Originally published at: Apple AI helpfully summarizes breakup text - Boing Boing
…
Sometimes I wonder what my dating life would have been like if I’d been a single person in the age of smartphones but mostly I’m just glad I missed it.
Just so I’m sure I understand: the contents of the email included the information that the poster’s significant other wanted to break up with them, or?
If that’s true, what is the problem with the summarization? The poster would be just as crushed if they simply opened up and read the email.
Are we actually asserting that the LLM should respond emotionally (I.e. with empathy)? We’re actually requesting this? If so, then I guess everyone else is much more ready than I am to:
a. believe that this is currently possible and
b. welcome artificial generalized intelligence into their lives.
Because sometimes human communication isn’t about maximizing efficiency?
For example, if a loved one died in a war would you rather get the news via
- A uniformed officer delivering the bad news in person,
- A sympathetic and tactfully written letter, or
- A gold star sticker unceremoniously slapped on your front window?
you are writing this as if its a fact, without any quotes, gail sherman. -deleted- e/ @danimagoo; I dont know, the preceding sentence doesnt make it much better, does it? (included the preceding sentence in edit)
Maybe include the preceding sentence, which makes it quite clear that she’s just reporting what Apple is claiming about the AI:
Apple is promoting Apple Intelligence as a more personal AI. It will get to know you and help you write better, help Siri understand you better, and improve your life.
perhaps it can learn a little empathy before it is released to the public
After 30+ years in the tech industry, I can confidently say that – for a variety of reasons --incorporating empathy into any given product is a very low priority for most coders, executives, and funders. “AI” is definitely bringing that issue into stark relief, though.
Timing. Phrasing. Lack of user options. Inability to distinguish a deeply personal matter (clues: “relationship”, “belongings”, sender is the girlfriend) from a routine one.
Within five minutes of considering this scenario I came up with a more humane, empathetic, and respectful summary that should be well within the capabilities of any LLM. But hey, maybe I just “think different”.
As is often the case in the “move fast and break things” tech world, it’s less a question of consumers welcoming a half-baked (and perhaps never fully baked) technology into their lives than it is one of corporations foisting it on them.
What we get:
“Jeeves, summarize my mail for me.”
“Certainly, sir. The riding club is reminding you of the annual meeting, the business deal you were waiting for has been delayed, and it seems your girlfriend is ending your relationship.”
What we want:
“Jeeves, summarize my mail for me.”
“Certainly, sir. The riding club is reminding you of the annual meeting, the business deal you were waiting for has been delayed, and… sir, I think this letter from your girlfriend you should open yourself.”
Edit: as @gracchus notes, this is a challenge even for humans, so I don’t expect something ruled by maths to do a better job.
This is similar to the alternative I came up with: “This personal text from [girlfriend] is better read in full and might upset you on your birthday. If you’d like to wait until after your birthday to read it, we can reply to the sender that you’re doing that.”
Not perfect and still potentially upsetting, but much better than what was delivered. I’m not an “ignorance is bliss” person and would probably read it anyhow, but I can understand why a lot of people would appreciate the consideration.
I think it does. She’s not directly quoting Apple, so she can’t put that in quotes. She’s paraphrasing. The preceding sentence, followed by a sentence that is clearly connected to it, followed by a paragraph break, is pretty standard writing for something like that. Sure, she could have made it more clear by starting the sentence with “Apple goes on to imply that…” but that would read a little clunky, and I think it’s unnecessary. It was clear to me that she was still just saying what Apple was saying, and not that she believes it.
ETA: If this were legal writing, I would absolutely be expected to add that “Apple goes on to imply that…” We were trained in legal writing to leave absolutely no room for ambiguity in attribution. But we would also be expected to include formal citations. This isn’t legal writing.
And, bringing things 'round, most BB readers – especially those familiar with both the site and with Gail Sherman – can parse that graf better and with more nuance than an over-literal “AI”.
sure. am I biased when I think the headline alone sounds a bit like a sponsored post? the “helpfully” is what gets me. its all kinda softbally.
Since it wasn’t a BB Shop post, I assumed that it was the dry snark and sarcasm we expect and love from the BB Authors.
It doesn’t help my sympathy for this guy that he complained about his GF breaking up with him… either she doesn’t appreciate him (in which case he’s better off), or he fucked up and she’s fed up with his shit (I’m leaning to this because, again, publicly posting about his GF breaking up with him).
But yeah, maybe just read your GF’s email instead of asking an AI to summarize it. Unless he has ALL his email summarized by default, in which case YES emotionally triggering information will be summarized since there isn’t a “protect my feelings” setting. I’m just pleasantly surprised that the summary was actually accurate.
Does Apple let a user exclude messages from certain senders from “AI” summarisation? Similar to how I can turn off “do not disturb” mode on my phone for important people in my life?*
On the one hand, that kind of granularity would make sense as a direct and elegant solution (more than the one I proposed above) to the problem in this case; on the other hand, this is an industry that often defaults to an all-or-nothing binary for features, and a company with the reputation of thinking it always knows what’s best for its users.
[* It’s not just “protect my feelings” mode, it’s “some people/messages are more meaningful to me than others” mode.]
Did he ask an AI to summarize it though?
If this “feature” is anything like most rollouts of AI technology then it’s the default setting whether users actually wanted it or not, just like how every Google search result now returns a bunch of AI-generated summaries you have to scroll past.
Yes, and Co-pilot recently appeared in your Windows 11 start bar whether you asked for it or not. And everything in Windows and Adobe now defaults to cloud storage, making you do extra work to save files locally.
These changes are not implemented because the user asked for them.
very dry, then. thnx.