Paul Rowe
Paul Rowe
@pauldrowe.bsky.social
Applied logician. Polyglot. Interested in curiosity, not dogma.
697 followers223 following439 posts
This does seem like an important step in the right direction. I think it creates many new issues around managing the proofs (especially given the scope and scale). But that is a good thing to worry about. What would a Mathlib analog really look like here, and how could we leverage it effectively?
Formal methods becoming orders of magnitude easier to do is a huge deal for software security.
Paul Rowe reposted
Paul Rowe reposted
Paul Rowe reposted
This continues to be well outside my expertise, but an article by @kristinandrews.bsky.social brought my attention to the key analogy between (unknown and disputed) animal consciousness and AI consciousness.
"There's no teapot floating in the asteroid belt." That's a strong claim which implies a lot of things about teapots and orbital mechanics. It requires strong evidence.
This continues to be well outside my expertise, but an article by @kristinandrews.bsky.social brought my attention to the key analogy between (unknown and disputed) animal consciousness and AI consciousness.
"There's no teapot floating in the asteroid belt." That's a strong claim which implies a lot of things about teapots and orbital mechanics. It requires strong evidence.
Paul Rowe reposted
"There's no teapot floating in the asteroid belt." That's a strong claim which implies a lot of things about teapots and orbital mechanics. It requires strong evidence.
Again: “AIs aren’t conscious” is a strong, affirmative claim. It implies a lot of specific things about consciousness. It requires strong evidence. “We can’t say for sure if AI is conscious” is a weaker, cautious claim. It doesn’t make any particular evidence about consciousness.
My most charitable interpretation of what this could be* is to guess that they want to continue to train models on user inputs and don’t want to add abusive content to their training. But also don’t want to put actual guards in place to check for content. *not necessarily of what I think k this is
Anthropic's new User Policy forbids you from engaging in "sustained and needless abusive or cruel behavior toward our models". It does not, however, explain what the fuck this could possibly mean.
Paul Rowe reposted
Paul Rowe reposted
For anybody interested in philosophy and its history, this is podcast is for you. If podcasts aren’t your thing, he has been publishing book versions of the series. They are wonderful overviews and intros to philosophical texts. And the no gaps approach is real but not unreasonable.
As of today the original #HoPWAG series has reached 500 episodes! Thanks to all who have listened and the many who have supported the series over the years, not least all the interview guests. #philsky #podcast #philosophy
Paul Rowe reposted
👀 Florida attorney general to OpenAI: (www.myfloridalegal.com/sites/defaul...)
At 45k followers I will do the same. (Probably* just repost @lastpositivist.bsky.social ‘s post). * This is not probable. I will never reach this many followers.
At 45k followers I will reveal exactly what terminology correctly carves AI at its joints.
Paul Rowe reposted
Paul Rowe reposted
I often feel conflicted by art from the 1920s. This is no exception. I like the forward facing eye projected from the (sideways) figure. The angles evoke Cezanne (good), but overall effect is somehow … I don’t know. If this speaks to you, I’d love to hear why and learn.
Man in a Tower by André Masson, 1924 https://botfrens.com/collections/212/contents/137779
Paul Rowe reposted
this is perhaps the most useful thing that LLMs and GenAI have done: to provoke people into articulating (over again) that even the 'hardest' of 'hard sciences' are human endeavours, with human meaning, and concerned with human understanding of the significance of the world we all inhabit
“Mathematics only makes sense as a gift economy. It has no trade secrets. Substituting a monetizable ersatz culture for the one evolved through historical experience is a particularly insidious form of knowledge collapse.” Mathematician Michael Harris on the AI companies coming for math:
Paul Rowe reposted
With the start of the semester just a week away, here’s what I decided to say to my upper division math course about how AI will or won’t change what they’re learning… francissu.substack.com/p/the-enduri...
Paul Rowe reposted
Paul Rowe reposted
You know those viral fb posts that reduce to order of operations quibbling? Please deploy this associahedron-based interactive viewer in order to subvert those threads with interesting mathematics: kyleormsby.github.io/parse-explor... 🧮
Resolving ambiguity in order of operations is easy: simply enumerate all possible parsings of a given expression, compute each value, and then average. E.g., for 9-3-3÷3, ((9−3)−3)÷3 = 1 (9−(3−3))÷3 = 3 (9−3)−(3÷3) = 5 9−(3−(3÷3)) = 7 9−((3−3)÷3) = 9 so 9-3-3÷3 = (1+3+5+7+9)/5 = 5. Voilà! 🧮
Paul Rowe reposted
Paul Rowe reposted
No professional accomplishment has ever made me happier than this. theonion.com/translator-o...
Paul Rowe reposted
In case you were wondering WHY someone would mail their scabs, the earliest inoculation available for smallpox involved grinding up the scabs and inoculating someone with a scratch on the skin. This method reduced the fatality rate from 30% to about 1%. 🧪 en.wikipedia.org/wiki/Variola...
Historian friend was reading a civil war era letter in an archive, the letter said “In this envelope I include my smallpox scabs” and sure enough there they were. Had to call CDC for a hazmat team.