3 ms·
I sell and run electrical work and I don’t see any good use for LLMs for what I do on a day to day basis. LLMs don’t understand constructions drawings so they’
by quickthrowman 1y ago
I sell and run electrical work and I don’t see any good use for LLMs for what I do on a day to day basis.
LLMs don’t understand constructions drawings so they’re no help when it comes to doing a takeoff. Construction specifications are already well-organized and able to be searched so LLMs don’t help there.
They can’t synthesize information from different sources (words from a person spoken on a phone conversation, napkin sketches, information that is embedded in an electricians head about a specific facility, etc) or coordinate multiple parties through a variety of communication methods (email, text, phone calls, RFIs, in-person meetings, etc)
About the only use I’ve found for them in my line of work is cleaning up data from outside sources. YMMV. Construction is a very relationship and trust based business that has been around longer than almost every other profession.
- tsunamifury 1y ago“The internet is a passing fad”
- quickthrowman 1y agoI don’t think LLMs have zero applications, I just haven’t found any for my specific use cases. If someone is able to figure out a way for LLMs to interpret construction drawings, it would be immensely useful, I’d be able to price substantially more work.
- do_not_redeem 1y agoAh yes, the same quote often used by cryptocurrency enthusiasts. "One time a new tech succeeded despite skepticism, therefore all new tech will succeed despite skepticism"
- tharmas 1y agoDistributed ledger/blockchain is an interesting idea but far too energy intensive to be of practical use. And then theres the whole concept of what currency actually is.
- eiishiiiou 1y ago"[Distributed ledger/blockchain] | [LLM] is an interesting idea but far too energy intensive to be of practical use."
- do_not_redeem 1y ago"And then theres the whole concept of what intelligence actually is." Thank you, I'm glad someone understood my comment at least.
- sleepyguy 1y agoI think back to a 1991-1992 when someone stood over my shoulder watching and telling me what a waste of time it was and who they hell is ever going to use it :)). I remind them from time to time and we have a good laugh.
- tharmas 1y agoYes, totally agree. LLMs are largely pattern recognition. They cant reason.
- orwin 1y agoThat's arguably the most interesting about the discussion on LLMs. Can't they reason? If they do, reason is an emerging fonction. 2 years ago, before inference was truly added to LLMs,a Honk Kong university paper wrote that ChatGPT 3.5 reasoning was correct 64% of the time on a specific task (you asked it to make the same reasonning 100 time, 36% of the time it would be wrong for no reason). I'd like to see how modern LLMs with added inference matrices and a lot of helpers before the actual transformers perform on this test. If a consensus arrives in 5 years and we decided that yes, LLMs can in fact reason, reasonning would be an emerging capacity, and that would be incredibly interesting.