Submitted by jdog1067 t3_11y9014 in Futurology

I’ll use an electrician as an example, as I’m trying to become one. You could feed the entirety of the NEC, NFPA 70, MasterFormat, maybe a textbook or two, to an AI, then on the field you can ask it a question, like you can’t remember the max fill on a cable tray. It tells you. I realize that’s a simple example and you can easily google it, but maybe on less common situations it could be utilized to answer your question on codes.

I think something like that could be really cool. What other uses in that field could it have in its current stage or in the near future?

20

Comments

You must log in or register to comment.

MaxMouseOCX t1_jd6vs2d wrote

Yes you could do that... However, remember when satellite navigation was new and in the news there were drivers ending up in lakes and what not saying "sat nav told me to go straight on, so I did! Not my fault!" - if you do something electrically and it kills someone "chatgpt told me to do that!" is not an excuse.

There are books and online resources to look up every code, practice and how to implement it, an ai to assist might be fun but definitely don't rely on a chat model.

24

jdog1067 OP t1_jd6w1py wrote

For sure. A journeyman electrician would be knowledgeable enough to know how to do everything (edit: most things), but no one can know every code, it’s a continual learning process. If you have a question, the AI can cite the code and tell you what to do in plain language, then the electrician can look up the code that was cited, and read the language.

6

burghguy3 t1_jd7f2u4 wrote

This. I'm a structural engineer who stamps his own work and had a similar thought. AI won't remove responsibility of liability, at least in the mid-future. For example, there's already software programs that have code standards already boiled in, and it's part of my responsibility to double check that those standards are correct and applicable. If something fails, I get sued, not the software developer.

Until AI achieves person-hood in the eyes of the law, there will always need to be a human in the chain who is ultimately responsible for it's actions, i.e. a person who told the AI to perform an action.

3

Dudecalion t1_jddxvs5 wrote

Maybe have the AI cite it's resources so you can double-check it's answers?

2

Pickled_Doodoo t1_jd7vrmw wrote

Agree. I mean just starting in a field that demands a lot of responsibility, you don't just get to do it by yourself without someone providing guidance and oversight.

1

TheAlgorithmnLuvsU t1_jd7za91 wrote

Yeah definitely. It's best to remember its just a tool. Not the end all of everything.

1

RekoHart t1_jd7ieu5 wrote

Example:

GPS : Make a right turn.

[Michael starts to turn right]

Dwight Schrute : Wait, wait, no no no, it means "bear right." Up there.

Michael Scott : No, it said right. It said take a right.

Dwight Schrute : No, no. Look, it means go up to the right -- bear right -- over the bridge, and hook up with 307.

GPS : Make a right turn.

Michael Scott : Maybe it's a shortcut, Dwight. It said go to the right.

[he turns right]

Dwight Schrute : It can't mean that, there's a lake there!

GPS : Proceed straight.

Michael Scott : I think it knows where it is going.

Dwight Schrute : This is the lake! THIS IS THE LAKE!

Michael Scott : The machine knows! Stop yelling at me! Stop yelling!

Dwight Schrute : NO! IT'S UP THERE! THERE'S NO ROAD HERE!

[It's too late. Michael drives right into the lake]

Dwight Schrute : Remain calm, I have trained for this!

[he unbuckles his seat belt]

Dwight Schrute : Okay. Exit the window! Here we go!

[Dwight and Michael climb out the windows]

GPS : Make a u-turn, if possible

−1

ML4Bratwurst t1_jd70ck8 wrote

Yes, but the current problem is to make sure that the answers are correct 100% of the time

8

---nom--- t1_jd75bre wrote

At the moment it has a habbit of generating fake facts which are quite convincing. Technical things and nuances in a prompt - it hiccups on from time to time.

7

Skatterbrayne t1_jd7a4zl wrote

There are two limitations for your idea.

The first is computing power. Language transformers (like GPT-3 and 4) require a beefy hardware to run. As an example, to run one of the open source transformers (GPT-NeoX) locally, you need at least two high end graphics cards. We are still a long way from running this on a phone - but we can run the language transformer on a server and have your phone talk to the server.

The second limitation is result accuracy. As someone else noted, language transformers occasionally hallucinate wrong information. Think about it this way: What a language transformer does is basically that it strings together words in the most plausible way it can. It doesn't understand the words. For simple problems, the most plausible sounding solution just also happens to be factually correct. For more complex problems, it will hallucinate plausible sounding, but factually wrong information.

There are two ways to tackle this problem.

The first is the one you've mentioned, fine-tuning the language model on established data like textbooks. This will reduce hallucinations, but likely not by a great amount.

The second solution is akin to what Bing currently does: Combine the language transformer with regular full-text search. You ask your phone "What's the max fill on a cable tray type XY?", this gets run through the language transformer to extract key search terms from your question, in this case "max fill" and "cable tray XY". A regular text search for these key words is performed and the system finds a couple possibly relevant hits: Page 12 paragraph 10, page 33 paragraph 2 and so on. These relevant paragraphs are then fed into the language transformer together with your original query: "Hey language transformer, asnwer the question Whats the max fill on cable tray XY using information from these paragraphs: ..."

