Brb gonna ask Claude to overthrow the Standard Model and do a breakthrough
The Weekday
@professormusgrave.bsky.social
Professor of political science. Willing expat becoming an unwilling emigre. Essentially millennial, analytically millenarian. https://musgrave.substack.com
The thing is we can't check either way, the pile of tokens used is so enormous, it makes it nearly impossible to suss out how they arrived at this. And so, unlike with Buckmaster-Alpöge's work, we can't really any valuable insights from this result. Nor does it have any real practical. Implications.
I have the answer to Navier-Stokes ... but do you have the right _question_
Does anyone understand the solution or is it just 42?
A story about how LLMs are force-multipliers for highly skilled individuals, demonstrated in a high profile manner like this, should be a huge deal for OpenAI but that's not what they want to sell. They want to sell autonomy. They want to sell cutting out the human entirely.
If OpenAI solved Navier-Stokes in several days, albeit at the cost of a literal moonshot (one-way), then we really need to start speaking 🦜
Wait they just started training this
Gotta play on medium, otherwise you're just guessing. Save the hints for White people under 60. Education is your best heuristic for the men and marriage for the women. If you run out of marriage hints, you can use kids.
Had some fun trying to think about how to dramatize voters' choices in the 2024 election, so I pulled CES data and let you guess how respondents (validated voters) actually voted (or said they did) in 2024. profmusgrave.github.io/votergame2024/
ooof this is tough Who did I vote for? I earned 14,794 points over 20 voters on medium mode (avg 740). Model 894, base rate 760. profmusgrave.github.io/votergame2024/
Had some fun trying to think about how to dramatize voters' choices in the 2024 election, so I pulled CES data and let you guess how respondents (validated voters) actually voted (or said they did) in 2024. profmusgrave.github.io/votergame2024/
Does anyone have access to The Edinburgh Companion to Science Fiction and the Medical Humanities (2025)? Am in need of one chapter from it.
Good God. This really is the most surreally stereotypical Ugly American administration ever. It’s like if the French ambassador looked like this:
This is the putz who is getting called in, it won't go well for him.
You can't solve for problems you can't see. You can't see problems that you haven't measured. You can't measure problems without dedicated capacity.
The real upshot of this shitshow is that OpenAI can and will do ruthless corporate espionage against you and you should not let their tools see anything you’re not OK with Sam Altman personally possessing a full copy of.
everyone's focusing on the (seemingly likely) possibility that OpenAI scooped this proof via user data, but imo that buries the lead that they allegedly wanted to strike an author's name from the record because he works for a rival company deeply fucked if true
So where are people hosting personal websites if they want to run movable type and basic scripting but they don’t want to spend a lot of money etc
This one would’ve been a lot easier if I knew whether or not the voter drove a cybertruck.
Had some fun trying to think about how to dramatize voters' choices in the 2024 election, so I pulled CES data and let you guess how respondents (validated voters) actually voted (or said they did) in 2024. profmusgrave.github.io/votergame2024/
Who did I vote for? I earned 16,002 points over 20 voters on medium mode (avg 800). Model 786, base rate 768. profmusgrave.github.io/votergame2024/ If any polling firm wants my services, I charge 50 dollars an hour.
Who did I vote for?
profmusgrave.github.io
Now I understand why the best computer rating most Starfleet officers ever get is an A4
They never taught us the Washington signle handedly bumble fucked the entire 7 years war into existence at school.
“They never taught this in school” - person who didn’t pay attention in school This is also the average content creator online and their legion of viewers.
Oh, thank you for this, @professormusgrave.bsky.social. Who did I vote for? I earned 7,873 points over 10 voters on ultimate mode (avg 787). Model 821, base rate 679. profmusgrave.github.io/votergame2024/
Who did I vote for?
profmusgrave.github.io
Who did I vote for? I earned 15,543 points over 20 voters on medium mode (avg 777). Model 841, base rate 776. profmusgrave.github.io/votergame2024/ Neat project, thanks for making it.
Who did I vote for?
profmusgrave.github.io
Ultimate mode is a treat.
Had some fun trying to think about how to dramatize voters' choices in the 2024 election, so I pulled CES data and let you guess how respondents (validated voters) actually voted (or said they did) in 2024. profmusgrave.github.io/votergame2024/
Now you too can experience the plight of this chimpanzee yourself.
Had some fun trying to think about how to dramatize voters' choices in the 2024 election, so I pulled CES data and let you guess how respondents (validated voters) actually voted (or said they did) in 2024. profmusgrave.github.io/votergame2024/
There is a lot of variance within all these universities' AI recommendations, so have a look at who asked for them and what the committee was charged to do. Why I built the inventory, what it shows, and how to use it: kylesaunders.substack.com/p/what-are-u...
What Are Universities Doing About AI?
I gathered 96 AI task force report records at 74 institutions into a searchable inventory. Find your campus. Compare the recommendations.
kylesaunders.substack.com
Now with difficulty levels! @robfarley74.bsky.social I think this is definitely a minimum viable product
Had some fun trying to think about how to dramatize voters' choices in the 2024 election, so I pulled CES data and let you guess how respondents (validated voters) actually voted (or said they did) in 2024. profmusgrave.github.io/votergame2024/
Who did I vote for? I earned 7,575 points over 10 voters on ultimate mode (avg 758). Model 797, base rate 744. profmusgrave.github.io/votergame2024/