121–130 of 172 posts
I always find it surreal when this super science fiction tech is possible in our time but basic stuff like this can't be implemented correctly.
Turns out that cache invalidation and naming things are still the hardest problems in computer engineering.
I always find it surreal when this super science fiction tech is possible in our time but basic stuff like this can't be implemented correctly.
Turns out that cache invalidation and naming things are still the hardest problems in computer engineering.
Earlier quoted context omitted.
What are you doing with it. I am seeing this kind of comment a lot but nobody seems to actually say what it is this tool can actually do for them. My personal experience with it makes me really wonder.
I was stuck on a bug for half a day. Even ChatGPT couldn't solve it. The next day, I get access to GPT-4. I put the code in, describe the expected behavior I want, and describe the actual buggy behavior I'm getting. GPT-4 gave me the solution immediately. I'm definitely not considering this to be a low-value "toy". This is powerful stuff.
- you learn a lot of stuff otherwise you wouldn’t have learned
- you have material to talk about in your daily standups
- your colleagues recognize your effort and once you fix it you get their admiration
Using GPT to fix the bug for me would only mean: ok bug is fixed… now I’m gonna pick up the next Jira ticket. So boring.
Earlier quoted context omitted.
What are you doing with it. I am seeing this kind of comment a lot but nobody seems to actually say what it is this tool can actually do for them. My personal experience with it makes me really wonder.
By far the biggest timesaver has been having it write shell scripts for any task that's even a little tedious. I'm not good at shell scripts, so this would typically never be worthwhile for me to do for any one-off task. Generally anything where I'm working with technology I'm not an expert on, and might not find it worth it to do learn the thing just to get the task at hand done. Prompt: "write a powershell script t…
I get that this isn't a trivial service to scale and deploy, but the $20/month fee to use ChatGPT reliably is one of the most unreliable and low-value products I subscribe to. It's basically a toy, and while I'm curious to learn to use it effectively and understand how it might fit into my workflows, I can't help but feel like I'm doing them more of a favour than they are me. It's kind of like ChatGPT Plus is more on…
Earlier quoted context omitted.
There are literally thousands of chat clients based off their API. So are they also chatGPT or the model they specifically trained and aligned for chat?
From a branding perspective I'm not sure it's even clear what ChatGPT is. If GPT-3 is ChatGPT, then what is GPT-4?
Note that the API docs refer to the chat api as the "OpenAI Chat API" and not the "ChatGPT API".
Earlier quoted context omitted.
Maybe that's true (and hopefully it is) but when I was seeing this behavior earlier today it showed me a series of different conversations with related, but slightly different titles- it wasn't just random unrelated ideas, it really seemed like someone was workshopping an idea over a period of time.
If you asked the language model to generate a list of chat titles in a single query, that's what it would do. It's going to try to make them seem related. I don't know whether they do it in a single query, though.
Earlier quoted context omitted.
I used it last night to write serialization code for EnTT (an entity component system) and cereal (a serialization library). The code needed to save/load all relevant state for entities, component data, and context (sort of like a singleton entity). The amount of magic C++ required here is beyond me, but I was able to hobble through it thanks to ChatGPT. I also received help on a Discord which was essential because a…
Instead of the final code which may or may not have time saved from chatgpt, it would be more interesting to see your chat history for that specific conversation with it and see how easily or poorly it understood what was needed, and how close to the actual end result it got.
The service is down now so I can't post it, but there's definitely an art to phrasing it to do what you want. I had some cases where I told it what I wanted to do at a certain section in high level terms, but when given together it would produce garbage (using non-existent functions).
For example, if told to loop through the components from the argument list and add a component to the archive if it exists, it'd fail due to trying to accomplish the task in a more straight forward way (with functions that don't exist). However, I could break it down over the course of more prompts, so one prompt would be asking it to do a loop through components,
auto process_component_type = [&](auto component_type) {
};
(process_component_type(std::type_identity{}), ...);
using ComponentType = decltype(component_type)::type;
std::cout if (registry.ctx().contains())
{
std::cout ());
}Earlier quoted context omitted.
chatGPT definitely refers to the frontend since their API uses the actual model name (gpt-3.5-turbo) chatGPT isn’t referenced. And their API has been spotty/down all morning.
There are literally thousands of chat clients based off their API. So are they also chatGPT or the model they specifically trained and aligned for chat?
OpenAI also has an API which includes chat completion (using GPT models), image generation (DALL-E 2), and more. This has also been down this morning and is still super spotty.
Other websites using “GPT” in their name are probably using their GPT models through their API and/or are riding on chatGPT’s incredible success and public exposure.