This article is a rant disguised as data analysis.
I don't know how to word this in a non-confrontational, respectful way, but this article just feels like ammo for your next "debate with your anti-AI ennemies" where you get to say "look, I proved with data that those people had a disproportionate reaction and have double standards, therefore anyone who dislikes LLMs or their impact are the same!". Like, sorry, I know that sounds really reductive, but this really is the vibe I get when reading this and your other replies where you repeatedly talk about "showing the hypocrisy and double standards".
The global LLM discourse has grown massive, it spans trillions of dollars in promises and investments and affects pretty much everyone, so it's the easiest thing in the world for both sides to just find some people being assholes in the other camp and say "look, here's how [other camp] behaves".
The irrational, extreme, and heinous reactions are partly bandwagoning, and you can go about your day thinking that anyone who reacts like that is evil. But if you wanna dig a bit further, you'll notice that the entire media sphere has been screaming in everyone's ears for a few years now, that they're expandable, low-value-human-capital. All the money in the world (exaggerating a little) is being spent on making sure to remind anyone who opens a computer, opens a website, looks at a billboard, or turns on his tv... that their boss really really really wants to replace them.
Now you'll say that the friendly rsync contributor has nothing to do with any of this and... well yeah he doesn't. You don't need to agree with an emotional response to understand where it's coming from, and even if you're still dead set on considering them "the enemy", then understanding why the anti-AI crowd reacts like that is STILL a positive for you.
Why is the guy being rigorous worthy of criticism, but the guys being idiots aren't? Did you post any similar calm-down comments in either of the HN threads on the original attacks?
I am more inclined to be critical of AI boosters, so what? Am I supposed to crumble under the weight of immense cognitive dissonance because I have... a stance in the discourse?
These guys on the github thread aren't my friends, I have no concern for them embarrassing themselves or leaving a bad digital footprint by drawing ms paint gore. I also have no concern for OP, but it just so happened to be the post I found, and I just so happened to be in the mood to leave a comment.
Engaging in LLM discourse is already a waste of my time, I'm not going to waste more of it just to avoid fallacious accusations of double standards because I didn't "do the same for the other side".
Strange thing to say. The post does a good job of showing that there is no evidence Claude had anything to do with the regressions.
Your problem is that this was shown? You don't value epistemics -- you care about the ideology more than truth. Even if you don't like AI you should still do it in the right way.
Your comment comes across as more unself aware and more destructive. Lets keep this place truth first and ideology second.
Yeah yeah the usual "look, these anti-ai people are so EMOTIONAL and HYSTERICAL while we're very logical and fact-based", I've lurked for a while so I read that one plenty of times.
OP spends a lot of time time doing statistics, but when another person replies "hey, 2 claude-authored features is not really statistically significant", the author literally agrees and says "my point was to show that you can't draw conclusions". Direct quote:
> I'm only trying to show there's no evidence for the anti-AI hypothesis
---
And here's another thing. How exactly is it self-unaware to say "Hey, I get your frustration with people being assholes, I'm not excusing it, but it can never hurt to understand why some of them have become so extreme in their vitriol, here's a few reasons for their feelings".
Is "feelings" a curse word or something? What's so wrong with understanding the emotional component of the AI discourse?
Talking about "emotions" is not destructive when the topic at hand is literally people being driven by emotion under a github thread.
What the article is saying is "these people are acting irrational because they're evil and the enemy, so here's how I prove them wrong with statistics!", and my comment to the article was "hey, you seem to go with the assumption that this is all based on pure evil, here are a few reasons why people might get tired, and then angry, about this whole thing".
You quite literally exemplify my point when I said that the analysis is mostly just >ammo for your next "debate with your anti-AI ennemies", it's a tool that allows you to not engage and dismiss any argument as "not on the side of truth because not on the side of the numbers".
All of this even though, and I need to state this again, I never once rejected the analysis or the results that OP came to, all I did was point out that OP is also engaging in "us vs them" think with the occasional "wink wink, CLASSIC AI hater amirite?" sprinkled in the article.
If someone spouts untruths, that is a failing of theirs. Talking about how there's emotions of theirs that back this promulgation of untruths, emotions which you could understand if you tried, is putting the cart before the horse. First express things that are not trivially false, then I will try to understand what you mean by it.
What in the AI discourse is trivially false? On one side we have almost all the money in the world, owners of giant social media companies, with easy access to media, who buy ads, etc. to tell you that half of us will be out of jobs in the next 6 months, that AI superinteligence is coming next year, that singularity is upon us, that Claude Mythos will cause so much harm it has to be contained, etc. etc.
I consider all of the above trivially false untruths. But on the other hand you have rsync users who are fed up with all the lies, all the propaganda, all the fear mongering brought up by literally the richest people the world has ever seen, who have experienced AI slop first hand, who have been tricked by AI music, are fed up by all the AI generated posters at their local coffee stand, have given up on trying to correct their coworkers AI generated code during peer review. And now these rsync users see AI slop has been pushed to their favorite tools, and see a new bugs popping up at the same time.
I consider the latter a completely understandable reaction. Yes they jumped to conclusion, but in doing so they have evidence. They have both anecdotal and circumstantial evidence for their conclusions. In comparison, the AI singularity people have nothing but vibes and science fiction behind their conclusions.
