Forum: Author Hangout

most content is randomized and original

If it's randomised, it can't be original.

it was worst times of the times of was it best the ;-)

AJ

plagiarism or not, they can't hold copyright, because they are not persons as defined by law. Their output can be regarded as 100 monkeys pounding on keyboards, and chance took a part.
The machine owner can't claim copyright, because the owner didn't create the work.

This is common sense (no relation to "The Law") which is sadly lacking, in today's society.

Disclaimer: I am not a lawyer, and I didn't stay at a Holiday Inn Express, last night. YMMV

Not sure that has been defined yet. Corporations are persons under the law and can hold copyrights so if a person or a corporation held a license for a particular chat bot the copyright would be in the name of the person or company. Whoever set the parameters for the story would be the creator. Unless it duplicated something already copyrighted or written it would not be plagiarism and could be copyrighted as an original work.

There is huge can of worms regarding inclusion of copyrighted material in the training data, but that's a separate discussion.

How could it work without the training data?

AJ

As a language model, it predicts the most likely next words; it doesn't recreate any source text (unless prompted to do just that).

I would strengthen this. It can't recreate any non-trivial source text even if prompted to do so with any reliability. It could recreate 'close' source text (where close may be in the eye of the beholder), but the difference between model size and training dataset size is beyond the current technology for data compression in many cases.

An enormous amount of information is 'lost' in training, but that's the goal - this isn't compression, it's modeling.

It's (of course) far more likely than the monkies and typewriters to manage that, but there is a ton of data loss. I'm more familiar overall with the AI for image synthesis, but in either case there are inevitable and enormous data losses.

Some of it is that both image synthesis and chat also tend to incorporate random seeds, so the same prompt might not reproduce the same thing.

And, yes, I expect it would go meandering off, which prompts an enormous philosophical debate about the nature of creativity.

We could even lose this forum.

WLPC is in Canada, so SCOTUS doesn't much matter. What matters is Canadian law.

Someone will still TRY to sue them.

They'd have to go to Canada. No US court would have jurisdiction and without personal service (i.e. the paperwork being served) to a US address, good luck even getting in front of a judge in the US (beyond him telling them to get the f-ck out of his courtroom and stop wasting his time).

IANAL, so take my comments with that in mind.

Lawsuits per se aren't my worry (barring some reciprocity agreement that I'm unaware of).

What would worry me would be a change in US law that affected transmitting the data itself. If transmitting the data from a website was banned, we'd need VPNs (and VPNs would likely be on the chopping block at that point), etc.

I'm worried that we're seeing increasing openness towards simply banning content, and it's a very short step from that to banning transmission of data to and from sites that hold banned content.

I'm worried that we're seeing increasing openness towards simply banning content, and it's a very short step from that to banning transmission of data to and from sites that hold banned content.

That is m worry as well, but end-to-end encryption and VPNs will get around that, and it's not difficult to run

your own VPN endpoint in a country that doesn't band the material.

Sadly, though, the totalitarians are winning with the 'think of the children' arguments and soon enough encryption will require a license and any use of encryption without said license (and 'golden keys for the government') will itself be a crime with lengthy mandatory prison sentences.

Orwell was a f-cking optimist.

I agree with what you are saying. That's why the word TRY is in bold type. We are talking about the USA where you can find a Lawer willing to file anything and just maybe a judge to agree with them.

And WLPC is in CANADA. US Courts and US Court rulings do not apply to what can and can't be published.

In order to sue, you need standin and jurisdition. No US Judge has jurisdiction over WLPC. Lawyers know this and could be sanctioned or even disbarred for filing a suit when they know it has no chance of having standing/and or jurisdition.

I assume you don't live in Canada though. In the UK, there's a blacklist of sites that ISPs aren't allowed to connect to. Mostly child porn, but sometimes sites are banned for political convenience.

AJ

I assume you don't live in Canada though. In the UK, there's a blacklist of sites that ISPs aren't allowed to connect to. Mostly child porn, but sometimes sites are banned for political convenience

I'm in US.

And that's what VPNs are for. I never connect to the internet without a VPN and my exit node is almost never in the US (unless I need to access something geoblocked).

Avoiding those blocks is trivial.

A good choice for VPNs is Mullvad (based in Sweden) with no-logging, completely anonymous access (you can buy time via a pre-loaded card from Amazon), and post-quantum encryption.

That still doesn't stop them does it? Just look at all the bogus lawsuits in the last 2 elections.

And sanctions have been issued and are forthcoming. At least in those suits they had plausible standing and jurisdiction claims, it was their legal theories that were bogus.

A suit against WLPC in a US court wouldn't survive a motion to dismiss that anyone who speaks English could write. :-)