And also for the sake of completeness, I don’t have unlimited deference to the professions. As an AMA member: AMA is largely a lobbying body for the economic interests of well-off physicians. But it *also* has many people who genuinely care about our professional ethics, and that is a real check.
I know this does nothing because we’re like post-search and maybe post-web now, but I would say I think it’s important to hold on to the notion that we did a fairly sh*t job regulating those, too lol
I guess my position boils down to:
Someone on a thread mentioned ~Hey what happens when someone says “I think I’m trans,” silence? My response is, Actually that shouldn’t be up to AI companies—look at OpenAI’s govt relations, you think that’ll be objective? Professions give some insulation there.
I can definitely see trends toward anti-AI regulatory being more of posturing than actual genuine reckoning with the new environment afforded by these systems 😔 +
For sure, though, there is very good reason to have a certain degree of regulatory introspection over these interactions, in the same way that we have regulatory scrutiny if your psychic performs heart surgery. I mean that isn’t censoring your psychic, it’s just assessing risk and benefit.
I think the “good outcome” would be a version of this, where interactions that fall within licensed domains must conform to professional standards for scope of practice for AI systems in that profession. Then professions ideally would assess the evidence and delineate what can be done autonomously.+
It’s a tension, right—right now, it is very bad that it’s pretty much the Wild West out there for AI services to provide medical advice etc. with ~zero oversight as to quality. That also harms people. But so does not recognizing that there is a genuine, if different, role for AI in these areas. +
Christ almighty what’s five months from now
The person this is for knows it’s for them :)
youtu.be/JUl4kxTfzKE?...
we must go deeper.
Radar Detector Detectors: Everything You Need to Know By Vortex © November 8, 2017 & 10 mins to read In areas where radar detectors are illegal, police commonly use radar detector detectors (RDD's) to locate drivers who are running radar detectors. How do they work? How easily can they spot radar detector users? How effective are "undetectable" radar detectors at avoiding detection from a radar detector detector? Let's take a look and cover everything you need to know about radar detector detectors. Photograph of a radar detector detector, a somewhat honestly equipment box suctioned to a window, on a car dashboard
Event description from a community-building gathering at a private equity knock-off WeWork: In Full Bloom: Celebrating Women March 3, 11:00 AM - 2:00 PM Mindspace Market Street 4th Floor W RSVP # Add to Calendar Join us in celebrating Women's History Month as we honor the strength, resilience, and creativity of women past and present at our bouquet workshop. Create your own beautiful floral arrangement while reflecting on themes of growth and empowerment that flowers naturally represent. Everyone is welcomed!
… oh, the dissertations that could be written about all of the concentric layers of semiotic offense in this one event advertisement.
Claude Code screenshot showing "Let's go." next to a white action dot.
god dammit.
the free market speaks volumes.
As I’ve said for a long time:
1. These things on code are essentially local k-NN on the internet.
2. The benchmarks are fake PR gadgets and don’t mean anything at this point.
3. That’s fine, local k-NN on the Internet for code isn’t Jesus but it’s a hell of a tool.
wait aren’t these the “you can’t tell me to make a gay cake” people
Screenshot of Bluesky discover feed Hmm, some kind of issue occurred when contacting the feed server. Please let the feed owner know about this issue. Message from server: NotEnoughResources View profile
not enough minerals
you heard of webassembly? well how about
interesting asymmetry between this and the “build apparatus to rip apart every book known to man to train our models” energy.
sounds like a skill issue, Anthropic.
Whoa, this is such a cool result!!
Thank you for posting this—it’s a good reminder, and it’s very easy to get swept up in the thumos and stop actually evaluating the structural scenario that drives our purpose, what we actually all should be here to do. I find myself unproductively swept up in it sometimes too.
Claude Code + Railway = I dunno kind of not really
Screenshot from Claude Code in orange text: "Surveying nonlinear Perron-Frobenius theory..."
uh, ok that escalated quickly
… and that’s why the goal for AI should be semiosis, rather than benchmarks comprehensible to humans. This paper’s title has the obvious implication, and my conjecture is that this lies as a limitation in the notion of “benchmark” itself.
oof who’s gonna tell her about “on formally undecidable propositions of principia mathematica and related systems”
What these all get at is epistemic dynamism, or what was at the heart of the original analytic sense of "generativity".
The good AI is one that does not simply recapitulate the form of extant structures, but participates in the constant, roiling recombination necessary for continuous reinvention.
I have thought a lot about the state of AI, what we might build in the future, and the worlds that result. Through this, I've come to three core aspects of a future "good AI".
The "good AI":
1. is queer;
2. can undergo semiosis; and,
3. understands prajnaparamita—and so walks the Bodhisattva path.
just gonna go ahead and tick that one off the FALGSC checklist
might I offer Ed a suggestion
