# Be aware, this will kill us! 😂 **URL:** \ **Category:** General Dev Chat\ **Tags:** ai\ **Created:** [6 August 2024 01:29 UTC](https://forum.devtalk.com/t/be-aware-this-will-kill-us/165041 "2024-08-06T01:29:03Z")\ **Posts on this page:** 1\ **Showing post:** 3
**Author:** ![Eiji](https://forum.devtalk.com/user_avatar/forum.devtalk.com/eiji/32/6784_2.png) [@Eiji](https://forum.devtalk.com/u/Eiji)\ **Post date:** [19 December 2024 05:44 UTC](https://forum.devtalk.com/t/be-aware-this-will-kill-us/165041/3 "2024-12-19T05:44:33Z")
Here we go again .. `GitHub` made `Copilot` free (with limitations). I have tried it and I was a bit surprised - only a bit … Please keep in mind that `Copilot` does not officially support `Elixir`. ❗ ### GPT 4o ![obraz](https://forum.devtalk.com/uploads/default/original/2X/3/3d0f60399e8c55394df89f488522e6538969b4b1.png) Let’s make it quick to not repeat myself too much once again … 1. Why do `chatbots` always try to convince me that I want something else that I want? I don’t want a `String`! 🤦‍♂️ 2. The `chatbot` have changed an option key (correctly) and the value (valid, but wrong for this case) and didn’t tell about that. :icon_confused: 3. Once again `ChatGPT` tries to convince me that I want to print data as list … especially if it changed a `chalist` to a `String`! 😂 4. There is no false in last paragraph, but it’s an answer for a different question. I didn’t asked to change type, but to `fix` code. Is `GPT` based on commits with only `fix` as a message? 💡 ### Claude 3.5 Sonnet ![obraz](https://forum.devtalk.com/uploads/default/original/2X/9/9594e2a412bdb59016b1fb2c4c25d3e84319b278.jpeg) First of all I have respect for `chatbot` have written in first paragraph. To perfection there should be also note like: > Please pay attention that `Elixir` is not officially supported language by `Copilot`, see below link for more information: > [GitHub language support - GitHub Docs](https://docs.github.com/en/get-started/learning-about-github/github-language-support) Second paragraph is amazing! That’s a first `chatbot` mentioning `deprecation` and it even told me which `Elixir` version deprecated old naming. 👍 It’s still far away from perfection, but it’s already on good road. Better answer should include: 1. Mention that a change in naming is recommended since `Elixir` version `1.3.0` as since this version the documentation at `hexdocs` was has been changed with a new naming. 🗒 2. There should be a links to: changelog, discussion, issue, PR and maybe even a direct link to commit (page with commit also includes some helpful information, so for some cases it’s also worth to mention it). 🔗 The changed code however decreases the quality. 1. It does not suggest a naming change `'char_lists'` → `'charlists'` 🙈 2. It does not suggest a `c` sigil `'char_lists'` → `~c"char_lists"` 🙁 3. The option key is changed to proper one as described. However once again the value is valid, but not in context of the value in question. It already found a deprecation, so it could just follow that lead! 👎 Similarly to `GPT 4o` the next paragraph is almost correct, but answers wrong question. We still can notice that `chatbot` have a problem with figuring out a difference between a `charlist` and `string`. 🤯 I would not said any word about last paragraph if the `chatbot` would not stop half a way. Many additions mentioned before are “nice to have” and they are not as much required as the correct answer. Since it already realised the deprecation it should give a proper answer without any more context. 🙊 ## Summary `Claude 3.5 Sonnet` in `GitHub Copilot` is first `chatbot` that I can recommend to `Elixir` developers, but only for **seniors** who understand the problems with so-called "AI"s. I hope that in next version developers would be able to make `chatbot` keep one idea without telling developers that they want something else. Since it even found a version I also hope it would be much more descriptive and provide a links for learning purposes. 😍 I guess that until `Elixir` would not be officially supported we would still need to be extremely careful about `chatbot` responses. However it’s not really as bad a it looks. It shows that’s not an `AI` which suddenly “wake up”, but just another algorithm. I can’t blame any developer working on any `chatbot` for such mistakes as simply `LLM`s alone were never intended for that purposes and if they would not change narration then I would not see any problem with `LLM` part. 🙈 They already are doing well. We should also keep in mind that the more lines of code the more possible bugs would be introduced. I would say that changing a direction of response (from deprecation to changing code logic) may be considered as a bug, but I also understand that `LLM`s have problems with figuring out the context. 🤔 --- _[View the full topic](https://forum.devtalk.com/t/be-aware-this-will-kill-us/165041)._