duskdozer

This article is in favor of using LLMs.

I didn't even know Android has an "AI Assistant" built in

Kind of weird to me you'd want to talk or type to turn lights off instead of just having a widget button

How is it excessive and nuts? If anywhere else is like it is by me, they've been ramming a bunch of giant warehouses at the least, and the increased noise both from the operation itself and the loss of natural barriers blocking other noise, light, and air pollution is really noticeably degrading things. And from what I understand, data centers are even worse due to their increased use of cooling and things like gas generators. It shouldn't come as a surprise to anyone that people are pissed off and increasingly so.

It is really intriguing, though my small attempt at using nix with home manager on pc failed, so I'm not so confident on getting it working well on android. Especially considering that I do a decent number of apk mods

What I hate about this is that there is a minimal amount of connectivity that I would actually want. If there was a bad crash and I'm incapacitated, I would want emergency services to be called. But I want basically nothing else. Well unless you count an FM radio 'connectivity' I guess. I hate the coercive tactics so many modern tech devices use.

Yes, I also used Ti[tanium?] Backup, and now use NeoBackup. Those still don't quite hit the spot for me of a full and reliable backup. It's close in many ways, but I've found system settings to be largely nontransferable, and many times apps would just seem to break or crash after restoring. So I have no idea how people do anything close to it without root, although I suppose it's possible at this point if you use a Google account and apps they probably have something.

Explain how? I've yet to find something reliable like this. At best you can have a decent number of apps automatically reinstalled to the original state, but not all, and most settings don't transfer, etc. And that requires root access, which most people don't have afaik.

Is it really that crazy though? Or is it an acknowledgement that the people in power would think like Peter Thiel and if given the choice between ending capitalism or ending humanity, would pick ending humanity?

And do you feel comfortable drawing all that down so that you're left with nothing at the end of the decade?

It could. But I'm not confident there will be a broad recognition of the fact if things become much worse than they are, or of the cause, or that there will be a desire to make things be better. There has been a long decline in quality in favor of optimizing for monetization and engagement, predating LLMs. But I do think there will be some pullback in things that actually matter if not in the SaaS FOTM.

>but performance on low-end devices and compatibility with accessibility technologies are not.

These are probably features

From what I've seen, most of this energy-per-lumen efficiency has just been funneled into outputting more and more lumens, not actually saving energy. It's only going to get worse with the new 2027 regulations coming in.

This is just assuming that more and brighter lighting is objectively good and better. It's not.

This is a common myth. Increased lighting does not reduce crime or improve safety[0,1], especially not poorly designed lighting - like the excessively bright and blue-tinted spotlights - which can even make things worse[2], by increasing glare, preventing our eyes from adjusting to the night, and increasing the number and contrast of relative dark spots.

Regardless, if one still lights up their own property due to anxiety, it's not a justification for causing light pollution. That light should remain on their property with angling and shielding.

[0]https://www.ojp.gov/ncjrs/virtual-library/abstracts/chicago-...

[1]http://jech.bmj.com/content/early/2015/07/08/jech-2015-20601...

[2]https://theconversation.com/more-lighting-alone-does-not-cre...

No, it really isn't.

By that time it may be decided that he's too big to be jailed.

His big mistake was clearly giving away what he got for free instead of trying to make a lot of money off it

Housing prices also more than doubled in real terms in the period of that graph https://fred.stlouisfed.org/series/QUSR628BIS

Moderators are still able to see your post history without extra steps.

Well, some data is deliberately provided at least. It's not a situation of other apps managing to grab the data like the facebook-localhost exploit

>The most prevalent third-party services included in CSP headers belong to Google Tag Manager (googletagmanager.com), Google Analytics (google-analytics.com), and Google Ads (googleadservices.com and doubleclick.net). Yet, as Table 8 shows, CSP policies commonly include other prominent actors in the advertising industry, such as TikTok and Meta.

I'm not conflating them. They were one of the examples I've used to say that the messaging promoted by public health officials is not the same as the expert understanding.

Well, that's why I scarequoted it. Morally I agree with you, and for the same reason I will not support LLMs and the companies pushing them. But as it is, the techbros and the adbros will continue to move fast and break things, and nobody will check them, so if some people can divert them into breaking slightly less, it is a small consolation.

I actually think this is 100% the correct move practically, as someone who thinks advertising is unethical. Keeping it out of this popular option while remaining open source allows the original to be clear from "wrongdoing" here and does not put so much pressure on creating an arms race. Then, the fork just adds sponsorblock, and things are ok.

As I see it, it's similar to the situation with FOSS passkey options. Currently the big projects are complying with all of the (increasingly silly, imo) demands of the standards org. If they did not, there would be significantly more pressure for attestation, in which ultimately only Google/Apple/Microsoft-controlled options would be permitted. Complying allows a fork to patch out the user-hostile behavior.

I haven't been able to watch youtube in the browser for ages. I get like one 10-20s chunk then nothing further. Do you use just the standard block lists?

Freetube if it's still working? I haven't tried it in a while.

I'm saying that "experts" here is not one group. The politicians and public figures choosing what information to publicize and disseminate are not the same people as those who worked on developing the vaccines.

How would an LLM know not to take that road then?

If your answer is you tell it, then the obvious solution to improve the app is to provide a way of specifying a road exclusion list, and that is extremely feasible.

I agree - we should have many more good fully-keyboard-accessible UIs!

I don't want to talk to my computer to use it.

It's incorrect to assume that the information you heard is 100% an accurate representation of the scientific understanding and not information filtered through "public health" which balances that with other interests. I'm not aware specifically what was known at the time about the initial vaccine formulation, but there are several examples where there was a disconnect:

- ~Feb 2020: COVID is known to spread through and linger in the air, not just spread via droplets or fomites; public health messaging fixated on simultaneous 6ft distancing and surface cleaning rather than air filtration and masking

- Spring 2020: it's known that masks reduce the risk of acquiring and spreading infection; in the US, public health messaging stated the opposite, in order to keep as many masks in supply to healthcare workers despite a major shortage of masks

- Jan 2022: it's know that an infected person can remain contagious for over two weeks; in the US, public health isolation guidelines are reduced to 5 days due to noncompliance and in response to pressure from the airline industry, which was experiencing a significant shortage of flight staff due to COVID infection

The issue with LLM guarding isn't that it's "excessive" in outputting edge case handling, it's that the result often ends up just suppressing an error that actually indicates there is a bug or that should be handled elsewhere in a different way.