4 ms·

Why do you think this? I'm totally blind and find Chat GPT to be useful for quick answers to questions instead of searching through SEO spam and low quality tut

by jareds 3y ago

Why do you think this? I'm totally blind and find Chat GPT to be useful for quick answers to questions instead of searching through SEO spam and low quality tutorials. I find Copilot to be sueful for better code completion and helping to write basic unit tests. I don't find that it can actually write complex code for me. I very well may be using the tools wrong though.

peddling-brink 3y ago
Multimodal LLMs can take images and describe them as text, which I imagine could be useful in some cases, for the vision impaired.
- RobMurray 3y ago
  It can be useful, although the usual caveats with LLMs apply. for example I can get a description of the contents of a window or control with one keystroke. Here is what it says about this HN page: The image shows a screenshot of a text-based discussion on an online forum. The discussion involves several users exchanging thoughts on various topics related to accessibility for visually impaired individuals, multimodal language models (LLMs), and coding. It features typical forum elements like usernames, timestamps, and reply links. [... summary of the comments] The color scheme is predominantly white with text in black, usernames in blue, and links in light blue, creating a contrast that's commonly seen in web-based forums.
  - cies 3y ago
    Thanks for this. Pretty useful I'd say!
  - bitwize 3y ago
    > It features typical forum elements like usernames, timestamps, and reply links. God, that's so on-brand for AI. "The image shows a group of humans. They are engaged in human-like activities like talking, eating, and hiding their insecurities. They feature typical human elements like hands, ears, and nipples."
- jareds 3y ago
  That's a good point. This may have been useful when I was working on internal webapps to determine if the UI looked like I expected it to. That was in 2010 though when it was more common to use plain HTML instead of single page frameworks. I'm glad I moved into back-end development.

by jareds 3y ago

peddling-brink 3y ago

Multimodal LLMs can take images and describe them as text, which I imagine could be useful in some cases, for the vision impaired.

RobMurray 3y ago
It can be useful, although the usual caveats with LLMs apply. for example I can get a description of the contents of a window or control with one keystroke. Here is what it says about this HN page: The image shows a screenshot of a text-based discussion on an online forum. The discussion involves several users exchanging thoughts on various topics related to accessibility for visually impaired individuals, multimodal language models (LLMs), and coding. It features typical forum elements like usernames, timestamps, and reply links. [... summary of the comments] The color scheme is predominantly white with text in black, usernames in blue, and links in light blue, creating a contrast that's commonly seen in web-based forums.
- cies 3y ago
  Thanks for this. Pretty useful I'd say!
- bitwize 3y ago
  > It features typical forum elements like usernames, timestamps, and reply links. God, that's so on-brand for AI. "The image shows a group of humans. They are engaged in human-like activities like talking, eating, and hiding their insecurities. They feature typical human elements like hands, ears, and nipples."
jareds 3y ago
That's a good point. This may have been useful when I was working on internal webapps to determine if the UI looked like I expected it to. That was in 2010 though when it was more common to use plain HTML instead of single page frameworks. I'm glad I moved into back-end development.