This article merely showed that one particular sample (two releases of rsync) or not statistically significantly different (p > 0.05) on one particular metric (bugs per commit). All that says is that you cannot generalize group difference over the population using that sample.
This still leaves the anecdotal evidence. And anecdotal evidence is still evidence, and in the absence of better evidence, it is perfectly rational to react based on the evidence you do have.
There are folks incentivized by AI: engineers and managers working for AI related companies who justify their beliefs with selective facts. And then there are engineers who are threatened by AI and are extra-sensitive to slop.
The article is missing the point that once camps have made up their minds - no amount of analysis is going to change that.
Why it is probably and regrettably true that few people in camps will change their mind, data analysis can help people who haven't been captured yet to either stay away from the camps or at least fall into the "more correct" one.
I wholeheartedly believe it's 100% intentional of Anthropic to use "vulnerability" to describe something that ranges from "serious attack vector" to "you forgot to add this variable to the useEffect dependency array".
The difference between serving 1 and 1 billion http queries is not the same as the difference between generating 1 and 1 billion tokens.
The startup blitz-scaling-market-capturing playbook makes makes sense when you spend to scale, not when you spend because you scale, yeah, I understand that step 2 is "and now you squeeze the users", but it will need to be by such a bigger factor...
But if it's fully being amortized, then it means they don't buy new Nvidia GPUs anymore for a while. The situation is either "your GPU AND the datacenter infrastructure it's running on is obsolete", or "Nvidia's profits tank because people are staying with current-level infrastructure".
That would be true if everyone weren't supply constrained and buying literally everything they can find.
There are actual risks that this trend doesn't continue, but as long as the trend continues, it is pretty good for revenue. "AI shown to hit a wall/doesn't actually deliver/stops growing so fast", "massive improvement in hw efficiency or tech such that all the old stuff becomes obsolete", "bottleneck on power/regulations/etc such that no one wants anything but the most efficient cutting edge stuff" would be the ways it could end and then all these factors reverse. Right now, power is so constrained that old, inefficient power generation is actively being turned back on or set up at new sites (e.g. old aviation turbines which are very inefficient compared to combined cycle).
Good article, I think most hero images are pointless in the first place, so having them diffusion-generated now feels like a signal to just not read the piece. It goes from "look at this irrelevant image" to "look at this irrelevant image, and I have bad taste by the way!"
I'm reading "people don't care so it doesn't matter" in the replies, in that case can we agree to just drop all unneeded illustrations altogether when it comes to technical articles?
But does this translate as "one year of cumulative work" or rather "one year of rearranging your workflow and discarding obsolete ideas"?
If you spend a year walking in circles, someone can easily close the gap with one step. Especially if models and harnesses are supposedly getting more powerful all the time.
Not to be a hater or anything (I'm a hater), but I'm seeing people mention the potential of LLMs for the "grunt work" like retopo, but I can't really begin to imagine what the "correct" data representation and python api calls would even begin to look like in a training set?
Would an LLM really be querying vertices in relation to one another and estimate whether their distance "sounds" like good topology?
Fusion has great potential, but is probably being held back by a lack of community support, the slightly-higher-barrier-of-entry of node-based workflows, and the subtle but annoying ways in which the software can work against you.
TLDR: it does some stuff slower than ae, but nodes allows it to very easily do a lot of stuff that ae struggles with.
It's also a lot easier to parse since node->properties is less nested than comp->layers->effects->properties (and this makes a big difference on cognitive load).
I would also recommend Control Panel for Twitter, easy to install on PC, and you can also have it for your phone if you use twitter via the Firefox mobile browser.
If all you can see is the following tab, then any ragebait that gets in your way is much more actionable, simply unfollow or mute whoever got it on your feed.
Violence against datacenters or AI company CEOs is very bad, they must be allowed to fail organically so that they don't have any excuse.
The last thing I want is for someone, in 2029, to say "but LLMs just weren't given a fair chance last time, we would have definitely reached AGI with more funding if it wasn't for [targeted attack]"
I don't know how to word this in a non-confrontational, respectful way, but this article just feels like ammo for your next "debate with your anti-AI ennemies" where you get to say "look, I proved with data that those people had a disproportionate reaction and have double standards, therefore anyone who dislikes LLMs or their impact are the same!". Like, sorry, I know that sounds really reductive, but this really is the vibe I get when reading this and your other replies where you repeatedly talk about "showing the hypocrisy and double standards".
The global LLM discourse has grown massive, it spans trillions of dollars in promises and investments and affects pretty much everyone, so it's the easiest thing in the world for both sides to just find some people being assholes in the other camp and say "look, here's how [other camp] behaves".
The irrational, extreme, and heinous reactions are partly bandwagoning, and you can go about your day thinking that anyone who reacts like that is evil. But if you wanna dig a bit further, you'll notice that the entire media sphere has been screaming in everyone's ears for a few years now, that they're expandable, low-value-human-capital. All the money in the world (exaggerating a little) is being spent on making sure to remind anyone who opens a computer, opens a website, looks at a billboard, or turns on his tv... that their boss really really really wants to replace them.
Now you'll say that the friendly rsync contributor has nothing to do with any of this and... well yeah he doesn't. You don't need to agree with an emotional response to understand where it's coming from, and even if you're still dead set on considering them "the enemy", then understanding why the anti-AI crowd reacts like that is STILL a positive for you.
reply