We understand reasoning enough to know humans (and other animals with complex brains) reason in a way that LLMs cannot.

While our reasoning also works with pattern matching it incorporates immeasurably more signals than language - language is almost peripheric to it even in humans. And more importantly we experience things, everything we do acts as a small training round not just in language but on every aspect of the task we are performing, and gives us a miriad of patterns to match later.

Until AI can match a fragment of this we are not going to have an AGI. And for the experience aspect there’s no economic incentive under capitalism to achieve, if it happens it will come out of an underfunded university.

It still solves one end of the problem. The cable problem probably needs enforcement from law side.

Like all dating apps they have a strong incentive to promote superficial short-term connections over long term relationships so they have repeated customers. The worst thing that can happen to a dating app is their users finding a long term partner.

In other news, asking Nick Clegg before emptying out his home would kill the robbery industry.

There have been former Google employees saying they had solutions to deal with SEO, but they were vetoed by the ads side of business, since better results reduce “engagement”.

This reminded me immediately of The Coffee Machine short story published last year. Reality manages to be stranger and weirder than fiction.

And centralization solves this how? The other social networks are giving more checkmarks to grifters and scammers than they are giving them to honest people because, spoiler alert, con artists are very good at both building a following and paying bribes.

It’s inevitable for ads supported social media.

They need to keep you “engaged” to show you more ads, and the most effective way to do it is to foster conflict.

You underestimate how much interface and algorithms impact how people interact with eachother.

We didn’t have the Intel 8080 chip that the Altair computer ran on, So Paul got to work writing a program that would simulate one on Harvard’s PDP-10 mainframe. This allowed us to test our software without needing an actual Altair.

So, basically:

We are going to write a compiler, but first we are going to write a Virtual Machine from scratch because the hardware is too expensive.

Good times.

They can’t close the kernel. They already distribute Android with proprietary software - for example Google Play services and DRM services.

It’s almost like LLMs aren’t capable of understanding or something. /s

And analytics. And offloading as much computation to the client, because servers are expensive and inefficiency is not an issue if your users are the ones paying for it.

Just wait until the vibe coder overwrites 3 months of “work” with garbage because ChatGPT never told them about git and then decided to poop the pants.

Antennapod and not depending on the whims of whatever a CEO feels like today.

Please do. I want to train an AI on Disney movies. (Dear Disney lawyers please take the bait)

All of the above and more? There’s always the risk of something falling through the cracks, so the more layers of security measures you add/can afford the better.

90% of security vulnerabilities are caused by “let’s just use/do this for now and change it before production”.

Two Mac OS VMs enforced at Kernel level, which wasn’t a thing on Intel Macs.

Did they remove the stupid limit of 2 VMs per machine or is it still an expensive toy?