4 ms·

I am a bit of a greybeard around new tools, and even today my usage of ChatGPT is limited, so I will give my 2c. You're right, in most cases, Google + Github se

I am a bit of a greybeard around new tools, and even today my usage of ChatGPT is limited, so I will give my 2c. You're right, in most cases, Google + Github search is faster to achieve a certain task vs typing out your problem and waiting for ChatGPT to generate results (personal opinion). Nonetheless, here are a few things where ChatGPT saves a lot of time: 1. Generating config files: Eg, Webpack, Docker-compose. Describe your use-case and it will generate the config file you need. This is usually faster than perusing docs and looking for specific flags/options. 2. Generating CLI syntax: jq, cat, tar, find. The flags/syntax are so hard to remember when you want to do something specific. ChatGPT does a great job here in quickly generating the command you want. 3. Solving problems when you're stuck: When I feel stuck with a problem or an error message, ChatGPT has proven to be an invaluable resource. Even when the answer it gave was wrong, a little hint in there was enough for me to find the solution I needed. 4. Generating functions you're too lazy to write: Often, you know how to write a function but you're too damn lazy to write it. Eg, use axios, go to this URL, get the contents and write it to CSV. Writing a function to do this feels like a chore. ChatGPT is a helpful assistant in this case. Overall, I think there are plenty of ways the tool can save you time, even if you're skeptical about changing your whole workflow.

freitasm 3y ago
> 3. Solving problems when you're stuck: When I feel stuck with a problem or an error message, ChatGPT has proven to be an invaluable resource. Even when the answer it gave was wrong, a little hint in there was enough for me to find the solution I needed. > 4. Generating functions you're too lazy to write: Often, you know how to write a function but you're too damn lazy to write it. Eg, use axios, go to this URL, get the contents and write it to CSV. Writing a function to do this feels like a chore. ChatGPT is a helpful assistant in this case. A friend swears by it and has whole single-page sites written by chatGPT in his internal lab. Just to test your point, I've asked ChatGPT to write a VBScript function that returns false if the input is true and false if the input is false. It wrote a seven-line function testing true, false and otherwise (clever, testing for null). I would have just written a function to return false - one line that covers all conditions, including null. There is no testing, so it uses fewer CPU cycles. So that's my rational: yes it can be used but don't just take the results at face value.
- mmh0000 3y ago
  >> I've asked ChatGPT to write a VBScript function that returns false if the input is true and false if the input is false. <insert to-be-fair meme here> You purposefully asked GPT a trick question for which it likely has very little training data. No one would ever go on StackExchange et al. asking how to write a function that always returns false. So GPT had to take its training data on other, realistic questions and fumble something together. A GPT can only ever be as good as its initial training material. It can not have new ideas or innovations. There was an excellent post here a couple of days ago to an article, which explains quite well what a GPT can and can not do and why: https://explainextended.com/2023/12/31/happy-new-year-15/ https://explainextended.com/2023/12/31/happy-new-year-15/ EDIT. GPT4 was happy to give me a one-liner that meets the conditions and even explains why my request was dumb: https://chat.openai.com/share/1d949b5b-3c2b-4634-802e-3ea4c2a7a786 https://chat.openai.com/share/1d949b5b-3c2b-4634-802e-3ea4c2...
- fragmede 3y ago
  That's GPT-3.5. GPT-4 manages to identify that the function always returns false and even names it AlwaysFalse. It doesn't do well on trick questions, but that's a bit like calling someone stupid because they can't tell you the number you're thinking of. Use the right tool for the job. Instead of trying to prove it's dumb and therefore useless, ask it questions it can do well at. It really doesn't do well with trick questions. If you ask it the river crossing problem but the boat is big enough to bring all the items, it insists on taking multiple trips. But that's okay. Because of AI we have to develop a more nuanced view of intelligence. previously, you could just place people on a 2d spectrum from dumb as a door nail to frighteningly smart. Now, because we have these things that are impossibly smart compared to a human in some dimensions but not in others, we have to give that 2d scale a lot more dimensions. Which always existed before, but most of us aren't cognitive psychologists.
chatmasta 3y ago
> Google + Github search GitHub Search is my favorite tool and I don't think most people realize how powerful it is or how much they can rely on it. You can get so much intel from there, not just about how to use a library, but even esoteric URLs or some fingerprints that are only in a site you're trying to scrape or an app you're trying to reverse engineer... it's saved me a ton of time and is a great way to find specific resources that I'd never know existed otherwise. Co-Pilot is also great but it's hard to explain why without just telling you to try it yourself. It's basically a super advanced auto-complete, but it's also more than that... it feels like a second brain that is running one thought ahead of you at all times. Sometimes I'm thinking through a problem and I just press tab and it's already thought of the solution.
lk23nlk23 3y ago
This is why powershell is so much better than bash or alternatives. Just convert any data type to object and work with it using common syntax. No need to learn 300 different syntaxes/parameters for different parsers like jq/xmlstarlets/etc.
nunez 3y ago
idk I feel like [3] aside, doing all of those things doesn't take me very much time? Also, I think that turning to ChatGPT to compose a shell pipeline is a smell
- hnfong 3y ago
  Doing [4] yourself doesn't take a lot of time, but the time savings adds up. That said, I wouldn't recommend using the standard ChatGPT interface for generating code -- just use a solution that integrates with your IDE/editor, and use it as some kind of coding autocomplete. (You can sometimes even trick the AI to converse with you in the comments, eg. /* Question: what is the gcc option to enable all warnings? Answer: <wait for completion>... ), so it can be super convenient when you just have a random question while typing code in your IDE or editor. 100% would recommend everyone to try it out.

