Welcome! Here are the website rules, as well as some tips for using this forum.
Need to contact us? Visit https://heatinghelp.com/contact-us/.
Click here to Find a Contractor in your area.

If you are not using AI....

Options

to assist in calculating difficult loads, you should ask yourself why. Conversational input with verifiable, structured, data output. Here is a real example I did last night after putting the kids to bed. Took many hrs out of the work required to achieve results like this.

Comments

  • ratio
    ratio Member Posts: 4,065

    Careful. Like Googling spelling, "AI" is based on what everyone else is doing, not what is correct. It's a step or two before "trust but verify".

    Computers don't think, & there aren't any signs that they ever will. People are just really bad at telling the difference. They can, however, add really well, provided they get the right numbers.

    LRCCBJbburdAlan (California Radiant) Forbes
  • hot_rod
    hot_rod Member Posts: 25,344

    someone showed some engineering design done with AI a few years back and it had a number of mistakes. Maybe AI becomes more accurate over time as it gathers more data?

    Bob "hot rod" Rohr
    trainer for Caleffi NA
    Living the hydronic dream
  • Harvey Ramer
    Harvey Ramer Member Posts: 2,264

    It sure does. Whatever you knew about AI last week is outdated this week. I have been running AI heat load simulations for close to a year now. Initially it was very bad. Now it's looking very good. Still need to verify the output obviously.

  • EBEBRATT-Ed
    EBEBRATT-Ed Member Posts: 18,235

    One of they guys that ran a refrigeration supply house told me that in my area the cooling load is typically 1/2 the heating load that of course was residential. Only a rule of thumb you still have to check but there is some truth to it.

  • Harvey Ramer
    Harvey Ramer Member Posts: 2,264

    Things get a lot different when you are doing DOAS targeting positive pressure values and ACH. Not to mention, non-standard internal loads.

  • smore1945
    smore1945 Member Posts: 18

    I've had good results by asking very specific incremental questions that lead up to the answer for the bigger question. You can sort of hack the AI into showing its work and giving correct answers to smaller problems. Then you can ask it to consider it's previous answers when answering the bigger question. But yes, definitely be careful! I was working on some website integration coding stuff (of which I have zero knowledge) and using ChatGPT to guide me through the process. I had some initial success, but once it got really technical it led me on an hours long circular wild goose chase. One of the harder things to calibrate for initially is unbridled confidence with which answers are given. It's like your buddy who "knows" everything, but really knows pretty much nothing 😂

    PC7060
  • Kaos
    Kaos Member Posts: 676

    What I have found is most are generally right but when they are wrong often they are subtly wrong. This makes it hard to trust unless you are in the specific field.

    For example, I as asking about some building question regarding exit stairs. It gave me the correct answer on size and shape of winders from our code but incorrectly said they can be used in an exterior stair (exterior stairs can't have winders under our code).

  • RayWohlfarth
    RayWohlfarth Member Posts: 1,789

    I use AI all the time to give me suggestions and help me double check my numbers.

    Ray Wohlfarth
    Boiler Lessons
    Harvey Ramer