Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Ability to see expertise is a milestone worth aiming for (2022)

commoncog.com

Post reply on HN
Print view View original

41–50 of 177 posts

Re: Ability to see expertise is a milestone worth aiming for (2022)

#41

Post by ryanklee »

jiggawatts wrote: post #37
thesumofall wrote: post #33

Earlier quoted context omitted.

But is this really the point of the article? The article argues that it is difficult to identify expertise. However, the situation you describe is one where (if I read you correctly) not your expertise was questioned but your recommendation. If there wasn’t any tangible difference between the CTO’s and your suggested course of action on a metric that he as a customer cared about, then maybe his proposal was indeed th…

I'm finding this entire line of discussion incredibly curious because not only does it confirm my original argument and the that of the article, but it hammers the point home. Everything you've just said is gibberish to a master. It's like making a straight-faced argument that setting the speed limit to 1 mph on a motorway might make perfect sense, and the traffic engineer is actually at fault for not making it clear…

Really does sound like you have difficulties with communicating your points that do not have to do with external factors, but with your actual skill at communicating. Experts manage to successfully argue fine points to laypersons all the time to non experts in every field. You are coming across as deflective.

Re: Ability to see expertise is a milestone worth aiming for (2022)

#42

Post by giantg2 »

jiggawatts wrote: post #2

This is an extremely important article to read for professionals of all sorts, not just freelancers. It's an issue that I struggle with nearly daily. I can't convince customers to pay me what I'm worth, because they can't judge my true value. Meanwhile, they'll happily pay Deloitte or Accenture 3x as much as what I charge, and then take the misguided ramblings of a junior-sold-as-senior at face value, not knowing the…

"take the misguided ramblings of a junior-sold-as-senior at face value, not knowing the difference"

Based on my experiences, I'm throughly convinced that most of the time the consultants will tell management what they want to hear. If management doesn't like the answer then they go to someone else. How likely it is that they will accept an unwanted opinion from a consultant is solely based on prestige.

Re: Ability to see expertise is a milestone worth aiming for (2022)

#43

Post by whateveracct »

jiggawatts wrote: post #39
ZephyrBlu wrote: post #21

Earlier quoted context omitted.

> When I watched AlphaZero play chess, I was just a tiny bit impressed. I watched it beat the Go grandmaster and I was not very impressed. Then I watched it play Dota 2 -- in which I've sunk several thousand hours -- and holy crap. Wow! It was suddenly terrifying how good it was, because I could judge the nuance of what it was doing Funny, my game of choice is StarCraft 2 and when I watched AlphaStar beat StarCraft 2…

You should be more impressed knowing that that AlphaStar was handicapped . Its clicking was rate-limited and its inputs were delayed by something like 150 milliseconds to make it "fair" for the humans. If you remove those limits, it would absolutely wipe the floor with the twitching corpse of any human player. I saw it play Warcraft 2 without the limiters, and it was just ridiculous. It would dance the units around a…

But not having human reaction time in a real-time game isn't interesting. And the strategies that arise from that aren't either. If anything, human-level I/O constraints should make the AI more impressive because it's actually playing the same game.

Re: Ability to see expertise is a milestone worth aiming for (2022)

#44

Post by ZephyrBlu »

jiggawatts wrote: post #39
ZephyrBlu wrote: post #21

Earlier quoted context omitted.

> When I watched AlphaZero play chess, I was just a tiny bit impressed. I watched it beat the Go grandmaster and I was not very impressed. Then I watched it play Dota 2 -- in which I've sunk several thousand hours -- and holy crap. Wow! It was suddenly terrifying how good it was, because I could judge the nuance of what it was doing Funny, my game of choice is StarCraft 2 and when I watched AlphaStar beat StarCraft 2…

You should be more impressed knowing that that AlphaStar was handicapped . Its clicking was rate-limited and its inputs were delayed by something like 150 milliseconds to make it "fair" for the humans. If you remove those limits, it would absolutely wipe the floor with the twitching corpse of any human player. I saw it play Warcraft 2 without the limiters, and it was just ridiculous. It would dance the units around a…

The handicapping was very ineffective. AlphaStar had 250-300+ APM and it was basically 100% effective APM. Pros have 300-400+ APM, but their effective APM is much lower due to spam patterns and checking production among other things.

Almost everyone in the community seemed to agree AlphaStar was very unfair. Things like a 3 way surround and performing blink micro on individual Stalkers on opposite sides of the screen are completely inhuman.

