I hate that I also had this thought earlier tonight.
I hate that I also had this thought earlier tonight.
If it was the Gay of Hormuz it'd still be open
I think this is about the temperature but honestly it works for decades as well
This combo is...okay, in my experience. I'd rather use higher and lower Claude models for each (e.g. Opus for planning, Sonnet for simple execution) but tokens are expensive.
I could maybe see an argument that threads these two by pointing out that none of this shit affects cis people's lives materially and it's weird to care so much about it, but ultimately I'd really just like cis people to see as worthy of dignity and human rights. Why is that so much to ask for?
"You talk about trans issues by saying they are a distraction and also pointing out there are so few of us that it isn't worthwhile dedicating time to persecuting us" is not the fuckin woke win that I see people portraying it as and it sucks
This is a point I wish more LLM boosters would consider. Tools can be amazing for some tasks and terrible for others! That's how most tools are! People want LLMs to be some kind of magic wand and the truth is undoubtedly more nuanced than that.
Also, telling people that my skepticism was started because of an Email forward makes me feel old as hell. It feels like I'm admitting to reading urban legend on mimeographed zines. Time is a bitch.
My experience has been better than that but I agree with you that people are just treating it as a shortcut. Probably a fair number of them are just writing tests to the implementation instead of establishing well-conceived tests and then implementing and reviewing the code. That way lies danger.
Also, I work in software development, where the general vibe around LLMs has become "Adapt or die." People working in other sectors probably have more to fear than me.
I still fear the inevitable bubble bursting, which will impact the people who become most reliant on the tech, and I don't think there's any good reason to trust any of the companies in that space. For that reason I'd warn people away from it if they can, but some people probably can't avoid it.
I'm thoroughly in the middle here: I work with LLMs regularly and have seen the good, bad, and ugly from them, and I think there's definitely some uses that will probably persist. But I don't advocate for their use; I'm more of a resigned adopter who retains a significant degree of skepticism.
This is a completely fair statement, although I've definitely seen a high degree of reluctance even from folks who do have a reasonable amount of job security, in part because they think (and probably rightly so) that LLMs will disrupt that security.
In my experience the best way to handle that is by retaining robust human-led QA processes. People who just think they can give directions to LLMs and get a good product without a lot of human review are, IMO, going to end up creating a lot more headache for themselves.
I also wrote last year about another 9/11-related episode that also sent me a long way down that path:
I've written about this before, but the real spark for me was finding out about skepticism (methodologically speaking) after getting one of those urban legend Email forwards (the "NASA finds a missing day" one). That led me on a path that resulted in me no longer finding religious claims convincing.
the affordability and iran president
Gotta love that kind of power.
This repeats yet another pattern from the Iraq War. When the Iraqis did not behave as the Americans wanted β without the deference and gratitude that a foreign military occupier felt it was due β they became targets.
I keep seeing this epithet used for said person and I am so grateful that I don't know what it refers to. (Please allow me to remain ignorant. No good can come of this cursed knowledge.)
Are you telling me that a Brit* tried to "well, actually" Birth of a Nation? Because if so that is so much funnier.
*there needs to be a better demonym here
Second-campist American leftists are insanely gullible. Completely unable to see AI slop.
They're the mirror image of gullible MAGA imperialists.
I still see people sharing that obvious fake "US bombed F4 drawings" pic, where the planes are bigger than the buildings. Insane.
βthe fouling of an entire region in an act of mass ecological terrorismβ
- Statement by White House Press Secretary, Nov. 6 1991, describing the burning of oil fields in Kuwait
I wonder what percentage of Americans knows that the CIA helped overthrow a democracy in Iran in 1953, installing a brutal dictator who tortured and murdered people. Republicans keep citing the 1979 Islamic Revolution as if that was the start of U.S.-Iran hostilities. Go back 26 years earlier.
It's extremely clear that the US doesn't really have an objective here except to inflict maximum pain, and attacking water supply is going to end catastrophically for the entire region. It's not just a war crime, it's approaching a crime against humanity.
In other words Trump-Netanyahu have effectively carried out massive uncontained poison gas attacks on Tehran and its people.
And I say that as a former English teacher with ample knowledge of writing pedagogy and a current web developer whose work frequently involves the use of AI tools.
When you understand composition in this light, teaching "AI writing" becomes mostly superfluous. You're teaching students how to fake learning, to conceal aptitude, and you're doing so with a moving targetβevery tutorial that asks you to avoid em-dashes will be obsolete once models are retrained.
Writing is similar in this sense; you need to be familiar with how to structure arguments, employ rhetorical devices, and understand how to tailor language to specific audiences and purposes. But unlike computation, writing *is* the work itself, not something to be offloaded.