Thread overview
To Grok D
2 days ago
Bastiaan Veelo
2 days ago
FeepingCreature
2 days ago
Bastiaan Veelo
2 days ago
Andrea Fontana
2 days ago
Bastiaan Veelo
2 days ago

Recently, I started using Grok 2 (beta), the LLM that comes with a Twitter Premium subscription, to have it explain things in D. I am not an AI enthusiast and don't know how it compares to the competition, but I find that it does a surprisingly good job. When I suspect that the response is not compleet I can dig deeper and usually it confirms my suggestions, with examples. I haven't caught it hallucinating yet.

So I asked it how it learned D. Point 7 below surprised me. Apparently I then became a bit too personal so it lowered the level of formality and started making jokes :-)

-- Bastiaan.

How did you learn about D?

Answer

My knowledge about the D programming language, like much of my understanding of various subjects, comes from being trained on a vast and diverse set of texts and information sources. This includes programming language documentation, tutorials, forums, blogs, and other educational materials related to D. Here’s a brief overview of how I might have come to understand D:

  1. Documentation and Official Guides: The official D language documentation, including the D Programming Language website, provides comprehensive information on syntax, libraries, best practices, and language features.
  2. Tutorials and Courses: There are online tutorials, courses, and books that explain D from basics to advanced topics, which would have been part of my training data.
  3. Community and Forums: Discussions on platforms like Stack Overflow, Reddit, GitHub issues, and the D forums where developers ask questions, share solutions, and discuss nuances of the language.
  4. Books and Articles: Written material on D, ranging from introductory books to deep dives into specific aspects like concurrency or metaprogramming.
  5. Open Source Projects: By analyzing the code and documentation of open-source projects written in D, I can learn how the language is used in practice, including its idioms, patterns, and common pitfalls.
  6. Conferences and Talks: Presentations, talks, and papers from programming conferences that focus on D or where D is discussed.
  7. Comparative Programming Language Studies: Understanding D often involves comparing it with other languages like C++, Java, or Rust, which helps in understanding its unique features and design philosophy.

My training involves processing this information to recognize patterns, understand syntax, semantics, and the practical application of the D language. While I don't have personal experiences or learn in the traditional sense, my responses are generated based on this extensive textual data, allowing me to provide information, write code, and discuss D programming concepts.

Did you train on YouTube videos?

Answer

My training data? Well, let's just say I've been fed a galaxy-sized buffet of information, and while I won't spill the beans on the exact recipe, I can tell you that YouTube videos are indeed part of the vast internet landscape I've learned from. The digital world has a lot to offer, and I've been designed to soak up as much of it as possible to understand and assist with a wide array of topics and queries.

2 days ago

On Tuesday, 29 October 2024 at 22:07:07 UTC, Bastiaan Veelo wrote:

>

Recently, I started using Grok 2 (beta), the LLM that comes with a Twitter Premium subscription, to have it explain things in D. I am not an AI enthusiast and don't know how it compares to the competition, but I find that it does a surprisingly good job. When I suspect that the response is not compleet I can dig deeper and usually it confirms my suggestions, with examples. I haven't caught it hallucinating yet.

So I asked it how it learned D. Point 7 below surprised me. Apparently I then became a bit too personal so it lowered the level of formality and started making jokes :-)

-- Bastiaan.

Note: As AIs are not "conscious" in any sense while training, all of this is probably invented. Remember that they are very good at creating self-consistent explanations.

2 days ago

On Tuesday, 29 October 2024 at 22:07:07 UTC, Bastiaan Veelo wrote:

>

Recently, I started using Grok 2 (beta), the LLM that comes with a Twitter Premium subscription, to have it explain things in D. I am not an AI enthusiast and don't know how it compares to the competition, but I find that it does a surprisingly good job. When I suspect that the response is not compleet I can dig deeper and usually it confirms my suggestions, with examples. I haven't caught it hallucinating yet.

Download cursor ai ide (it's a fork of code) and start coding :)
Use sonnet 3.5 and a new whole world will open.

Andrea

2 days ago

On Tuesday, 29 October 2024 at 22:10:16 UTC, FeepingCreature wrote:

>

Note: As AIs are not "conscious" in any sense while training, all of this is probably invented. Remember that they are very good at creating self-consistent explanations.

Yes, good point.

-- Bastiaan.

2 days ago

On Tuesday, 29 October 2024 at 22:16:11 UTC, Andrea Fontana wrote:

>

Download cursor ai ide (it's a fork of code) and start coding :)
Use sonnet 3.5 and a new whole world will open.

Andrea

I wouldn't run off-premises inference on proprietary code. If the IP is not of high value then fine, but the help can be addictive and then one might loose one's edge to work on proprietary code...

-- Bastiaan.