I thought this was the weakest argument in the article. Yes, non-determinism is inherent to LLMs. It's also inherent to human brains, yet somehow we still manage to call (some) people reliable.
The bit about backwards compatibility doesn't make sense either. LLMs, just like humans, can take backwards compatibility into consideration — or not. Sometimes they'll forget, and sometimes they'll fixate on it even in a system that hasn't been deployed yet. It has nothing to do with LLMs never doing "the same thing the same way twice".
LLMs may or may not become as reliable as other productivity-improving infrastructure, but it's not the nature of their technology that will decide this.
By and large we do. Unfortunately, the losers don't care unless you identify them personally. For them, the thrill of cheating and griefing others easily overcomes some generalized cultural zeitgeist.
(Thanks for the well-reasoned reply! I'll preface mine by saying that I'm not terribly well-versed in the GDPR either... and that's really part of the problem.)
AFAICT, the GDPR consider an IP address to be PII. So collecting visitor logs or sticking an analytics script on the page -- even without asking people to sign in or otherwise identify themselves -- would fall under the purview of the GDPR. I don't think doing such things should cross the line into "active character".
It's also a terrible idea to infer intent from offering translations in other languages. As another comment pointed out (I presume correctly) there are more German speakers outside Germany than there are within its borders. Why should trying to improve the accessibility of a web site by people around the world make one susceptible to the laws of a language's native country?
As you say, the implied end state is the real problem: do I now need to learn about the relevant laws of every country whose residents (or even citizens?) might visit my web site? That's regulation without representation, and should be resisted on principle IMHO. Or will every site need a standard disclaimer that it's only intended for residents of X?
Perhaps some creative lawyer will figure out how to add a clause to Terms of Use that shifts responsibility for damages incurred by exposure to extraterritorial laws to the offending (offended?) user, and we can have a detente through mutually assured destruction of dueling lawsuits. :)
And just to preempt any criticism from the European peanut gallery, I'm in favor of stronger privacy protections on the Internet but they need to be worked out and agreed on globally, not imposed unilaterally by claiming extraterritorial jurisdiction.
> This is distinct from, say, a company explicitly providing a service to customers in a particular country (cf GDPR).
In what ways do you think the GDPR is distinct from the laws in the present case? Let's say, for argument's sake, that gutenberg.org starts capturing PII about visitors and storing it without conforming to the GDPR. Do you think it would be fair for the EU to sue them and (attempt to) enforce their law? If so, why?
The bit about backwards compatibility doesn't make sense either. LLMs, just like humans, can take backwards compatibility into consideration — or not. Sometimes they'll forget, and sometimes they'll fixate on it even in a system that hasn't been deployed yet. It has nothing to do with LLMs never doing "the same thing the same way twice".
LLMs may or may not become as reliable as other productivity-improving infrastructure, but it's not the nature of their technology that will decide this.