synsypa.bsky.social
@synsypa.bsky.social
Reposted by synsypa.bsky.social
You brain is optimized to find meaning patterns, and a good way to find meaning in patterns is to find it when there isn't actually meaning nor a pattern.
February 2, 2026 at 6:47 PM
You brain is optimized to find meaning patterns, and a good way to find meaning in patterns is to find it when there isn't actually meaning nor a pattern.
Reposted by synsypa.bsky.social
/3 Anyway my view of Scott Alexander was always like my view of a lot of those self-styled rationalists: there’s nothing they wouldn’t support if you gave them enough words to work their way there.
January 19, 2026 at 3:46 AM
/3 Anyway my view of Scott Alexander was always like my view of a lot of those self-styled rationalists: there’s nothing they wouldn’t support if you gave them enough words to work their way there.
Reposted by synsypa.bsky.social
When someone points out "The only reason corporations vocally support human and civil rights is because the market rewards them and it helps their brand."
a man speaking into a microphone with the word good written on his face
Alt: Donald Glove: Good
media.tenor.com
January 11, 2026 at 8:12 PM
When someone points out "The only reason corporations vocally support human and civil rights is because the market rewards them and it helps their brand."
Reposted by synsypa.bsky.social
I blame Sam Altman. Other people too, but mostly Sam Altman
January 5, 2026 at 7:36 PM
I blame Sam Altman. Other people too, but mostly Sam Altman
Reposted by synsypa.bsky.social
You can't really plan a business model around "I hope we are beloved enough that people will willingly hand massive piles of Money without protest."
December 22, 2025 at 10:43 PM
You can't really plan a business model around "I hope we are beloved enough that people will willingly hand massive piles of Money without protest."
Reposted by synsypa.bsky.social
The problem with AI really does just seem like nobody wants to actually put in the hard work to utilize it properly. They expect auto-miracles in a box rather than just miracles in a box
December 21, 2025 at 9:00 PM
The problem with AI really does just seem like nobody wants to actually put in the hard work to utilize it properly. They expect auto-miracles in a box rather than just miracles in a box
Reposted by synsypa.bsky.social
It’s really amazing how ML, RPA, and “conventional” AI practitioners are having to thread the needle between the new business hotness and the reputational damage of hooking their wagon to that association
December 19, 2025 at 4:19 PM
It’s really amazing how ML, RPA, and “conventional” AI practitioners are having to thread the needle between the new business hotness and the reputational damage of hooking their wagon to that association
Reposted by synsypa.bsky.social
"I don't care to talk about it this way, this makes me morally superior to the rest of you, who I consider to be sniveling cowards" shut the fuck up, actually.
December 15, 2025 at 9:18 PM
"I don't care to talk about it this way, this makes me morally superior to the rest of you, who I consider to be sniveling cowards" shut the fuck up, actually.
Reposted by synsypa.bsky.social
"Attention is All You Need" (the paper that begat GPTs) was published in 2017. Bycroft's LLM data flow visualizers (bbycroft.net/llm) went live in 2021. Anthropic's Claude "map" was released in mid-2024
Not knowing "how the mind works" is irrelevant: ppl know how LLMs work
bsky.app/profile/cybr...
Not knowing "how the mind works" is irrelevant: ppl know how LLMs work
bsky.app/profile/cybr...
No one knows how the mind works, and this guy is no exception.
Obviously LLMs are (at least) the closest thing we have created to human intelligence. It passes the Turing test so to start talking about what it doesn’t do you’ll need better tests.
Obviously LLMs are (at least) the closest thing we have created to human intelligence. It passes the Turing test so to start talking about what it doesn’t do you’ll need better tests.
This crosses my mind every time I see a breathless article about LLMs "developing survival instincts" or "resisting being shut off" or "blackmailing their operator"
That's not how this works. It's not how any of it works.
That's not how this works. It's not how any of it works.
October 27, 2025 at 7:12 PM
"Attention is All You Need" (the paper that begat GPTs) was published in 2017. Bycroft's LLM data flow visualizers (bbycroft.net/llm) went live in 2021. Anthropic's Claude "map" was released in mid-2024
Not knowing "how the mind works" is irrelevant: ppl know how LLMs work
bsky.app/profile/cybr...
Not knowing "how the mind works" is irrelevant: ppl know how LLMs work
bsky.app/profile/cybr...
Reposted by synsypa.bsky.social
/7 So I question the proposition that just because Charlie Kirk didn't go around killing people, but went around belittling and demeaning and dehumanizing them instead, that he's inherently admirable. Would I be a hero if I said I will debate ANYONE about whether black people are human? No.
September 11, 2025 at 5:54 PM
/7 So I question the proposition that just because Charlie Kirk didn't go around killing people, but went around belittling and demeaning and dehumanizing them instead, that he's inherently admirable. Would I be a hero if I said I will debate ANYONE about whether black people are human? No.
Reposted by synsypa.bsky.social
/3 Trumpists and most of the new Right are shitty people with shitty values from a shitty culture. Responding to their violence with violence may not solve a damn thing. There are good arguments against it. But don’t let them pose for a millisecond as anything but the gleeful thug trash they are.
September 11, 2025 at 1:14 AM
/3 Trumpists and most of the new Right are shitty people with shitty values from a shitty culture. Responding to their violence with violence may not solve a damn thing. There are good arguments against it. But don’t let them pose for a millisecond as anything but the gleeful thug trash they are.
Reposted by synsypa.bsky.social
If you write stuff for a living, please never report the upper end of a range without also including the lower end. Doing that is a thing that I like to call “lying.”
May 9, 2025 at 5:14 PM
If you write stuff for a living, please never report the upper end of a range without also including the lower end. Doing that is a thing that I like to call “lying.”
Reposted by synsypa.bsky.social
Court: we take very seriously these grave findings by Congress and two successive presidents
Congress and two successive presidents: honk honk toot toot weeeee oh no!
Congress and two successive presidents: honk honk toot toot weeeee oh no!
January 17, 2025 at 2:42 AM
Court: we take very seriously these grave findings by Congress and two successive presidents
Congress and two successive presidents: honk honk toot toot weeeee oh no!
Congress and two successive presidents: honk honk toot toot weeeee oh no!