I don't think there's anything wrong with those sorts of general observations (I mean, who remembers dozens of phone numbers anymore now that we all have smartphones?), but that said this non-peer-reviewed study has an awful lot of problems. I mean, we can focus on the silly, embarassing mistakes (like how their methodology to suppress AI answers on Google was to append "-ai" into the search string, or how the author insisted to the press that AI summaries mentioning the model used were a hallucination, when the paper itself says what model was used). Or the style things, like how deeply unprofessional the paper is (such as the "how to read this paper"), how hyped up the language is, or the (nonfunctional) ploy to try to trick LLMs summarizing the paper. Or we can focus on the more serious stuff, like how the sample size of the critical Section 4 was a mere 9 people, all self-selected, so basically zero statistical significance; that there's so much EEG data that false positives are basically guaranteed and they talk almost nothing about their FDR correction to control for it; that essay writers were given far too little time for the task and put under time pressure, thus assuring that LLM users will be basically doing copy-paste rather than engaging with the material; that they misunderstand dDTF implications; the significant blinding failure with the teachers rating the essays being able to tell which essays were AI generated (combined with the known bias where content believed to be created by AI gets rated lower), with no normalization for what they believed to be AI, and so on.

But honestly, I'd say my biggest issue is with the general concept. They frame everything as "cognitive debt", that is, any decline in brain activity is treated as adverse. The alternative viewpoint - that this represents an increase in *cognitive efficiency* by removing extraneous load and allowing the brain to focus on core analysis - is not once considered.

To be fair, I've briefly talked with the lead author, and she took the critiques very well and was already familiar with some of them (for example, she knew her sample size was far too small), and was frustrated with some of the press coverage hyping it up like "LLMs cause brain damage!!!", which wasn't at all what she was trying to convey. Let's remember that preprints like this haven't yet gone through peer review, and - in this case - I'm sure she'll improve the work with time.

Not even a Google Search. Literally just talking to it before giving the toy to their children to see if, asked to talk about something harmful, it does so or refuses. Are the parents in your mind too tired to literally speak?

I first heard this comparison back when IDEs were young (kudos to Larry Masinter, at Xerox PARC at the time).

Amplifiers don't really know or care what they are amplifying.
If you tell them to create good, bad, immoral, or dangerous code, they'll try to comply.
Laws against bad uses of LLMs just make them illegal - they don't make them impossible.

Mediocre programmers with IDE/LLM support will create reams of mediocre code, at best.

My preference for Firefox is based on the bookmark sidebar, with nested folders of bookmarks (similar to the way file are nested within folders on a disk drive).

The problem is that I've frequently heard of "anonymized" data being de-anonymized. It's not something I worry about a great deal, but perhaps I should worry more...and I've no real way to tell.

There are several alternatives, depending on your use-case. My fall-back is Falkon (from KDE).

FWIW, I didn't detest Pocket. OTOH, I never used it, either. But "universally detested" is wrong. I thought of it as "dead wood", but that's a very different category. There are LOTS of software capabilities that I don't use.

To me, panning a browser because it isn't optimized to run on a phone is silly. Saying I prefer a different browser on my phone would be sensible (if I though web browsing from a phone was sensible...but with my eyes that's never going to be true).

I have a paid subscription to the Washington Post (I live in the DC suburbs), so I get their content sans paywall. They let me create a few non-paywall links per month, and I share them when I see something the rest of the net should see without the paywall.

I pay Reddit annually, and I get their content sans ads. Whenever I see Reddit before I log in, I want to go wash my eyes out.

The real problem is I don't want to spend the money for a full subscription to every news source I read occasionally.

If there was a way to pay, say, $10/month to get 30 links from a basket of paywalled news sources, I'd be on it in a heartbeat.

Well..."sort of virus first" is probably correct, but that doesn't mean that things don't sometimes go into reverse. The "sort of virus" couldn't be like the current stuff, because it couldn't be parasitic. And there are arguments that a "sort of cell" evolved before the genetic machinery. Nobody really knows, (See "metabolism first" https://ancillary-proxy.atarimworker.io?url=https%3A%2F%2Fui.adsabs.harvard.edu%2F... for one argument.)

I, personally, suspect that the proto-"cell walls" and the proto-"genetic machinery" evolved their first stages independently, and mixed together later into something with superior "stability".

Well, they might let me remember people's names. There are probably other uses, but that's the one that occurs to me.

I accept that it is breaking the law. That doesn't mean any official is going to enforce that law. There are lots of laws that are frequently broken in obvious ways, and are enforced only when it is a convenient excuse.

How far in the future?
I can imagine useful AI based "smart glasses", but that description doesn't fit what I've heard so far, and my prescription means I couldn't try them if I wanted to do so.

My wife is a sign language interpreter, and does a lot of remote work, especially since covid.

To handle a meeting on Teams, sign language interpreters need to pin two video streams - the current speaker, and the deaf client(s).

It is essentially impossible to do this in Teams - they routinely open up a separate Zoom session for interpretation.

You'd think the inability to do this would be an ADA violation...

I'm not sure you understand what jailbreaking means in the context of AIs. It means prompts. E.g. asking it things and trying to get it to make inappropriate responses. Trying doesn't require any special skills, just an ability to communicate. Yes, I very much DO think most parents will try and see if they can get the doll to say inappropriate things before giving it to their children, to make sure it's not going to be harmful.

(Now, if Mattel has done their job right, *succeeding* will be difficult)

If it's doing auto-transcription it's probably training an AI with your information, This could be a major security problem.