Even generative Ai has minor uses… For example I’ve used before we had a database with multiple tables but one included the the device serial number, a test type, and a test guid. Another table actually had the details of each test keyed by that test guid.
My boss would commonly send me list of serial numbers(not formatted, just space separated) and a request the test details for each device. I would setup a query for 1 serial number that returned the requested information, go onto copilot/chatgpt and say something like “write a query like this but with with all these serial numbers”. Yes I could do it my self, copy the list in, erase the space, add a comma, add a new line, repeat for each serial number(largest request o got ever was 650). The Ai got it in seconds and I verified several serial numbers picked randomly. I don’t recall it ever not working as expected. That’s a small use case and I gave lots of context.
You could also have written a script to query the entries for you. Instead of asking chatgpt to do it very resource intensively for you, spend a minor amount of time writing the script and once its done you can just run it each time your boss sends you more serial numbers.
This is the thing with LLMs. Its not really solving anything. It’s wasteful to the point of absurdity to complete tasks that are already very easily completable in far simpler and less costly ways. Its like using a block chain to perform arithmetic. You could, you could do that, but why? Why would you waste all those resources to achieve the same thing you already can?
And thats all fine and well until you get a case where it trips up and gets something really important wrong. Then its a liability. And unlike if an individual makes a mistake, an LLM cant just be told to be more careful. Its not a person, it is static at the time it is built. It cannot learn, only rebuild which is also extremely resource intensive and costly. It cannot know anything either, it just generates outputs based on the static parameters defined and tuned when it was built.
You are right but I was using it mainly for formatting, not writing the script. Granted I could of written a single purpose script to strip space, add a comma and add a new line.
I wrote the script of how to pull the data, I was mostly using it to format the lists my boss sent me.
I didn’t just take the email, copy into Ai, copy result back to email and send. I wrote the script for 1 serial, asked Ai to modify it for the following 150 serials, run it in smms, then send results.
If you were my boss, would you want me to spend my time doing… Backspace, comma, enter, down arrow…150 times? Especially if you need results fast? Yeah I could but why? Unless it was just a lazy day and felt like drooling on the keyboard.
Even generative Ai has minor uses… For example I’ve used before we had a database with multiple tables but one included the the device serial number, a test type, and a test guid. Another table actually had the details of each test keyed by that test guid.
My boss would commonly send me list of serial numbers(not formatted, just space separated) and a request the test details for each device. I would setup a query for 1 serial number that returned the requested information, go onto copilot/chatgpt and say something like “write a query like this but with with all these serial numbers”. Yes I could do it my self, copy the list in, erase the space, add a comma, add a new line, repeat for each serial number(largest request o got ever was 650). The Ai got it in seconds and I verified several serial numbers picked randomly. I don’t recall it ever not working as expected. That’s a small use case and I gave lots of context.
You could also have written a script to query the entries for you. Instead of asking chatgpt to do it very resource intensively for you, spend a minor amount of time writing the script and once its done you can just run it each time your boss sends you more serial numbers.
This is the thing with LLMs. Its not really solving anything. It’s wasteful to the point of absurdity to complete tasks that are already very easily completable in far simpler and less costly ways. Its like using a block chain to perform arithmetic. You could, you could do that, but why? Why would you waste all those resources to achieve the same thing you already can?
And thats all fine and well until you get a case where it trips up and gets something really important wrong. Then its a liability. And unlike if an individual makes a mistake, an LLM cant just be told to be more careful. Its not a person, it is static at the time it is built. It cannot learn, only rebuild which is also extremely resource intensive and costly. It cannot know anything either, it just generates outputs based on the static parameters defined and tuned when it was built.
You are right but I was using it mainly for formatting, not writing the script. Granted I could of written a single purpose script to strip space, add a comma and add a new line.
I wrote the script of how to pull the data, I was mostly using it to format the lists my boss sent me.
I didn’t just take the email, copy into Ai, copy result back to email and send. I wrote the script for 1 serial, asked Ai to modify it for the following 150 serials, run it in smms, then send results.
If you were my boss, would you want me to spend my time doing… Backspace, comma, enter, down arrow…150 times? Especially if you need results fast? Yeah I could but why? Unless it was just a lazy day and felt like drooling on the keyboard.