[ home / bans / all ] [ amv / jp ] [ maho ] [ f / ec ] [ qa / b / poll ] [ tv / bann ] [ toggle-new ]

/qa/ - Questions and Answers

Questions and Answers about QA

New Reply

Options
Comment
File
Whitelist Token
Spoiler
Password (For file deletion.)
Markup tags exist for bold, itallics, header, spoiler etc. as listed in " [options] > View Formatting "


[Return] [Bottom] [Catalog]

File:82cf84b2d40c116c45febf459….jpeg (368.72 KB,1539x2048)

 No.165369

What does "competency" for kbowledge-based fields look like in a post-AI world? Before search engines, competency meant being able to recall certain information or at least be able to search through relevant reference material and so forth to find whatever you're looking for. With search engines, that changed to finding relevant information from trusted websites, but relying on them heavily (e.g. copy-pasting solutions from Stack Overflow) was discouraged. Now with modern LLMs, you can mostly trust the information they provide to be mostly correct, but they can essentially do all the work with enough proding in the right direction. So, what does competency look like when an LLM can be trusted to get things ~80% the way done?

 No.165370

File:__tanya_degurechaff_youjo_….png (295.4 KB,1600x1600)

Competency is doing the last 20% aka actually getting shit done. That's probably always been the real measure of competency it's just that before all these sophisticated knowledge aggregators it was harder or just slower to get shit done. Knowledge is useless by itself, you need to actually do something with the knowledge. Having knowledge and applying it to something useful is what competent people do.

 No.165371

File:R-1763092520307.jpg (24.81 KB,319x433)

>>165369
In my experience competence has always been a misnomer for w*rkplace compatibility.
Competency is "do thing good." Competency in the j*b market is "make it look like a risk and a loss that you're not an employee."
Even with the field leveled a bit, it's how well you use the tools and how little you mess up. However, because the field is leveled, there's also more "submit for getting thrown under the bus when the management is fucking up and does not want to take responsibility or even try to fix anything" for you to do if you want to look like a competent employee.


>with modern LLMs, you can mostly trust the information they provide to be mostly correct
Sure, if your job is to throw around lazy layman factoids and you're working under laymen that know no better.
>Now with modern LLMs, you can mostly trust the information they provide to be mostly correct, but they can essentially do all the work with enough proding in the right direction. So, what does competency look like when an LLM can be trusted to get things ~80% the way done?
Having sex with the prompting and the ways to get the really needed result sooner, without introduction of any vulnerabilities of repercussion debt. The sooner you get to the needed point compared to others, the more competent you are.
Also, being less of a layman than the LLM agent.

 No.165372

If you can't put something together at all without the LLM, you're not competent.

 No.165373

>>165372
Is copypasting from stack overflow/github/personal notes back on the table?

 No.165374

File:[SubsPlease] Mikata ga Yow….jpg (312.73 KB,1920x1080)

>>165371
AAAAH HUGE SURPRISE BOX!

 No.165375

>>165369
lots of experience?

 No.165383

Same with anything that gets mostly automated (IE CNC machines, sewing machines? printers..??), competence mostly falls down to how well you can operate the machine spirit.
Using search engines as a comparison point, you could be considered more competent than others if you can filter out answers you don't want from answers you do want. Ever had a case where you're searching for a niche problem, but had to word it in a very specific way and use word blacklists because otherwise you'd get dozens of thousands of answers on extremely basic things instead? It takes a specific type of wrangling to make the computer spirit interpret your question question the way you actually want it to.

 No.165385

>>165383
>Ever had a case where you're searching for a niche problem, but had to word it in a very specific way and use word blacklists because otherwise you'd get dozens of thousands of answers on extremely basic things instead?
wow. being a fringe fetishist helps at a j*b

 No.165387

>>165372
Yeah, this is true.

 No.165388

File:1744435632368.png (33.24 KB,554x439)

>>165369
nepotism

 No.165401

Some people even with all the information in the world wouldn't be able to apply it in a way that mattered or sort it for those who will be using it later. 100% just a knowledge check style jobs don't really exist, but some certainly come close to it. Competency was never about just regurgitating facts it was about implementing or using them effectively. The jobs I would say AI would be most able to take at the point that its in is secretarial work since it just involves handling existing data. A rando can program with AI now, but I wouldn't trust an app they make with any of my information because a rando hasn't been taught anything about cyber security and won't even realize they need to work that in. AI still needs someone who knows what to ask it, then what to do with what it spits out.




[Return] [Top] [Catalog] [Post a Reply]
Delete Post [ ]

[ home / bans / all ] [ amv / jp ] [ maho ] [ f / ec ] [ qa / b / poll ] [ tv / bann ] [ toggle-new ]