Lorgo Numputz<p>I use a an open-source tool called "rclone" to back up my data to the AWS S3 service; this data is then quickly migrated from the base S3 storage tier to another tier called "Glacier", which is less expensive.</p><p>The tradeoff for the savings is that files in the Glacier class are not immediately available; in order to be able to restore them I need to request that they be restored in S3 so I can copy them. Typically you restore them for a limited number of days (enough time for you to grab a copy) before it then reverts back to Glacier class.</p><p>The other wrinkle is: The files are encrypted. Not just the files but the file names and the file paths (enclosing folders/directories).</p><p>Here is the tricky part: The backup software does not have the ability to request a file be restored from files stored in the Glacier tier. I have to do that using the aws command line or the console. This is doubly tricky because I will have to request the exact file using the encrypted filename and path... not the name I actually know the files as.</p><p>So it turns out that rclone can actually tell me the encypted filename and path if I ask it correctly because of course they've dealt with this problem already. :) </p><p>I thought to myself "Here is a chance for ChatGPT to show its quality".</p><p>I'll skip to the chase:</p><p>ChatGPT gave me exactly the *opposite* instructions of what I asked for. </p><p>Instead of telling me how to get the encrypted filename path from the unencrypted equivalent it, instead, told me how to get the plaintext from the encrypted filename - which I didn't have. This is using the latest ChatGPT 4o, the very latest.</p><p>I question the usefulness of this kind of tool (meaning ChatGPT) for anyone who isn't an expert. I've done this long enough that I know of other sources to look at (such as the manual pages) but if you aren't that savvy I'm not sure how you would find the right answer.</p><p>The ability to regurgitate unstructured data with LLMs is amazing - almost magical when I compare it to other efforts to do the same that I have been involved in previously.</p><p>But the ability to summarize and present the data in an accurate, useful form is nowhere near acceptable.</p><p><a href="https://beige.party/tags/ai" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>ai</span></a> <a href="https://beige.party/tags/llm" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>llm</span></a> <a href="https://beige.party/tags/llms" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>llms</span></a> <a href="https://beige.party/tags/chatgpt" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>chatgpt</span></a> <a href="https://beige.party/tags/aws" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>aws</span></a> <a href="https://beige.party/tags/s3" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>s3</span></a> <a href="https://beige.party/tags/rclone" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>rclone</span></a></p>