Adverb
@adverb.bsky.social
ML, Psychology, Art, Materials Informatics, Espresso, &c.
he/him
he/him
November 7, 2025 at 3:36 PM
Like it's hard for me not to take this as sort-of charmingly old-fashioned 😅
November 7, 2025 at 4:36 AM
Like it's hard for me not to take this as sort-of charmingly old-fashioned 😅
Guess I'll mention that there was this lol
Not sure how easy it would be to hotwire the site's chatbot like this but maybe it's not bad on the bot's side? Whole site is mostly boilerplate shopify.
Not sure how easy it would be to hotwire the site's chatbot like this but maybe it's not bad on the bot's side? Whole site is mostly boilerplate shopify.
November 7, 2025 at 4:24 AM
Guess I'll mention that there was this lol
Not sure how easy it would be to hotwire the site's chatbot like this but maybe it's not bad on the bot's side? Whole site is mostly boilerplate shopify.
Not sure how easy it would be to hotwire the site's chatbot like this but maybe it's not bad on the bot's side? Whole site is mostly boilerplate shopify.
OP is deleted, so I wonder how long it'll take for them to make fixes.
Contents of the article were also in error; it seems like they confused articles under review with surveys, reviews, position papers, leading to the "blanket ban" misreading.
Contents of the article were also in error; it seems like they confused articles under review with surveys, reviews, position papers, leading to the "blanket ban" misreading.
November 3, 2025 at 7:23 PM
OP is deleted, so I wonder how long it'll take for them to make fixes.
Contents of the article were also in error; it seems like they confused articles under review with surveys, reviews, position papers, leading to the "blanket ban" misreading.
Contents of the article were also in error; it seems like they confused articles under review with surveys, reviews, position papers, leading to the "blanket ban" misreading.
Incredible content is being made every day.
November 3, 2025 at 5:31 PM
Incredible content is being made every day.
Please consult someone from arxiv before writing something like this :/ low standards here
November 3, 2025 at 5:16 PM
Please consult someone from arxiv before writing something like this :/ low standards here
From their chatbot. For those who don't know, the model is clearly based on a facebook model using typical web data. Please please please use your brains. Their are smart people who are turning their brains off over this type of thing.
October 29, 2025 at 8:19 AM
From their chatbot. For those who don't know, the model is clearly based on a facebook model using typical web data. Please please please use your brains. Their are smart people who are turning their brains off over this type of thing.
I can understand reading someone and thinking "ah, they have interesting points on this subject" but before you base your worldview on theirs, look into whether you have any good reason to do that.
And look into whether they're consistently wrong. (Note the date!)
And look into whether they're consistently wrong. (Note the date!)
October 29, 2025 at 6:22 AM
I can understand reading someone and thinking "ah, they have interesting points on this subject" but before you base your worldview on theirs, look into whether you have any good reason to do that.
And look into whether they're consistently wrong. (Note the date!)
And look into whether they're consistently wrong. (Note the date!)
I'm not a credentialist; there are tech journos doing great work in this space!
But uh, when you see predictions that are routinely wrong and statements about the tech, it's just like... why in the world *would* anyone expect him to have insights here beyond doing literal PR for himself?
But uh, when you see predictions that are routinely wrong and statements about the tech, it's just like... why in the world *would* anyone expect him to have insights here beyond doing literal PR for himself?
![Bibliography
Zitron, Ed (2013). This Is How You Pitch: How To Kick Ass In Your First Years of PR.[6]
Zitron, Ed (2018). Fire Your Publicist: The PR and Publicity Secrets That Will Make You and Your Business Famous.[6]](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:p4rn5lfn7ymuxgerh6zk4mrt/bafkreihxeyunwf5omslnxhriif3omrt7tteq4vpfdyzir7xcmzi62zv3ae@jpeg)
October 29, 2025 at 5:47 AM
I'm not a credentialist; there are tech journos doing great work in this space!
But uh, when you see predictions that are routinely wrong and statements about the tech, it's just like... why in the world *would* anyone expect him to have insights here beyond doing literal PR for himself?
