antonkar

Serious

The proposed project is a non profit, I don’t think it can be a for profit legally (it didn’t stop AI companies, though)

Imagine taking the whole Web, removing spam, duplicates, bad explanations

It will be the free new Wikipedia+ to learn anything in the best way possible, with the best graphs, interactive widgets, etc

What LLMs have for free but humans for some reason don’t

In some places it is possible to use copyrighted materials to educate if not directly for profit

A bit related: AI companies distilled the whole Web into LLMs to make computers smart, why humans can't do the same to make the best possible new Wikipedia with some copyrighted bits to make kids supersmart?

Why kids are worse than AI companies and have to bum around?)

The only way to prevent robots from being jailbroken and set to rob banks is to move GPUs to private SOTA secure GPU clouds

The topic-starter added a comment there:

"So, yep, I really wasn't even using my VPN that much recently - it's the same one I used for months.

And even bigger problem, of course, is - why just looking at my problem where a user with almost 1000 followers and a community that actively chats and spends a lot of time on X (usually, it's good for ad supported platforms).

So why even quite active paid users can't be checked in at least a day (I think it should be minutes but alas), why 5-7 days! ;-)"

It's a lot, but you're right that the more realistic scenario is just running an existing open source model - sadly those are trivial to misalign

So we better have SOTA clouds: people don't update computers for months and as Bredolab shows - their GPUs are up for grabs

Ideally and int. org. will do it but they are slow and countries don't cooperate enough, so a startup is more realistic

Sadly it is almost inevitable we'll have another Bredolab-size botnet (30 mln computers, 1% of all) by cybercriminals but this time it'll be an AI botnet.

As soon as AI components will be useful, they'll be in botnets. And with such a size, you have more compute than OpenAI and can train a frontier misaligned model or modify an open source one:

The solution is to not just have 10% of compute in clouds like we have now but at least 50% in SOTA clouds, to have more compute than botnets. Have AI model App Store, probably NVIDIA will become like Apple:

Will have at least the most minimal checks on what runs on their hardware

We'll have to do it but better to do it early not late

It can be a unicorn startup that NVIDIA will want to buy (motivating gamers to put GPUs in clouds is easy, same with others really, you can share $30-1500/month with them by renting GPUs from your cloud to corporations and others)

There is Salad but they don't secure the hardware and software, so can't really have corporate clients. Amazon AWS, Azure, others, show that it's a real business. Unicorn-sized just in the USA (even bigger if global) if you'll do the math

Yes, it’s great.

The densest UI is a direct democratic simulated multiverse: imagine a long exposure photo as a 3D scene, in which you can walk or fly.

“Point and scroll the mouse wheel” to focus on a particular moment of time, make it crisp, or “scroll back” to see billions of years of time as a hazy long exposure scene.

Example: 14 bln years of our planet look like a hazy ocean (our planet was a water-world most of the time), with the Sun arcs static in the bright sky.