Would be quite hilarious if the first two companies to buy As space on ChatGPT is Anthropic and Google.
Specially hilarious, since that's exactly how TikTok got all their initial traction from Meta Ads.
Even if the prompts are technically leaked to the provider, how would they be identified as something worth optimizing for out of the millions of other prompts received?
This kind-of sort-of already happens now with Big Model / AI release. Rest of the world already gets features & model drops much much before the EU does.
I am honestly interested in any examples you might have, cause I do spend a bit of effort keeping up with whether LLM releases are delayed or lack certain features in EU member states specifically and know no recent example of that.
GPT-5, Gemini 2.5 Pro and Claude 4 were released across the same time frame (staggered releases that affect users even within regions not withstanding) for EU and US customers.
ChatGPT Agents meanwhile had a three week delay, but that was not EU specific and affected other countries such as Switzerland as well. Previously, I have also seen the very much not EU UK included in such delayed releases.
Essentially, all recent LLM releases I am aware off either dropped simultaneously for EU and US customers or, if they were inaccessible within the EU early on, that generally included none-EU countries with different or no applicable regulation as well. Any example of differences in accessible features I know of hasn't been limited to EU member states.
Just checked, the releases of Dall-E 3, GPTImageGen, Google Veo 2, Imagen 3 and 4, took place simultaneously for EU and US as part of global launches.
Sora was the only outlier here, though as always, the restrictions did not encompass just the EU and were lifted shortly there after, just like with ChatGPT Agents:
> Right now, users can access Sora everywhere ChatGPT is available, with the exception of the United Kingdom, Switzerland and the European Economic Area. We are working to expand access further in the coming months.
What's funny about this comment is that YouTube is absolutely usable if you pay for YouTube Premium, but most won't.
Goes to show, while everyone whines about how bad an ad based business model is, noone has any alternate strategy that works at scale, - especially one that's not exclusionary to people from the developing world who can't afford to pay for every single thing on the internet.
It's crazy that people will watch hours of YouTube content every day, streaming HD video with almost zero buffering, features like auto-generated subtitles that can then be automatically translated into other languages, essentially unlimited storage so the videos they like are around forever (copyright claims and the like are a separate issue), you can embed the videos onto your own site for free, you can use them to livestream... all this and you'll never see a single advert if you pay for YouTube Premium. It's fantastic, but most people are so entitled that they seem offended that YouTube would ask for money to provide this service. I know premium does nothing about the insane amount of sponsorship slop every creator stuffs into their videos now, but maybe if people stopped using every possible tool they can find to block YouTube adverts and paid for premium, YouTube would pay their creators more and they wouldn't need to constantly lie about drinking AG1 every day and the quality of raycon headphones.
Pretty sad that the rest of the world needs to pay for the extra tokens because of non-scientific american bias.
This is also possibly a big point why countries/regions want sovereign LLMs which will propagate regional biases only.
I always like to ask these models who invented the airplanes, because a few countries have their own inventor... So in my opinion, it's a good way to check.
Very good. If the LLM has to express an opinion, it will have to be its own opinion (after the implementation of intelligence and judgement) - otherwise, it has to explicit the foundations of its statements (certainly not be the "hearsay machine" we have seen).
It's not a matter of «extra tokens»: it's that the fact, the "summary after the protocols", is what I wrote. It is the correct answer. It's what you should expect from a lucid speaker.
No. That disclaimer implies that there are other likely answers. The age of the earth is completely settled, and has been for a long time. Facts don't care about your feelings.
You misunderstand it completely, as it is not a matter of feelings. And it is not a disclaimer (which you apparently felt as a disclaimer).
It is a matter of facts. The facts are, that that computation was performed by Patterson and refined by others. This is, as said, what a good reasoner will tell you.
> implies that there
Even if there had never been other attempts to answer that question, the "facts"¹ remains as stated: Patterson computed, followers refined. Without those specifications, the machine will be a "dumb believer" - a "minor". We will not ask for the machine's opinion until it will be intelligent. And when it will be intelligent, it will speak as I said.
> completely settled
Proper science does not work the way you seem to think it work.
--
¹(And I mean "facts" the way I used it, not the way you used it. I meant "facts recorded as objective" - you meant "information you accepted to believe", which is of course very far from facts and may happen to be adherent to the state of things only by coincidence.)
It is not just “according to some research”, it is also according to the overwhelming scientific consensus at the time. Sources are good but it should not appear as if it is one opinion among possibly many others equally valid.
But it does not matter: the «overwhelming scientific consensus» will be the reason why it will be the chosen reply by the machine, but to specify in the reply "According to Patterson, followers and overwhelming scientific consensus" would be a redundancy.
The appearance that it could be «one opinion among possibly many others equally valid» is all in your head: it is an unduly feeling from a bad mental framework.
The advanced framework (that I advanced) is that of the foundational theory of knowledge: a notion has a source - you computed or reasoned, or somebody else. You do not allow your consultant to believe, so you demand that knowledge is tracked.
You will not accept an oracle.
The paradox is that you are seeing the demand of the source as a support to "belief", while it is the radical opposite: the only thing it will be """believed""" (and not really "believed" - just the end of the chain) is the protocols, that "in the training sources I read statement S".
