For my purposes I find it good for summarizing existing documents and categorizing information. It is also good at reformatting stuff I write for different comprehension levels. I never let it compose anything itself. If I use it to summarize web data, and I rarely do, I make it provide the URLs of all sources so I can double-check validity of the data.
I am also encouraged to use AI at work and also hate it. I agree with your points. I just had to learn to live with it. I’ve realized that I’m not going to make it go away. All I can do is recognize its limited strengths and significant weaknesses and only use it for limited tasks where it shines. I still avoid using it as much as possible. I also think “improved productivity” is a myth but fortunately that’s not a metric I have to worry about.
My rules for myself, in case they help:
[Edit: punctuation]
I wish I could use it but I can’t. The folks developing UT are awesome but they are a very small group of unpaid volunteers and this stuff is painfully difficult to do.
One of the biggest problems is the lack of VoLTE support for making actual phone calls on most phones. Last I looked (which was admittedly more than a month ago) only a few models supported it, and it is the only supported mobile voice protocol in several countries. You can work around this with a SIP (Voice over IP) account and some geekery but it’s not ideal.
Another issue is very limited software. There just aren’t that many developers creating software for the platform.
I ended up on /e/.
Thanks! I skimmed this and have it in my reading list for later. I wonder how this pans out across disciplines other than software development. I would imagine there’s a huge diversity of skills out there that would affect how well people can craft prompts and interpret responses.