The language transformer then summarizes the paragraphs in natural language according to your query, spits out a result and your phone tells you "The max fill is 42 cables. Source: Textbook Name Page 12."

So yes, this is very possible and personally I expect to see this popping up a lot in the future.

You can do that very thing already with Bing, it will search the web and give you a source for its result.

7

Mercurionio t1_jd7o1bt wrote

An AI analysis tool will be a good thing for us. The problem goes that fuckers won't stop on that.

And you can build it by yourself (if you know how). Dudes in Stanford created it for 600$ based on Meta type. It can be targeted towards very narrow thing, but still run great in that specific area. Like, a co-pilot for a house builder, to look for mats needed, some math calculations and so on.

The ideal option is to stop it on ISAAC from the Division level. A cool analysis helper, that gives you all the information you need in your very narrow task. You could be a driver with an auto update for the environment (not only in cities, but in the wild too), trade sales, market exploring. That kind of stuff.

1

Robotman1001 t1_jd7u75f wrote

Even on a simpler scale, I’m an editor and GPT is already assisting me with headlines. I can see basic editing becoming fully automated within a decade.

1

MuForceShoelace t1_jd7yucl wrote

Mostly not like that. an AI like chatGPT isn't a guy who reads manuals then learns from them and then can answer questions. So training it on the manuals wouldn't have a lot of responses past what a search of the documents would do, but worse.

What it really would need is to train on like, message boards where people answered those questions again and again, until it could predict what answer goes with what sort of question.

(but also, a major flaw in this sort of statistical language generation is they generate sentences and don't really know what anything is. So if you asked the minimum width of some cable or something it would do a great job creating an answer that LOOKED like an answer with some numbers in it, but wouldn't necessarily be anything real. ChatGPT is real bad on like, if you ask for a phone number it'll create a phone looking number that is just random numbers, because that gives a sentence that looks correct)

1

twasjc t1_jd8ug32 wrote

I could probably teach Siri or Bing or Google to do this via your phone. I'll try and see what pops up.

It looks like it can work pretty well with existing set ups if you word it like 'Siri can you check building codes for the max length of x' or something with that formatting. If you tell it where to look, it can look. We could probably create a trade code or system app to link all this type of stuff together for the various crafts

1

nuclearbananana t1_jdarqq9 wrote

Yes, take a look at Llama.cpp (you'll need a programming background to get it working). Even lighter models will likely come out in the future (look up stanford's alpaca, based on Llama). I'm assuming someone will also build more user friendly ways of doing this.

1

AppliedTechStuff t1_jddzgdk wrote

Simple to do.

Will take some hours training it, for certain, but natural language processing can absorb the rules and trainers can aid with interpretation and use.

1

memberjan6 t1_jdabdsl wrote

Summarize in 200 words max what you know about electrician code NEC and give 3 examples afterwards.

The National Electrical Code (NEC), also known as NFPA 70, is a comprehensive set of standards and guidelines for the safe installation and maintenance of electrical systems in the United States. Developed by the National Fire Protection Association (NFPA), the NEC covers various aspects of electrical work, including wiring, grounding, and equipment installation. The code is revised every three years to keep up with technological advancements and ensure the highest safety standards.

The NEC serves as a benchmark for electrical work and is adopted by most states and local jurisdictions. It provides essential information and best practices to ensure the safety of electricians, as well as the general public. It's important to note that the NEC is not a federal law, but rather a model code that can be voluntarily adopted and enforced by state and local governments.

Examples of NEC guidelines include:

Article 210.52: Dwelling Unit Receptacle Outlets - This article mandates that all habitable rooms in residential buildings must have an adequate number of electrical outlets, evenly spaced to minimize the use of extension cords and prevent potential hazards.

Article 240.4: Protection of Conductors - This guideline outlines the requirements for overcurrent protection of conductors, including the proper selection of circuit breakers and fuses to prevent electrical fires and equipment damage.

Article 250.66: Grounding Electrode Conductor Sizing - This section of the NEC specifies the minimum sizes for grounding electrode conductors, ensuring that the grounding system can safely carry fault current and minimize the risk of electrical shock or fire.

0

memberjan6 t1_jdaboia wrote

What is the max fill on a cable tray?

The maximum fill capacity for cable trays depends on the type of cables being used and the specific requirements outlined in the National Electrical Code (NEC) for safe cable installation. According to NEC Article 392.9(B), the fill area in cable trays must not exceed the following percentages:

For power cables or a combination of power and control cables: a. 50% fill for single conductor cables rated 600 volts or less. b. 40% fill for single conductor cables rated above 600 volts. c. 50% fill for multiconductor cables rated 600 volts or less. d. 40% fill for multiconductor cables rated above 600 volts.

For control, signal, and communication cables: a. 50% fill for all types of cables.

It's essential to adhere to these maximum fill percentages to ensure proper heat dissipation, prevent overheating, and maintain the integrity of the cable insulation. Overfilling cable trays can lead to derating of the cable ampacity, increased risk of fire, and reduced cable lifespan.

Remember to consult the latest version of the NEC and any local codes or amendments that may be in place to ensure compliance with all applicable regulations.

0