> Read what follows as a map: where open AI is winning — some numbers surprised even us — and where it is exposed. A case that hides its weak points is an advertisement. In the six weeks since our first draft, a major US lab returned to open weights and Washington debated a ban and declined it. The map moved while we drew it.
Some interesting numbers throughout, but I'd love to read the presentation without the obvious Claudeisms.
Agreed; not writing your own prose is lazy and disrespectful. If you expect/hope that humans will read something, you should demonstrate that it is important to you by spending your own time on it, and you should respect their time by being as clear and eloquent as possible (which Claude et al. are decidedly not, for now anyway).
ctrl+f for "distill" is pretty enlightening as to the bent of this publication, for better or worse. They do cover the idea to some extent, but mostly to dismiss threats from DC.
If the creator wanders through here: this is a fantastic presentation from a dataviz perspective, but next time it'd very much be worth the money to pay someone to rewrite 100% of the copy. The Claudisms are just impossible to ignore, and makes you constantly wonder "is this sentence hallucinated, or real?"
I originall clicked this to see discussion of safety, but I guess I'm not sure what I was expecting. Not sure what to say, either. I guess I just hope the mass casualty event that changes everyone's minds on this topic is in the thousands or tens of thousands instead of the millions or billions :(
It would have been nice if someone put some thought and effort into this, but it's just classic AI-generated slides designed to push the author's point.
I scrolled to the slide titled "Eight of the top ten models by token volume are open weights" and if you're familiar with the space, you can probably already guess the mistake it made: They put up the chart of OpenRouter's token volume to show the open models ahead of the closed models. If you don't know, most traffic to frontier labs doesn't go through OpenRouter, so any chart showing OpenRouter stats does not capture the actual market. They (or their AI) added a little footnote box about this, but that stats are actually useless for judging the entire market.
It also cherry picks stats and benchmarks all over the place. It tries to make useless stats like saying the best open model only lags behind frontier by "4 points" into a huge accomplishment.
7 comments
[ 0.18 ms ] story [ 9.2 ms ] threadhttps://github.com/orgs/openapi/discussions/categories/open-...
Some interesting numbers throughout, but I'd love to read the presentation without the obvious Claudeisms.
If the creator wanders through here: this is a fantastic presentation from a dataviz perspective, but next time it'd very much be worth the money to pay someone to rewrite 100% of the copy. The Claudisms are just impossible to ignore, and makes you constantly wonder "is this sentence hallucinated, or real?"
I originall clicked this to see discussion of safety, but I guess I'm not sure what I was expecting. Not sure what to say, either. I guess I just hope the mass casualty event that changes everyone's minds on this topic is in the thousands or tens of thousands instead of the millions or billions :(
I scrolled to the slide titled "Eight of the top ten models by token volume are open weights" and if you're familiar with the space, you can probably already guess the mistake it made: They put up the chart of OpenRouter's token volume to show the open models ahead of the closed models. If you don't know, most traffic to frontier labs doesn't go through OpenRouter, so any chart showing OpenRouter stats does not capture the actual market. They (or their AI) added a little footnote box about this, but that stats are actually useless for judging the entire market.
It also cherry picks stats and benchmarks all over the place. It tries to make useless stats like saying the best open model only lags behind frontier by "4 points" into a huge accomplishment.
Don't waste your time.