After all of curl’s noise about their deluge of slop reports and halting bounties because of it, I was surprised to see they would be so amenable to slopping curl itself. I suppose it’s yet another case of “all those other AI users are doing it wrong; because I understand the codebase and definitely totally use the bullshit machine responsibly then for sure I won’t produce slop”
On the other hand, that artist might have some sort of creative vision of their own that could irrevocably taint Yudkowsky’s Authorial Intent. Even worse, he would have to prominently credit their contributions, rather than attributing to Claude only the menial labour of piecing together his sacred vision.
I think the ‘trap’ might just be something akin to addiction? Once you’ve had your dose of the premium slop, you have to keep paying to get your fix. Except that completely fails to address why ‘people [shouldn’t] individually refuse to avail themselves’ of chatbots – it rather seems like a pretty good reason to stay away!
Right, deconstruction and lampshading and playing straight and exaggerating tropes are all themselves tropes – but our esteemed gentlemen is above tropes, so all he can do is glance vaguely in these various directions without actually going anywhere.
I was skimming over the script to try and patch some gaps that review’s excerpts glossed over, which also carries the benefit of avoiding all the remote analytics that seems to be included for some reason.
Naturally it’s 90% comments left by the chatbot describing all the fixes it had to make to its own errors. We can also see that Eliezer is still having some character recall obvious genre conventions of the setting, declaring “wouldn’t it be very rational to do this boring and direct thing instead?”, not getting a good answer in his own narrative, and then doing the genre convention anyway. Some real authorial growth here.
(For the otherwise confused, I do not advocate that people individually refuse to avail themselves of pre-superintelligent AI or AGI. Because AI is not a linear system where slightly less input results in slightly less damage; because the "now you have to pay us $20/month or fall behind" trap is real; and because I personally think it wise to learn and understand the frontiers in whatever ways I most easily find energy to invest there. I would give it all up in a heartbeat if there were an effective and collective treaty to stop it all at once. Don't be the "Yet you participate in the system. Curious!" meme.)
And yet you fail to name ‘Mr. Gotcha’. Curious!
I still haven’t managed to decipher what he means by ‘the "now you have to pay us $20/month or fall behind" trap’, or how that apparently justifies sending money to the organisations most at risk of destroying the world? But it sounds like submitting to a kind of blackmail.
If he just meant that not having access to chatbots put you at some sort of societal disadvantage, that isn’t a trap. It’s just something you would do. ‘Trap’ implies that, what, this is only a problem if you start slopping things? So how does that mean that you should?
Mozilla do not operate a pervasive network of internet monopolies that can relentlessly push for Firefox adoption and sabotage compatibility of other browsers. If Mozilla had any sense, they might dedicate some more budget to advocating antitrust and campaigning other counter-monopolistic policies, but instead… they want to make the corners rounder. And flip back and forth on supporting a compact layout, while deliberating over whether they need more AI or even more AI.
There are few parts I thought about sneering about in individual comments, but I think there are… enough here.
Still, one particularly galling part was this remark by the author:
Firefox’s new redesign is—at least to me—a welcome one. It no longer looks like a browser from 2015.
What on earth does a ‘browser from 2015’ look like? That was barely a decade ago! You mean an interface that still managed to function rather than whatever inscrutable nonsense the phone OSes are doing this week? Who wants software that has to be completely redesigned every five years?
So bizarre that other software outfits can’t work out how to put together a policy like this (although ‘substantially’ is doing a bit of heavy lifting)
Although the anchor in the link doesn’t point to anything on the main page; looks like you have to view at the README directly:
I see they’re still offering free copies “For US Kindle readers, while supplies last.” Perhaps they’re not that popular 🤔 (per this lesswrong, there were 1000 on offer)
edit: you know… if they were really serious about the doom and gloom, wouldn’t they just offer free ebooks to literally anybody? Would their site not just have a prominent link to a PDF of all the important things we need to know to navigate this tumultuous time? Beyond the core doomer audience, I’m not really sure who the target market is for “The survival of literally everybody is at stake and action is needed now, please give me $20 to learn how”
Well now I’m very confused. As far as I can tell, there’s nothing particularly special about ‘töten’, and its object is accusative. So why should it be dative in this… uh, case?
I think this perplexes us because we’re tying ourselves in knots trying to rephrase it that it might be deemed more ‘mature’. But that is missing the point: bringing up the concerns of the users at all is just a fundamentally immature remark.
We, the developers, are the adults in the room, and we want to slop it. So fuck off.
After all of curl’s noise about their deluge of slop reports and halting bounties because of it, I was surprised to see they would be so amenable to slopping curl itself. I suppose it’s yet another case of “all those other AI users are doing it wrong; because I understand the codebase and definitely totally use the bullshit machine responsibly then for sure I won’t produce slop”