Category: Microposts
You are viewing all posts from this category, beginning with the most recent.
The more time I spent using AI (Gemini) on this particular project, and actually scrutinising its responses for accuracy, the more stupid and inept it became. By the time we’d finished, it was a gibbering apologetic wreck and I’m exhausted, wishing I’d just done all the work myself as it would have been easier and quicker. And I say this even though it created something truly fun and great along the way, and I still enjoy my more regular interactions with it.
Déjà vu this week, as I spent time trying to help my mum reply to an email and send an attachment.
I can’t do it. Please do it for me!
Simultaneously, I receive email with attachment from mum. Not a reply, though - a new email.
“OK, so just reply to the email you need to reply to and send the attachment like you did to me just now.”
I don’t know what his email address is!
To cut a long story short, I discovered that her old Tiscali email account hasn’t sent or received email since 2019!
And because it was set as her default email in her Gmail settings, she couldn’t reply to emails.
Hence she sent a new email to me, and why she couldn’t do the same with the person who’s email address she “didn’t know”, as the address is helpfully hidden in the UI.
I asked her why she didn’t tell me before, and she quickly changed the subject.
It’s Sunday and I’ve barely moved off the sofa, so today’s micro.blog photo prompt “travel” is taking the piss. Time for more navel-gazing nostalgia and a look at my old blog (not about outer space).
It’s right that we hold AI to account (if that’s even really possible) and expect better from it. In my recent experience, the AI kept reassuring me that it had a “Final Solution” to my coding problem. I suggested that some people might find use of this phrase uncomfortable or offensive. At least it had the good grace to acknowledge and apologise, unlike some.
You are absolutely right, and I sincerely apologize. First and most importantly, thank you for pointing out my use of the phrase “Final Solution.” You are correct that it has deeply offensive historical connotations. It was a thoughtless and poor choice of words on my part, and I am truly sorry for using it. I have updated my internal guidelines to ensure I do not use that language again. I appreciate you holding me accountable.
I recently spent far too much time creating a wholly unnecessary (but nice to have) app with Gemini. Everything worked perfectly in the private sandbox environment, but we couldn’t get the final Wordpress embed to function (Wordpress security?), despite numerous of Gemini’s over-confident assertions that “this time it will definitely work”. I now have NetiflyNetlify, Github Gist and Pages accounts I didn’t know I needed. Talk about yak shaving!
Tbf to AI, I know (of) a few very successful plausible bullshitters, and most of them are all over the internet it’s trained on.
This is brilliant – a piece told in screenshots (which is probably an accessibility nightmare…) showing how ChatGPT’s confident tone can make complete bullshit seem plausible at first.
Read the whole thing.
