3 ms·

Exactly what you said Andrew. The comparisons are lacking and are almost at whataboutism level. The amount of actual 'work' that AI does versus the tools of y

by ulfw 2y ago

Exactly what you said Andrew. The comparisons are lacking and are almost at whataboutism level. The amount of actual 'work' that AI does versus the tools of yesterday are an order of magnitude away

kruxigt 2y ago
[dead]
smokel 2y ago
Going from punched cards to interactive terminals surely must have been a big productivity boost. And going from text based CAD to what is possible on modern workstations has probably also helped a bit in that field. In that view I'd say the productivity boost by LLMs is somewhat disappointing, especially with respect to how amazing they are.
player1234 2y ago
Quantify it, show us the numbers.
- NitpickLawyer 2y ago
  I think the field is too new and the successful stories too private atm. However I think the best apples to apples example in this context is Amz's codebase update project that they've blogged about. From memory, they took some old java projects, and had some LLM driven "agents" update the codebase to recent java. I don't know java enough to know how "hard" this task is, but asking around I've heard that "analog" tools for this exist, but aren't that good, bork often, are hardcoded and so on. Amz reported ~70% of code that came out passed code review, presumably the rest had to be tweaked by humans. I don't know if there are any "classical" tools that can do that ootb. So yeah, that's already imrpessive and "available today" so to speak.
  - skydhash 2y ago
    Java is intent as code. It’s so verbose that you have to use an IDE to not go crazy with all the typings. And when using an IDE, you autocomplete more than you type because of all the information that exists in the code
- fragmede 2y ago
  quantifying programmer productivity has been a problem since its inception. lines of code is a terrible metric. so is Jira ticket points. I can tell you that using an LLM, I can make a chrome extension to put a div that says "hello world" at the top of every webpage far quicker than if I had to read the specifications of extension manifests and how to do it manually but how do you quantify that generically? how do you quantify that vs the wasted time because it doesn't understand some nuance of what I'm asking it to do, or when it gets confused about something and goes in circles?
  - skydhash 2y ago
    You could also download a sample extension code, strip out the text and put “Hello World” instead. As fast and no need to train a model to do that.
xbmcuser 2y ago
The problem is not what ai can do rather most people in the workforce don't how to use the current generation of Ai. As the children that grew up with using chat gpt etc get into the workforce then only will we see the real benefits of AI.
- rwyinuse 2y ago
  Oh yeah, the "digital native" myth. I'm not convinced children using ChatGPT to do their homework will actually make them more productive workers. More likely it's going to have the opposite effect, as they're going to lack deeper understanding that you can build only through doing the homework yourself. Really it's not about just using technology, but how you use it. Lots of adults expected kids with smartphones to be generally good with technology, but that's not what we're witnessing now. It turns out browsing TikTok and Snapchat doesn't teach you much about things like file system, text editing, spreadsheets, skills that you actually need as a typical office worker.
  - xbmcuser 2y ago
    That's different from what I talking about it's the problem of inertia people already in jobs are used to doing them in a particular way. New curious driven people that get into the work force would optimize a lot of office work. A 10-12 year old that has learned how to use Ai from the very start will be using an AI that has 12-15 years of incremental improvements when he or she gets into the work force. A lot of people here on hacker news disparage newer generations. But how many of you can run a tube based or punched based computer. So if you don't know are you an idiot?
- xxs 2y ago
  ...indeed just wait few more years.
blharr 2y ago
Moore's Law? The computer does about 30,000 times more than it did 30 years ago. Order of magnitude shifts are just common

by ulfw 2y ago

Exactly what you said Andrew. The comparisons are lacking and are almost at whataboutism level. The amount of actual 'work' that AI does versus the tools of yesterday are an order of magnitude away

kruxigt 2y ago

[dead]

smokel 2y ago

Going from punched cards to interactive terminals surely must have been a big productivity boost. And going from text based CAD to what is possible on modern workstations has probably also helped a bit in that field. In that view I'd say the productivity boost by LLMs is somewhat disappointing, especially with respect to how amazing they are.

player1234 2y ago

Quantify it, show us the numbers.

NitpickLawyer 2y ago
I think the field is too new and the successful stories too private atm. However I think the best apples to apples example in this context is Amz's codebase update project that they've blogged about. From memory, they took some old java projects, and had some LLM driven "agents" update the codebase to recent java. I don't know java enough to know how "hard" this task is, but asking around I've heard that "analog" tools for this exist, but aren't that good, bork often, are hardcoded and so on. Amz reported ~70% of code that came out passed code review, presumably the rest had to be tweaked by humans. I don't know if there are any "classical" tools that can do that ootb. So yeah, that's already imrpessive and "available today" so to speak.
- skydhash 2y ago
  Java is intent as code. It’s so verbose that you have to use an IDE to not go crazy with all the typings. And when using an IDE, you autocomplete more than you type because of all the information that exists in the code
fragmede 2y ago
quantifying programmer productivity has been a problem since its inception. lines of code is a terrible metric. so is Jira ticket points. I can tell you that using an LLM, I can make a chrome extension to put a div that says "hello world" at the top of every webpage far quicker than if I had to read the specifications of extension manifests and how to do it manually but how do you quantify that generically? how do you quantify that vs the wasted time because it doesn't understand some nuance of what I'm asking it to do, or when it gets confused about something and goes in circles?
- skydhash 2y ago
  You could also download a sample extension code, strip out the text and put “Hello World” instead. As fast and no need to train a model to do that.

xbmcuser 2y ago

The problem is not what ai can do rather most people in the workforce don't how to use the current generation of Ai. As the children that grew up with using chat gpt etc get into the workforce then only will we see the real benefits of AI.

rwyinuse 2y ago
Oh yeah, the "digital native" myth. I'm not convinced children using ChatGPT to do their homework will actually make them more productive workers. More likely it's going to have the opposite effect, as they're going to lack deeper understanding that you can build only through doing the homework yourself. Really it's not about just using technology, but how you use it. Lots of adults expected kids with smartphones to be generally good with technology, but that's not what we're witnessing now. It turns out browsing TikTok and Snapchat doesn't teach you much about things like file system, text editing, spreadsheets, skills that you actually need as a typical office worker.
- xbmcuser 2y ago
  That's different from what I talking about it's the problem of inertia people already in jobs are used to doing them in a particular way. New curious driven people that get into the work force would optimize a lot of office work. A 10-12 year old that has learned how to use Ai from the very start will be using an AI that has 12-15 years of incremental improvements when he or she gets into the work force. A lot of people here on hacker news disparage newer generations. But how many of you can run a tube based or punched based computer. So if you don't know are you an idiot?
xxs 2y ago
...indeed just wait few more years.

blharr 2y ago

Moore's Law? The computer does about 30,000 times more than it did 30 years ago. Order of magnitude shifts are just common