This is also what made the Dota 2 bots feel somewhat disappointing. Of course perfect mechanical skill beats humans. Watching the SF mid 1v1 matches illustrated this very clearly to me. The AI chipped the player down with perfect AAs, then finished them off with perfect distance razes.

The bots weren't 100% mechanical skill, but it was a large enough proportion that it was uninteresting.

Re: Ability to see expertise is a milestone worth aiming for (2022)

#45

Post by SanderNL »

candiddevmike wrote: post #32
SanderNL wrote: post #29

Earlier quoted context omitted.

If anything it’s more important. Without it you cannot separate its confabulation from its useful output. Knowing when to use what tool, how to do it properly and to what extent takes expertise as well. Carpenters and craftsmen in general are infinitely more productive with woodworking and related tools than I will ever be. My destiny is assembling IKEA cupboards. The same principle will apply with AI tools. Sure you…

Probably not the best comparison, as a carpenter or craftsmen can't tell a LLM something to build X out of wood and have it come out mostly OK (yet?). Anyone can tell ChatGPT to write some code to do X, not just expert programmers. It may not be perfect, but it's only going to get better, no?

This metaphor is not about the tools so much as it is about the nature of expertise and its effect on the results of said tools.

Someone with expertise will use any tools, including GPT, in ways a novice can’t even understand. Same tools, different outputs. Not the tools, because basically anyone has access, but the one using it will determine the outcome. In a capitalist society you will have to compete otherwise known as staying employed.

Someone with expertise can hook up AIs in ways and with a velocity that will leave amateurs in the dust. The amateurs will still be fiddling around with “prompts” and Googling or ChatGPT’ing what “latent” means while the expert has working systems ready to go.

I don’t see this chasm getting smaller. These tools magnify your inherent (in)competency. Try to compete with a professional artist competent with Stable Diffusion if you want a taste of that.

Re: Ability to see expertise is a milestone worth aiming for (2022)

#46

Post by yard2010 »

ZephyrBlu wrote: post #15
jiggawatts wrote: post #2

This is an extremely important article to read for professionals of all sorts, not just freelancers. It's an issue that I struggle with nearly daily. I can't convince customers to pay me what I'm worth, because they can't judge my true value. Meanwhile, they'll happily pay Deloitte or Accenture 3x as much as what I charge, and then take the misguided ramblings of a junior-sold-as-senior at face value, not knowing the…

It sounds like you were just telling the CTO "this is wrong" instead of offering an alternative solution? If that's the case it's not surprising it didn't sink in. You're just giving him another problem. > I've spent my career performance tuning, but then how do I explain this to people who think a 60-second timeout is a performance problem best solved by buying another million dollars worth of kit? I'm surprised you…

Did you read the article? This is the point, selling a good service to people that has no idea is inherently hard

Re: Ability to see expertise is a milestone worth aiming for (2022)

#48

Post by ZephyrBlu »

yard2010 wrote: post #46
ZephyrBlu wrote: post #15

Earlier quoted context omitted.

It sounds like you were just telling the CTO "this is wrong" instead of offering an alternative solution? If that's the case it's not surprising it didn't sink in. You're just giving him another problem. > I've spent my career performance tuning, but then how do I explain this to people who think a 60-second timeout is a performance problem best solved by buying another million dollars worth of kit? I'm surprised you…

Did you read the article? This is the point, selling a good service to people that has no idea is inherently hard

That is very much not the point of the article. The article is very focused on exploring the idea of identifying expertise and only briefly mentions money in the beginning of the piece.

Cedric explicitly directs people to another article which discusses how this idea relates to money: "Bellotti’s observations are worth reading in their entirety, especially if you’re interested in the actionable implications of the idea. But I’d like to take this property of expertise in another direction. Now that we know that expertise is more legible to experts, could we perhaps invert this observation and turn it into a goal by itself? Could you turn this into something that helps you acquire expertise?".

Re: Ability to see expertise is a milestone worth aiming for (2022)

#50

Post by darkerside »

jiggawatts wrote: post #2

This is an extremely important article to read for professionals of all sorts, not just freelancers. It's an issue that I struggle with nearly daily. I can't convince customers to pay me what I'm worth, because they can't judge my true value. Meanwhile, they'll happily pay Deloitte or Accenture 3x as much as what I charge, and then take the misguided ramblings of a junior-sold-as-senior at face value, not knowing the…

Isn't it possible that you may not understand the CTO's priorities? Security is clearly essential to him, and performance is essential to you. I don't doubt there are other ways to satisfy both criteria, but frankly, it just doesn't sound like the CTO cared about those milliseconds.
Post reply on HN