Krys Dolega
@krysdolega.bsky.social
Philosopher by education
Akademischer Rat @ruhr-uni-bochum.de by employment
Silly goose by calling
He\Him
www.krysdolega.xyz
Akademischer Rat @ruhr-uni-bochum.de by employment
Silly goose by calling
He\Him
www.krysdolega.xyz
Oh, I didn't mean trawling the internet boards for user generated content. I mean that they literally torrented all the possible copyrighted cultural texts to create the training corpus:
arstechnica.com/tech-policy/...
arstechnica.com/tech-policy/...
“Torrenting from a corporate laptop doesn’t feel right”: Meta emails unsealed
Meta’s alleged torrenting and seeding of pirated books complicates copyright case.
arstechnica.com
December 16, 2025 at 10:51 PM
Oh, I didn't mean trawling the internet boards for user generated content. I mean that they literally torrented all the possible copyrighted cultural texts to create the training corpus:
arstechnica.com/tech-policy/...
arstechnica.com/tech-policy/...
I mentioned two parallel development paths, because I don't think this would stop the advent of "microcomputers" and, eventually, the internet. After all, the ARPANET was started in the 60's and became operational by 1971
December 16, 2025 at 9:06 PM
I mentioned two parallel development paths, because I don't think this would stop the advent of "microcomputers" and, eventually, the internet. After all, the ARPANET was started in the 60's and became operational by 1971
My first thought was that we would probably end up with two completely parallel research programs using completely different architectures. As to data availability, the current industry leaders just stole copyrighted materials. I could easily imagine Cold War US gov just using public libraries etc.
December 16, 2025 at 9:05 PM
My first thought was that we would probably end up with two completely parallel research programs using completely different architectures. As to data availability, the current industry leaders just stole copyrighted materials. I could easily imagine Cold War US gov just using public libraries etc.
Sorry for all the mistakes, I saw and wrote this on the run
December 16, 2025 at 8:53 PM
Sorry for all the mistakes, I saw and wrote this on the run
Probably still terrible things, governments able to fund such research weren't much better in the past. But there is also a chance that these technologies would become a kind of shared service over time, just as telephone lines. Imagine time-sharing LLMs...
December 16, 2025 at 3:36 PM
Probably still terrible things, governments able to fund such research weren't much better in the past. But there is also a chance that these technologies would become a kind of shared service over time, just as telephone lines. Imagine time-sharing LLMs...
I imagine a world where some of the current breakthroughs happened way earlier and through completely different implementations. What would LLMs look like and what would they be used for if they weren't developed in the era of oligarchs of social media?
December 16, 2025 at 3:33 PM
I imagine a world where some of the current breakthroughs happened way earlier and through completely different implementations. What would LLMs look like and what would they be used for if they weren't developed in the era of oligarchs of social media?
Can you recommend some secondary literature on this?
December 15, 2025 at 8:38 PM
Can you recommend some secondary literature on this?
What paper\book is this in?
November 26, 2025 at 6:48 PM
What paper\book is this in?
More seriously, I expect that platforms like MTurk and Prolific are probably already scrambling to implement more and more sophisticated methods aimed at preventing widespread fraud through agentic models and AI-powered browsers
November 21, 2025 at 12:35 AM
More seriously, I expect that platforms like MTurk and Prolific are probably already scrambling to implement more and more sophisticated methods aimed at preventing widespread fraud through agentic models and AI-powered browsers
Sure, but:
- a lot of survey studies require relatively large samples (though, from what I've read, not everyone is concerned about statistical power)
- most people are just gonna recruit students, which is problematic since x-phi seems to be affected by demographics more than, say, psychophysics
- a lot of survey studies require relatively large samples (though, from what I've read, not everyone is concerned about statistical power)
- most people are just gonna recruit students, which is problematic since x-phi seems to be affected by demographics more than, say, psychophysics
November 21, 2025 at 12:30 AM
Sure, but:
- a lot of survey studies require relatively large samples (though, from what I've read, not everyone is concerned about statistical power)
- most people are just gonna recruit students, which is problematic since x-phi seems to be affected by demographics more than, say, psychophysics
- a lot of survey studies require relatively large samples (though, from what I've read, not everyone is concerned about statistical power)
- most people are just gonna recruit students, which is problematic since x-phi seems to be affected by demographics more than, say, psychophysics
Reposted by Krys Dolega
Happy partially muscled skeleton screaming by the perimeter fence day to all who celebrate. A solemn 30 second screaming before vanishing will take place at 11am.
November 14, 2025 at 9:47 AM
Happy partially muscled skeleton screaming by the perimeter fence day to all who celebrate. A solemn 30 second screaming before vanishing will take place at 11am.