freitasm 3y ago

> 3. Solving problems when you're stuck: When I feel stuck with a problem or an error message, ChatGPT has proven to be an invaluable resource. Even when the answer it gave was wrong, a little hint in there was enough for me to find the solution I needed. > 4. Generating functions you're too lazy to write: Often, you know how to write a function but you're too damn lazy to write it. Eg, use axios, go to this URL, get the contents and write it to CSV. Writing a function to do this feels like a chore. ChatGPT is a helpful assistant in this case. A friend swears by it and has whole single-page sites written by chatGPT in his internal lab. Just to test your point, I've asked ChatGPT to write a VBScript function that returns false if the input is true and false if the input is false. It wrote a seven-line function testing true, false and otherwise (clever, testing for null). I would have just written a function to return false - one line that covers all conditions, including null. There is no testing, so it uses fewer CPU cycles. So that's my rational: yes it can be used but don't just take the results at face value.

mmh0000 3y ago
>> I've asked ChatGPT to write a VBScript function that returns false if the input is true and false if the input is false. <insert to-be-fair meme here> You purposefully asked GPT a trick question for which it likely has very little training data. No one would ever go on StackExchange et al. asking how to write a function that always returns false. So GPT had to take its training data on other, realistic questions and fumble something together. A GPT can only ever be as good as its initial training material. It can not have new ideas or innovations. There was an excellent post here a couple of days ago to an article, which explains quite well what a GPT can and can not do and why: https://explainextended.com/2023/12/31/happy-new-year-15/ https://explainextended.com/2023/12/31/happy-new-year-15/ EDIT. GPT4 was happy to give me a one-liner that meets the conditions and even explains why my request was dumb: https://chat.openai.com/share/1d949b5b-3c2b-4634-802e-3ea4c2a7a786 https://chat.openai.com/share/1d949b5b-3c2b-4634-802e-3ea4c2...
fragmede 3y ago
That's GPT-3.5. GPT-4 manages to identify that the function always returns false and even names it AlwaysFalse. It doesn't do well on trick questions, but that's a bit like calling someone stupid because they can't tell you the number you're thinking of. Use the right tool for the job. Instead of trying to prove it's dumb and therefore useless, ask it questions it can do well at. It really doesn't do well with trick questions. If you ask it the river crossing problem but the boat is big enough to bring all the items, it insists on taking multiple trips. But that's okay. Because of AI we have to develop a more nuanced view of intelligence. previously, you could just place people on a 2d spectrum from dumb as a door nail to frighteningly smart. Now, because we have these things that are impossibly smart compared to a human in some dimensions but not in others, we have to give that 2d scale a lot more dimensions. Which always existed before, but most of us aren't cognitive psychologists.

chatmasta 3y ago

> Google + Github search GitHub Search is my favorite tool and I don't think most people realize how powerful it is or how much they can rely on it. You can get so much intel from there, not just about how to use a library, but even esoteric URLs or some fingerprints that are only in a site you're trying to scrape or an app you're trying to reverse engineer... it's saved me a ton of time and is a great way to find specific resources that I'd never know existed otherwise. Co-Pilot is also great but it's hard to explain why without just telling you to try it yourself. It's basically a super advanced auto-complete, but it's also more than that... it feels like a second brain that is running one thought ahead of you at all times. Sometimes I'm thinking through a problem and I just press tab and it's already thought of the solution.

lk23nlk23 3y ago

This is why powershell is so much better than bash or alternatives. Just convert any data type to object and work with it using common syntax. No need to learn 300 different syntaxes/parameters for different parsers like jq/xmlstarlets/etc.

nunez 3y ago

idk I feel like [3] aside, doing all of those things doesn't take me very much time? Also, I think that turning to ChatGPT to compose a shell pipeline is a smell

hnfong 3y ago
Doing [4] yourself doesn't take a lot of time, but the time savings adds up. That said, I wouldn't recommend using the standard ChatGPT interface for generating code -- just use a solution that integrates with your IDE/editor, and use it as some kind of coding autocomplete. (You can sometimes even trick the AI to converse with you in the comments, eg. /* Question: what is the gcc option to enable all warnings? Answer: <wait for completion>... ), so it can be super convenient when you just have a random question while typing code in your IDE or editor. 100% would recommend everyone to try it out.