It was so refreshing to see a video with a clear and direct title. In 2026 this would have been "We cracked open a motor. You won't believe what we found inside! <emoji>". Checking the channel, it seems they have succumbed to this for their latest videos.
Some here are arguing that mechanisms used by LLM providers already derail the goal of "the very best, most precise words at every single decision point", therefore the author is misguided.
The author has expressed a preference. Assume that there is a sequence of tokens, such that it is considered the absolute best by the author. This particular method of watermarking makes it less likely to generate that sequence, by definition.
I feel their argument would have been clearer and stronger if they had spent more time exploring the alternatives, and whether these alternatives would be just as effective. It is trivially easy to remove invisible tokens.
Like it or not, there is a public good to being able to identify AI generated content, and a small degredation in quality is tolerable in my opinion.
I don't think anybody has to worry about this issue though. Manual writing, coding, and proof reading continues to be an option. Where AI output is nothing to be ashamed of, the tools are available. For everyone else, there will be LLM providers that ignore EU law.
If the author has preferences on their "own writing" that conflict with Anthropic's, then they should actually write it themselves rather than paying Anthropic to do it. Private companies don't owe you anything, even less so when they're beholden to laws in foreign jurisdictions.
>Assume that there is a sequence of tokens, such that it is considered the absolute best by the author
You can't assume that because if that was the case he'd already know what sentence to write, because that's what that means.
The notion of a best sentence requires a final cause, an end to write to. By their very nature that's not how LLMs work, so you can't 'degrade' them on that front. They can't lose a property they didn't have.
I also suspect those who have decided to leave their employer will be more likely to create, maintain, and even keyword stuff the skills listed on their LinkedIn profile. This would be a poor proxy for real world skill level and their impact on the departed organisation.
Edit: Ah wait, I think I get what you mean. By "global news" you mean it's been deemed worthy of sharing to an international audience by media outlets, as opposed to a crowdsourcing news source like HN. Is that right?
I mean when you write 'UK researchers find', especially to a global audience (which UoKentucky PR dept may not particularly have but HN does (Hello from the (the!) UK)) that it sounds more like the country UK than the.. no offence intended but not exactly world renowned university. (I've heard of Kentucky, probably my first time hearing of its university in its own right - though I might have assumed it had one.)
I wouldn't read too much into it. UK is one of my alma maters. Everyone in that area of the US means "University of Kentucky" when they say "UK". It isn't a dig at the United Kingdom nor is it (I assume) an attempt to gain undue credibility by associating with the country. For the people there, UK as the University is simply the first order association for that acronym, rather than what is to them a faraway country that has no bearing on their day-to-day lives.
I don't think I am? I'm just agreeing the title should be different here, and it has now changed so either the submitted or a mod agreed.
> Everyone in that area of the US means "University of Kentucky" when they say "UK". It isn't a dig at the United Kingdom nor is it (I assume) an attempt to gain undue credibility by associating with the country. For the people there, UK as the University is simply the first order association for that acronym, rather than what is to them a faraway country that has no bearing on their day-to-day lives.
Right, that's what I meant about audience, and it making sense for the UoK's PR dept.
I don't feel like that is a correct way to measure an institutions mediocrity or not. I have found that frequently cited Universities that publish many very popular articles per year per journal become less credible with fame. Hubris aside objectivity becomes lost in the money being made.
A far better measure of a University would be the people it graduates or by assessing the actual effects and impact of a research University with their published studies.
If the University of Kentucky specializes in agricultural research, that's most likely why you've never heard of them - perhaps they've made huge contributions to the elimination of terrible fungus plaguing a specific crop, nobody would hear about that.
Fixing the proverbial fungus is a far bigger deal tho than most of the ambiguous pop science publications coming out of far more "prestigious" institutions these days.
a lot of southern schools publish in a different fashion. Most of my agricultural and husbandry knowledge of the area i live in comes from various universities .pdf files. My point is merely not publishing in Nature (or wherever) doesn't mean that serious research doesn't occur at that campus. Their "customers" may just be citizens of that state (or in some cases, that county).
Argument from authority is a pernicious fallacy, and typically effective too. You were right to call it out. I must admit I overlooked the sarcasm, however.
Yeah, it was kind of weird - pretty sure the positive comments about it weren't organic, but there are some other anti comments that didn't feel quite right either. I'm not even sure if the comment section works any more!
> In a practical world this is an outrageously useless platitude.
As an industry we went from treating engagement as an art form, to treating it as an optimisation problem. Along the way we got very, very good at it, at scale. While profitable, we now know that the methods used to drive this engagement are harmful. Many countries recognise this harm and have introduced legislation to limit or ban methods used on vulnerable groups, such as children.
It is interesting that we, as an industry, do not talk more about the harm we are complicit in causing. Somewhere along the way we normalised and accepted the idea that addicting features are desirable, and that we are not responsible for the consequences.
With all said, an industry is not an individual. You and I may care about this problem, but it is not clear how to fix it. At the very least, as individuals it would be good to avoid contributing to addicting features as a matter of principle wherever the opportunity arises, lest we become the equivalent of digital drug dealers.
While the analogy they offered may not be a good fit, I agree with the sentiment. My takeaway was not that addicts have zero agency or ability to help themselves, but that it's unethical to rely on this as the sole mechanism for minimising harm.
