People actually reparing their own cars! It is practically Communism! and has never happened before in the US!

Yes, it is a clear case of "follow the money"...

Nokias days were numbered long before..

When their primary business collapsed, they were basically left with the mobile network business and the huge number of patents they had from the time they actually did a lot of engineering and innovation.

both of those braches left seem to be doing "kinda ok"

I do not know why that post is modded negative, but the fact is that many companies are making that choise.

In case anyone is intrestested Lois Rossman (of the right to repair fame) did a lot of videos on the topic of how New York treated his company: https://www.youtube.com/playli...

History shows that when there are big changes in technology resulting in disruptions, we get riots, revolutions, wars and such.

The bigger the scale of the change, the bigger the disruption.

It is like that
"I don't always test my code but when I do, I do it in production" meme that was around a lot at some point.

Like the one on: http://pbs.twimg.com/media/BY5...

.. that there might be still some more money to be squeesed from farmers with new tech that is even harder to bypass the DRM on.

>I'm not seeing the "change the world" aspect of this.

It is not changing the world, but it is changing many situations.

I know of 2 such personally.

My friends cabin: It used to have a large external mobile phone antenna and it allowed him to get 3g, that was basically good enough to read emails and do light browsing, but now that 3g is shut down his options are 20 miles of cable, so microwave link option, 2g Edge , traditional satellite or starlink. He went with starlink.

Some of my wife's family lives in Philippines and not in any of the big cities, there they used to get bad LTE connection that quite often was zero bars, and cost a bit above 2000 pesos/month(about $35) at their data usage levels as it was based on used data. They now have a MUCH faster and more reliable Internet with starlink at 2700 pesos/month (about $46). Using as much data as they currently use on the old connection would be way more than that.

Neither of those are "world changing" but both provide a very good internet connection to compensate for other bad infrastructure.

It is not trying to compete with services in places where you can get a good wired internet connection.

People living that close to each other would require much more bandwitdh than is possible.

There are a lot of people in the world who live in places where such is not avaible due to things like being too sparesely populated, so it is not financially feasible to put cables in place.

Such places have been traditionally either without connections, served by microwave link or by "old style" satellite network.

In such situation there will likely not be too many other uses in the same cell and thus you get better service and a comparable price to old solutions.

Well, I use windows 11 daily and it is only a minor downgrade if you take (a lot) of time to tweak it to be usable.

I disagree on the base so trying to point out why:

>LLMs are not the same thing as AI

LLMs are AI. They may not be a GAI(general artifical intelligence), as there are many things they cannot do, but sayin LLMs are not AI is saying that a Ford model A was not a car because it could not do all the things we do today with cars.

>but when this stupid bubble bursts it will drag the whole industry down into another 20 year AI winter.

No, there will invariably be a bubble bursting, like we had in internet in 2000, but it did not result in "internet winter" for 20 years. What will likely happen is that all the silly companies in the space will die, but the ones doing actually useful will continue and new ones will rise up, but there will be less forcus on AI for a while, sure.

>While they have been an interesting curiosity for the last few years LLMs were always going to be a technological dead end.

Maybe, maybe not, we are far from knowing the results of the experiment.

>This is because there is literally nowhere for them to go - other than to try to scale up the models.

Except there is. More effcient use, specialist models, step by step reasoning and such have shown that there is a lot that can be done without scaling up.

>Getting them to do anything useful - like reasoning - has to be a bolt on.

Most likely yes, that is the current approach that you actually sub models and a master model that mostly allocates things. It is interesting to see where that leads.

>In any event, hallucinations mean they can never be used for anything serious.

Three things on that:
1) Humans do errors, forget things and so on too. The difference is that the hallucinations feel different than human errors and that is why they currently feel more wrong.
2) There is some interesting research where they can in small models already predict based on some behavior of the model if it is hallucinating. So it is interesting to see if that can be scaled up.
3) It is currently already being used for serious things.

>However, the funniest thing here is the idea you can manufacture 'high quality synthetic' training data. Nobody, at any level, has an actual clue what they are doing.

This part I mostly agree on except that the idea is not funny in it self and there is some research that suggests there might be ways, but the end part is way too accurate in most situations with the low current understanding of what goes on inside the model.

>As long as the money keeps washing around, everything is great. I suppose if you're paid millions then every month that you manage to hold on to your job is an achievement.

Oh yeah, that is obviously a big achievement..

Finland is already seeing a huge increase in both announced and under way large scale battery storages. That is because so much of the power comes now from wind that the prices have huge variation, so the expected pay back time is quite short.

.. how many whistleblowers and other inconvinient people die of "no suspected foul play" according to police.

US starts to at times feel like Russia, where inconvinient people tend to fall out of windows.

Except that then advertisers will look for other places to advertise on if the ads are not actually seen.

A more likely scenario is :
Step 1: Google will do the stuff automatically and then display the result along with ads.
Step 2: Recognise when it is their AI agent in ads and no give any money to sites for AI agent visits as they are not "real views", thus keeping all the money for the ads.
Step 3: The ad supported part of the web dies with much lower income.

Well.. I do not know... someone seems to disagree with you at least: https://autotribute.com/auto-i...