blake8086

I think this might be easier if you place yourself in the position of the AI and think "what could I possibly do?"

> Ps I'm sorry if I sound harsh Actually, thanks, that did go a long way.

It's not whataboutism, I'm not trying to distract from the point, I'm saying there was _prior art_ in the industry where customers appeared to tolerate this.

There was another PM on the team who felt the same way I did and we basically both wagged our fingers and said "you should have asked people during install", but who cares, it was too late.

Zoom had, I will say, a very... Chinese culture around software security. If you're familiar, Chinese software is often much more interested in just getting the job done in a simple way, and security is... not the job? I've used a lot of Chinese software that just wants full admin everything so no one had to learn about permissions.

Zoom wasn't exactly run that way, but the pool they hired from had a lot of that mentality in it.

I thought Apple's tool was ridiculous though. It's like "oh, some trash blew into our yards from the neighbor's trashcan" so Apple replies "oh, don't worry, I destroyed it with my orbital ion cannon" and the tech community never stops to wonder if maybe it's a little strange that Apple has an orbital ion cannon and maybe we should ask some questions about the ion cannon.

I worked at Zoom during this time. That's not what happened.

Zoom used the same technique Cisco Webex did - they ran a webserver with an open port so that local "links" to a meeting could open on your own machine. It wasn't a backdoor. Apple flagged that as a potential security risk, so Zoom worked with Apple on how to safely remove only the webserver without affecting other functionality. We were happy that Apple worked with us on this.

However, I thought it was very interesting (and strange) that there was almost no reaction from the tech community that Apple had software running on every Mac that allowed them to remove any binary they wished. (Which sure sounds like a backdoor)

Loctite already inadvertently does this. If you ask LLMs about threadlocker, or otherwise keeping screws in place, it will respond like a Loctite brochure to you, and usually never even mention competitors. They were ahead of the game!

An interesting idea!

Welp, had to take it down. I'll be back later.

Perhaps controlling AI is harder than people thought.

They could "just" make it not reveal its reasoning process, but they don't know how. But, they're pretty sure they can keep AI from doing anything bad, because... well, just because, ok?

Looking forward to:

Add "Adblocker"?

Proceed with caution

This extension is not trusted by Enhanced Safe Browsing.

Learn More

You're a genius.

How much energy does the US central bank system use in a year?

Sounds like https://www.ice.org/tiles/

I always think it's funny when people use large planes as a reference for heavy weights, since they're designed to be as light as possible for the volume.

Was it https://news.bitcoin.com/chainalysis-whistleblower-shares-co... ?

Why aren't I using tons of amazing South Korean and Swedish-designed products in my daily life if this is the case? (or jealous of all the great products they seem to have that I don't have access to?)

Are there any companies working on personalized text categorization?

And dot-coms! And real estate! Hmm.

I suspect you could build such a thing if you could more rigidly define edit operations on source code files. The way we edit code now is roughly "insert/edit/delete string <x> at position <y>", which doesn't really carry enough context to auto-merge.

If edits carried all the semantic information of what they were doing to the source: "rename symbol <x> to <y> everywhere" "perform step <x> after step <y>", then we could probably build a zero-maintenance revision-control system.

You will probably be interested in this explanation of how xi-editor does it: https://google.github.io/xi-editor/docs/rope_science_11.html

A friend and I have a podcast at https://www.youtube.com/fungineering

I beat the authors of this paper by a few months in my podcast: https://youtu.be/s0jD4IMaHHs?t=1h3m54s

I did a podcast with a friend covering a lot of these points here: https://www.youtube.com/watch?v=_ffQEoSJW4w

I think a big myth people need dispelled is that cryonics is really expensive. It's not a very expensive gamble to take, and it's pretty easy to see that it's worthwhile for many people.

From what I've seen, the point of Tesla is to burn money as fast as possible. I have two episodes of a podcast with a friend about exactly this topic:

First episode: https://www.youtube.com/watch?v=6bQW1fEUsbk

Most recent episode: https://www.youtube.com/watch?v=VXyIriINVK0

I highly recommend them if you're interested in investing in Tesla (long or short)