The computers had a lot of RAM, I suspect...

Excellent post. I think it pulls the curtain back further on how these LLMs work, and what they are really doing. With so much hot air around the topic it is pretty hard to get hard info if you are a non-expert, but the hangman example really illustrates that these are modeling language without underlying understanding at all. Makes me concerned for how people are using these LLMs (like the pythia at Delphi or some other oracle). Thanks for your post!

I wonder if HSR could be placed on parts of the existing interstate highway system to save costs and time.

Thank you, these are good points! I will look at Betterbird, I had not heard of that before. Also, the tendency to make things "Microsoft version" is a terrible habit that company has- examples I can think of from the top of my head include their development of ooxml for MS Office, rather than sticking to the existing ODF. Nominally, MS states that they do things like this for added features (for the doc example), but the outcome is reduced interoperability and lock in. I think a similar dynamic exists for Outlook/Exchange. Its really too bad.

I can't fully defend Outlook, but I do believe many serious professionals use it, limitations and quirks and all.

I am curious, though- what email client do you recommend instead of Outlook or other "poorly designed, poorly executed, feature lacking, functionality lacking email clients"? I have been looking, use Thunderbird myself (its basically OK), but I don't see a lot of great stuff in the "email client" space. I'd love to be wrong, though.

In my experience with it (I am using it daily, have been since 2021) Outlook (Classic) does not have very good search. Maybe they should nail that down before making it Outlook (Copilot edition)? Just an idea..

Remarkable how many responders to your comment quickly illustrated your point. And missed it. You may be on to something!

I also suspect there is a technical+legal reason for the bizarre limitation. A better way to handle all this: Should be OFF by default. You can turn it ON. If you toggle OFF and ON a couple more times within X time (a year, whatever), then the button should be grayed out, stuck to OFF. There can be a link to explain why (you are limited to toggle X times...) and maybe the date it can be toggled back on.

Back when I was on MacOS last time ended in 2021. I think?) I recall iPhoto/Apple Photos coming up collections of photos based on faces it recognized. I don't recall any toggle or opt in. It may have been on-device, though, which would be significantly more appealing to me than any cloud implementation. Does anyone know the story around that feature?

Here is to hoping these sorts of things do move on-device, with increased processing power and NPU going into new processors. Unfortunately I fear our data are too valuable to tech companies for them to implement this; unless an outfit like Microsoft has some sort of major change in their business orientation (toward the interests of those who buy their OS, apps, and devices, not those buying data about them) we will probably not see it from the usual suspects.

Good questions. I think its interesting in that it seems this round of Meta glasses users might end up as popular as the "glassholes" of about a decade ago. If so, that would be the death-knell for Meta's new product. I'd find that notable, as tech companies have been trying to succeed at some version of "strapping a screen to your face, with cameras" for some time now.

what about all the folks working "gig economy" jobs in big cities, like Uber eats? I looks to see if there are required checks for those, and I see there has been a recent crackdown. I wonder if digital IDs will make it harder for those businesses, and for folks in the cities to get their food deliveries? Or at least make them costlier?

..the merger of MacOS and iPadOS?

Excellent comment, you give the broader picture that many are missing. I have a somewhat related response when folks say they must pirate some particular media because the companies make it difficult to see or own. People often forget that you don't have to consume that content.

Critical thought is quite rare online, but its good to see some out there! Cheers.

Excellent question! My thought was that these computer programs are not really "knowledge" engines- they are engines that produce a statistically-driven facsimile of a reasonable answer to a question. Though clearly the area is developing very fast and I am not keeping up with everything, I think this is still the case.

Great post- This is some helpful insight for those in the US to understand differences with other parts of the world. I am in the EU at the moment and I can attest that life for drivers here is much tougher (and more expensive). In EU cities though, there are solid public transport alternatives. Add to that the fact that these older (compared to most US) cities were not designed for large numbers of cars and .. well, it makes sense here. People (and to an extent, systems) will do things that make sense in their context.

Ol Olsoc, thank you- interesting information there, and certainly worth considering. I believe at least the current administration's focus on the "TQ++" part of "LGBTQ++" bears out what you are saying. And I agree with you, I think very few people in the U.S. can be accurately classified as fascists.

As always, an amalgam of different factors were at play for different people; the issues you identify might indeed have been important to many. Giving some attention to what the Democrats have done to contribute to the current state of affairs makes a lot of sense.

In that vein, my wife feels another important factor in the election (and shunning of the Democrats) is that people felt Democrats lied to them regarding the condition of Pres. Biden. Especially by his VP. There might be something to that as well.