But uh, when you see predictions that are routinely wrong and statements about the tech, it's just like... why in the world *would* anyone expect him to have insights here beyond doing literal PR for himself?
(hopeful voice)
Is that good
Is that good
October 28, 2025 at 2:42 AM
(hopeful voice)
Is that good
Is that good
Whispering three letters that would ruin my and everyone else's day in the thread
October 27, 2025 at 2:45 PM
Whispering three letters that would ruin my and everyone else's day in the thread
If I had to make a guess it'd be they're just going to delete heuristics and use a typical recommender algo that they put under grok branding.
hypothetically could do some base model finetuning for sequential recommendation or whatever but like you said, no way they do that haha
hypothetically could do some base model finetuning for sequential recommendation or whatever but like you said, no way they do that haha
October 26, 2025 at 7:00 PM
If I had to make a guess it'd be they're just going to delete heuristics and use a typical recommender algo that they put under grok branding.
hypothetically could do some base model finetuning for sequential recommendation or whatever but like you said, no way they do that haha
hypothetically could do some base model finetuning for sequential recommendation or whatever but like you said, no way they do that haha
Actually bothering to look into it, he removed the dataset after one day. So I think you're confusing two people & uhhh the first one was polite and got death threats. This is the crux lol
October 26, 2025 at 6:14 PM
Actually bothering to look into it, he removed the dataset after one day. So I think you're confusing two people & uhhh the first one was polite and got death threats. This is the crux lol
I don't have actual thoughts on this but it always reminds me of that medical paper that accidentally reinvented the trapezoidal rule haha
October 26, 2025 at 5:50 PM
I don't have actual thoughts on this but it always reminds me of that medical paper that accidentally reinvented the trapezoidal rule haha
I made it to youtube
Any goog folks know if I can collect one (1) refurbish android phone worth of bounty? Lolll
Any goog folks know if I can collect one (1) refurbish android phone worth of bounty? Lolll
October 26, 2025 at 2:23 AM
I made it to youtube
Any goog folks know if I can collect one (1) refurbish android phone worth of bounty? Lolll
Any goog folks know if I can collect one (1) refurbish android phone worth of bounty? Lolll
Lol I'm so close to free
October 24, 2025 at 5:47 AM
Lol I'm so close to free
Lmao this is I think a for real security vulnerability? I'm supposed to be fully locked out of here
October 24, 2025 at 5:28 AM
Lmao this is I think a for real security vulnerability? I'm supposed to be fully locked out of here
People will see a dude who's smart and has been right before and just completely forget about uncertainy and human fallibility.
Cognito hazard for folks in the thread who don't want to know why he's followed haha
Cognito hazard for folks in the thread who don't want to know why he's followed haha
October 19, 2025 at 6:01 PM
People will see a dude who's smart and has been right before and just completely forget about uncertainy and human fallibility.
Cognito hazard for folks in the thread who don't want to know why he's followed haha
Cognito hazard for folks in the thread who don't want to know why he's followed haha
I don't think LLMs are more likely to be conscious than trees fwiw, but:
Closing sentence here is concretely not held by the majority of the field.
Otherwise we would be... done with studying LLMs, training dynamics, interpretability, etc. etc, etc.
Closing sentence here is concretely not held by the majority of the field.
Otherwise we would be... done with studying LLMs, training dynamics, interpretability, etc. etc, etc.
October 5, 2025 at 1:31 AM
I don't think LLMs are more likely to be conscious than trees fwiw, but:
Closing sentence here is concretely not held by the majority of the field.
Otherwise we would be... done with studying LLMs, training dynamics, interpretability, etc. etc, etc.
Closing sentence here is concretely not held by the majority of the field.
Otherwise we would be... done with studying LLMs, training dynamics, interpretability, etc. etc, etc.
Aaah is it the type of thing where you grow the little cells and reuse them like kefir/sourdough? Or different?
I've had this alarm for a few months btw. Best food.
I've had this alarm for a few months btw. Best food.
October 4, 2025 at 5:08 PM
Aaah is it the type of thing where you grow the little cells and reuse them like kefir/sourdough? Or different?
I've had this alarm for a few months btw. Best food.
I've had this alarm for a few months btw. Best food.
