One comment made me think for quite some time: “This is literally the Microsoft Copilot pitch.” First you have AI create a beautiful artifact in Powerpoint et. al., then you waste other people’s time, and they evade your move by letting AI boil it down to the essentials.
I have three exhibits from Apple to illustrate how AI is marketed to stupid and lazy people.
I likened it to the ouroboros, the ancient dragon that eats its own tail! https://benpoole.com/weblog.html#29-apr-2024-ai
The time economics and appeal are very similar to voice messages. A simple point that could have been single paragraph of text that could have been easily understood, consumed, and responded to even while performing another task, becomes a convoluted 2-minute recording that puts the burden of making sense on the recipient, but demonstrates that the sender took 2 minutes of their time to make their point, rather than sending a short paragraph with their ask that could have been handled in 15 seconds.
The common element is an implied notion that the sender’s time is more important than the recipient’s, and signaling through time consumption, not the density of the information, how much effort the sender invested – when they have really not.
Do you think LLMs have benefits? Which?
I am still unsure about it. But my gut feeling is, that the advantages exist but are a lot harder to describe to critical people.
And this makes me uncertain, whether the advantages outweigh the disadvantages. Or if the latter are just easier for me to grasp.
I have a broader perspective on this. Yes, it’s true that for some, it’s a way to accomplish tasks they wouldn’t have done otherwise. It’s also clear that with AI, I can “inflate” things and just as easily “deflate” them. For me, there’s a deeper meaning here.
Summarizing meetings – as we see in Video 2 – is incredibly helpful. We’ve already gone through this… when meeting attendance is less and the benefits of interaction diminish, it becomes a problem. In many meetings, people are now actively participating again, and AI functions generate added value, allowing us not to take notes constantly.
The deeper meaning: It’s obvious that those who don’t contribute and only rely on AI will face challenges. Whether in their career or family… these things work for a while before they backfire.
Anyone who realizes that AI is taking over parts of their tasks should start thinking seriously. This can be positive – we will certainly need skilled workers in the future, and perhaps their talents lie elsewhere, not just between the office chair and the keyboard.
I’m not an Apple user, but if I was I’d be rather worried using “Apple Intelligence” when I see reports like these (which are not a one-off):
https://www.bbc.co.uk/news/articles/cx27zwp7jpxo
Now imagine it makes similar mistakes when summarising your presentations or emails to your boss and/or customers with similar mistakes.
(And from my own experience others are just as bad, I’ve had ChatGPT recommend me a beach I know very well as not safe for swimming as a popular swimming destination)
Having just finished a website (nodejs, django, html, js, css) with VSCode (and Copilot enabled) I hereby welcome our new overlord! It saved me from tons of Cmd+C/Cmd+V keystrokes. 🙂
At the same time, I hereby admit that programming websites and services is very deterministic and apparently an easy job for LLMs.
So your milage may vary, I guess.
For the last year, I’ve given sales people pitching AI two use cases: students expanding ideas and their professors summarizing those papers.
It reminds me of something that always bothered me with textbooks: what’s the difference between publishers? I mean: the fluff, the redundant words that tie the actual information together. Can we read and process bullet points or do we need 30-page pdf’s? A bit like 300-slide powerpoints and their complete opposite.
From your FAQ: “You will find the slides completely useless if you have not been present.”