I tried chat gpt for the first time today. I asked it to write me a script to fetch some data from an API. It didn't succeed, and this kind of common task from a well-documented API should (I thought) be a slam dunk.
I tried driving a car for the first time today. I just wanted to drive from my house to the end of the block. I crashed into a fire hydrant and almost killed three people. This should be a very simple and common task, going from your house to the end of the block, and yet the car failed miserably.
Sarcasm aside, at my company I've noticed a very disturbing split among my co-workers. One group of co-workers hold the above attitude, that they should be able to just use a tool on day one and it should be able to solve problems that even just a couple of months ago would have been unimaginable. That group of co-workers looks at ChatGPT and thinks it's basically underwhelming and a big nothingburger.
The other group of co-workers are learning how to use ChatGPT to solve their problems. They're learning clever ways of prompting ChatGPT, how to give it a proper system message, how to take large tasks and break it down into smaller tasks to offload a lot of the grunt work onto ChatGPT while they focus more on the overall architecture. That group of co-workers is seeing excellent gains in productivity.
The idea that you would just use a tool on day one, probably don't really know much about it or understand it that well, and expect it to magically solve your problems is about as foolish as driving a car for the first time, not really knowing anything about cars, and expecting nothing bad to happen.
ChatGPT is a tool, the people who take the time to learn that tool and understand how it works will become much more productive software developers, and those who don't will get left behind.
I am reminded of this intro by Marc Andreessen for Breaking Smart[1]:
> A great deal of product development is based on the assumption that products must adapt to unchanging human needs or risk being rejected. Yet, time and again, people adapt in unpredictable ways to get the most out of new tech. Creative people tinker to figure out the most interesting applications, others build on those, and entire industries are reshaped.
There will high group of people who will tinker and adapt to the way the gpts work and make the best of it - creating new apps, industries, and so on. This has been the case all through the tech cycles.
We will change ourselves to the new tech and we won't even realize we changed.
I am re-reading Breaking Smart book in the context of chatgpt and I'm getting way more insights than I read the first time.
You have no information on how I approached using chat gpt, what prompts I used, how close its replies came to being correct/useful or not, but you decided to regale us with an irrelevant car metaphor anyway.
A big one for me is being able to decide when to use GPT3.5 and when to use GPT4 for a given prompt. Perhaps it will be short term that this is a "skill", sooner or later they will lift the GPT4 cap and speed it up, but for right now I can stretch out a GPT4 session indefinitely if I know what gpt3.5 can and can't handle.
Besides that though, chatGPT has been nothing short of a huge product....I don't even want to say a "huge productivity boost", that implies like good sleep or drinking coffee, its been like an augmentation that allows be to bat above my league. Like a free (well $20/mo) co-worker who knows a lot about programming.
GPT-4 is a lot better at things like that. That said, you'll either have to pay for it, or instead use some severely rate-limited but free interface you might find by asking around.
Prompt engineering is a thing for sure. Might want to search the news here for some clues on ycombinator to get better at it. And some tasks are beyond it.
Kranar|2 years ago
Sarcasm aside, at my company I've noticed a very disturbing split among my co-workers. One group of co-workers hold the above attitude, that they should be able to just use a tool on day one and it should be able to solve problems that even just a couple of months ago would have been unimaginable. That group of co-workers looks at ChatGPT and thinks it's basically underwhelming and a big nothingburger.
The other group of co-workers are learning how to use ChatGPT to solve their problems. They're learning clever ways of prompting ChatGPT, how to give it a proper system message, how to take large tasks and break it down into smaller tasks to offload a lot of the grunt work onto ChatGPT while they focus more on the overall architecture. That group of co-workers is seeing excellent gains in productivity.
The idea that you would just use a tool on day one, probably don't really know much about it or understand it that well, and expect it to magically solve your problems is about as foolish as driving a car for the first time, not really knowing anything about cars, and expecting nothing bad to happen.
ChatGPT is a tool, the people who take the time to learn that tool and understand how it works will become much more productive software developers, and those who don't will get left behind.
jjude|2 years ago
There will high group of people who will tinker and adapt to the way the gpts work and make the best of it - creating new apps, industries, and so on. This has been the case all through the tech cycles.
We will change ourselves to the new tech and we won't even realize we changed.
I am re-reading Breaking Smart book in the context of chatgpt and I'm getting way more insights than I read the first time.
[1]: https://breakingsmart.com/en/season-1/introduction-by-marc-a...
throwaway8689|2 years ago
Workaccount2|2 years ago
Besides that though, chatGPT has been nothing short of a huge product....I don't even want to say a "huge productivity boost", that implies like good sleep or drinking coffee, its been like an augmentation that allows be to bat above my league. Like a free (well $20/mo) co-worker who knows a lot about programming.
ogogmad|2 years ago
vajrabum|2 years ago