I think I can contribute something to the “privacy” aspect. But I’ll say first that I have noticed the same thing. There are some toxic behaviors that feel more common in these circles than what I have experienced elsewhere.

There is a lot of confusion around European data protection rights and privacy. EG the GDPR is often wrongly called a privacy regulation. In reality, privacy and EU data protection rights are entirely separate.

In the Charter of fundamental rights of the European Union, you will find privacy in Article 7 and data protection in Article 8.

EU data protection works similar to copyright in that you have rights over data. Personal data is defined as any data that is “directly or indirectly related” to you (GDPR). It does not matter if the data is public or private, sensitive or banal. It doesn’t even matter if the data can be connected to your real identity. That’s quite unlike what one would think of as privacy.

So, it does not matter if people expected their communications to be secure or not. “Reasonable expectation of privacy” is a concept in US law.

Comments, posts and DMs are personal data because they are connected to a user who is a person. If any other person is mentioned, then this mention is their personal data. You could even argue that some post or comment also becomes someone else’s personal data when they reply to it. Such texts cease to be personal data only when the connection is irreversibly broken. As long as the connection can be restored, it remains personal data, even if that requires access to information that isn’t readily available.

When a DM is sent to some unauthorized recipient, that is literally a violation of the senders fundamental rights. In truth, this is relatively serious compared to some other stuff that causes outrage or gets the authorities involved.

It might have been legally required to notify the authorities of such a data breach within 72 hours.

Turning that on is probably a GDPR violation for those in Europe.

ETA: Don’t shoot the messenger. I won’t be suing.

That gives 30% to Steam though. Better use Itch.io as linked on the github page.

https://anuke.itch.io/mindustry

I prefer Threads.

You are a very brave person.

Defeatist opinion.

The commercial alternatives hope to make money with every additional user. They use AB testing and statistics to streamline the on-boarding and to increase engagement. The result may not be in the user’s interest (doom-scrolling, ragebait, …) but it works.

For a fediverse instance, any additional user is a cost, not the promise of money. Financially, you wouldn’t want that. Those who fund instances are giving a gift to the world for their own reasons. You can accept the gift or not. Those who keep instances running with donations will usually want to sustain the community of which they are part. They probably don’t want it to change very much.

So, I don’t think matters will change. Partly because the psychological engineering is antithetical to the fediverse ethos (as I see it, in my humble opinion). But mostly because the outcome we see is an inherent result of the incentive structure.

Link to bill (SB-1047):

https://leginfo.legislature.ca.gov/faces/billTextClient.xhtml?bill_id=202320240SB1047

It’s not clear if this is piracy. In the US, it’s obviously an ongoing fight. Basically, what you describe is “books3”, put together with scripts by Aaron Swartz.

It’s legal in Japan, if the purpose is only AI training and not enjoyment. I’m not sure if there are issues regarding DRM or such.

In the EU, the dataset and resulting model would be illegal. Any business offering the model would be in hot water, but I think internal use would be fine.

It probably says somewhere where you dled the model. It’s also in the metadata. I forget where it’s displayed. Maybe in the terminal window.

Things you should know:

• What a token is depends on the model.
• Context takes a lot of (V)RAM.
• People modify models to increase the context but that often doesn’t work well. Watch out for the model missing things, esp. in the middle of the document.

L3 is probably not the right base for the task. Maybe Phi-3 or Cohere.

Haywire. You don’t see that often anymore.

It remains a dangerous dead end. Any competent fraud will remove the watermark, or use a generator that doesn’t add one. Giving people the idea that the absence of a watermark makes something trustworthy, can only help bad actors.

Is that the 2B or the 7B?

@[email protected] Can you draw ascii art of a tractor?

This sounds like some weirdly petty political wrangling that would delight any full-blooded bureaucrat.

The desire to make demands about training data is weird. Open source has never included a requirement to provide documentation of any kind. If there was some requirement for documentation, few would care and most just do their thing anyway. FOSS licenses facilitate sharing by giving people an easy way to make their code legally usable by others.

There’s nothing that quite matches source code + compiled binary. There are permissively licensed datasets and models. I’ll call either open source. Neither is equivalent to source code but either can be a source.

minimize their own liability

I think maybe that is behind some earlier releases but Google is already operating Gemini. It’s not going to help them with liability.

Eventually, the cost of making/releasing these models is simply insignificant. Maybe they hope that people messing around with them will give them tips for Gemini. Maybe they hope to familiarize potential future employees with Google specific AI (giving them a larger and more qualified hiring pool).

They may also hope for a bit of good will, but that may easily backfire.