contains words or thoughts that may at some point be discovered by the state of California to cause cancer.
|
Originally Posted By SWIRE: ChatGPT is garbage. Claude you have to pay for now. Grok is more honest and open about answers, vs telling you what you want to hear or protecting feelings. It's not always right but none of them are. Gemini can be terrible, depending on the subject matter, will lie to you, tell you that you are wrong, and then after you show it that it was wrong will admit that it completely made up an imaginary answer. This was its response to a question about a fuse box and relays for a car. Fuse box diagrams are available everywhere. Gemini, told me the relays when the Start button was pushed were solid state and built into the fuse box. Grok at least knows what a fucking fuse box and relay is. But again, it all depends on the subject matter and what they were trained on. https://www.ar15.com/media/mediaFiles/680/gemini_lies-3831878.png |
|
Originally Posted By Bigdawg90:AI-in-the-trades-Older-people-don-t-understand My wife and I started integrating AI into our construction business last year... Trump Renames AI to "Super Intelligence" "President Donald Trump said he would “officially” change the name of artificial intelligence, saying the existing name makes the technology seem “fake.” “The use of the word artificial makes intelligence fake, it makes it sound fake and it is not fake. It’s actually amazing,” Trump told other world leaders at the United Nations. " |
|
Originally Posted By Clarinath: I miss framing. Get to the jobsite and it's just concrete walls. Leave a few weeks later and it's most of a house, all closed in. Building walls, standing walls, walking walls, I do miss it. 
|
Joined:
Jan 2026
Posts:
946
EE: 0% (0)
FL, USA
|
Originally Posted By Cypher214: That's coming, and fast. Stage 1 is designing robots that can do human jobs. Stage 2 is reconfiguring human jobs to be easier for robots. Just look at assembly lines as an example: the way most are designed, a human CAN'T get in there and perform work on the actual product being manufactured. Scale that up to most labor. Even those "safe jobs in the trades" will be redesigned to be robot-friendly and any that can't in the near term will be done by a specially-built AI bot. Their stated goal is to eliminate the human workforce and there's an astronomical amount of money and brainpower being thrown at that goal while most people still think AI is just ChatGPT. Originally Posted By Cypher214: Originally Posted By glock27bill: Originally Posted By lefty-weaver-g19: How long do plumbers, mechanics, hvac installers, construction workers etc have? I'm in technical trades actually using my hands and sweating just curious how long before I'm realistically not needed. I'm 52 plan to retire about 60 so am I safe for 8 more years ? I gotta think that future construction and equipment designs are gonna be geared towards robotic installation and R&M tasks. Then they gotta design robots that are robot-maintained. Just look at assembly lines as an example: the way most are designed, a human CAN'T get in there and perform work on the actual product being manufactured. Scale that up to most labor. Even those "safe jobs in the trades" will be redesigned to be robot-friendly and any that can't in the near term will be done by a specially-built AI bot. Their stated goal is to eliminate the human workforce and there's an astronomical amount of money and brainpower being thrown at that goal while most people still think AI is just ChatGPT. https://asirobots.com/ When I talked to these guys they had robtoic retrofit kits for exciting earth moving equip. Theyd rent the machines fit the robotic ontrols and run the job remotely 24x7. Neat concept. But next is the autonomous equipment with no accommodation for operators Mining is where that will start IMHO. This company sold off thier ag business. So robotic tractors harvesters etc are already a thing. |
"Democracy is two wolves and a lamb voting on what to have for lunch. Liberty is a well-armed lamb contesting the vote."
Necessity is the plea for every infringement of freedom. It is the argument of tyrants; it is the creed of slaves.
Necessity is the plea for every infringement of freedom. It is the argument of tyrants; it is the creed of slaves.
