@sarker's banner p

sarker

competency crisis actor

0 followers   follows 0 users  
joined 2022 September 05 16:50:08 UTC

Suddenly I cannot remember the color of your eyes

Or the things we said as we stood together for the last time


				
User ID: 636


				
				
				

			

		

	














	


		


			



	

		

			

			

				

					

						

							
						

						

							
							

								
sarker


								

								

								
								
								

								
							
					
 
				


				
competency crisis actor

				
				
				

				

					0 followers  

					follows 0 users  

					joined 2022 September 05 16:50:08 UTC
				


				
					

					
Suddenly I cannot remember the color of your eyes


Or the things we said as we stood together for the last time



				

				

				

				

				

					

					

					
					
					

					
User ID: 636


					
					
					
				

			


					

						
					



		

	

 











			
			

			
			

			




	


		

			
‎

			

				
				

					Hour
					Day
					Week
					Month
					Year
					
				

			


			
‎

			

	
	

	 Hot Bump Top Bottom Old Controversial Comments
	



		

	






	

		
		

		
		
	

	

  







		

			
			
				Culture War Roundup for the week of August 3, 2026
			
		

	

	

	

		
			

			

				

			

		

			


				

					


I can't think of any other technology in history that has achieved 50 percent market penetration in 3-4 years.




What did you think takeoff meant? Vibes? Papers?




and they both start to take off in early 2023, which was about how I remembered when I first started hearing about them.




That's when GPT-4 hit the public imagination.




HALF of Americans are doing this weekly?




Yes!




My circle must just be in the stone age, then, because I still know multiple people who don't own computers.




About 9% of American adults don't have a smartphone, so sure, it's not out of the question that you know several of them.


				

				

						

							

								
							

						


						
					

				
			


			
			
			
				

			
		

	






		

			
			
				Culture War Roundup for the week of August 3, 2026
			
		

	

	

	

		
			

			

				

			

		

			


				

					


I suppose it would include things like Alexa/Google Home/Siri?




Well, no, those are not chatbots.




I wasn't even hearing the term "A.I." as anything other than the movie (or Allen Iverson) in any remotely mainstream sense, unless post-Covid.




Huh? The prominence of the AI boom is largely a post Covid phenomenon.


				

				

						

							

								
							

						


						
					

				
			


			
			
			
				

			
		

	






		

			
			
				Culture War Roundup for the week of August 3, 2026
			
		

	

	

	

		
			

			

				

			

		

			


				

					
Almost half of Americans use chatbots specifically, with about a quarter doing so daily.


				

				

						

							

								
							

						


						
					

				
			


			
			
			
				

			
		

	






		

			
			
				Culture War Roundup for the week of August 3, 2026
			
		

	

	

	

		
			

			

				

			

		

			


				

					


I'll bet less than 5% of people have ever used AI in any form,




Almost a supermajority of Americans use AI at least multiple times a week. Let me know how you want to pay up.


				

				

						

							

								
							

						


						
					

				
			


			
			
			
				

			
		

	






		

			
			
				Culture War Roundup for the week of August 3, 2026
			
		

	

	

	

		
			

			

				

			

		

			


				

					
Ironically, the pizza in the third world is much cheaper.


				

				

						

							

								
							

						


						
					

				
			


			
			
			
				

			
		

	






		

			
			
				Culture War Roundup for the week of August 3, 2026
			
		

	

	

	

		
			

			

				

			

		

			


				

					
14" Domino's "ultimate pepperoni" is $26 plus $6 delivery to my house.


				

				

						

							

								
							

						


						
					

				
			


			
			
			
				

			
		

	






		

			
			
				Culture War Roundup for the week of August 3, 2026
			
		

	

	

	

		
			

			

				

			

		

			


				

					


The cost savings feel minimal on a per meal basis




What? A pizza for "one" (10 inches) from a chain pizza joint in town is $20 before any delivery fees.


				

				

						

							

								
							

						


						
					

				
			


			
			
			
				

			
		

	






		

			
			
				Small-Scale Question Sunday for August 2, 2026
			
		

	

	

	

		
			

			

				

			

		

			


				

					
Don't worry, now that AIs are using search people are actually going to give a shit about this. exa.ai surfaces the obituary as the top result.


				

				

						

							

								
							

						


						
					

				
			


			
			
			
				

			
		

	






		

			
			
				Weekly Finance Thread - 2026-08-01
			
		

	

	

	

		
			

			

				

			

		

			


				

					
You should probably just start by looking at ACE recommended charities which mostly work on improving the lives of farmed animals.


As a vegetarian, I'd love it if we just stopped slaughtering and torturing animals, but the first goal is pretty remote. I'd be happy if we stopped putting male chicks in shredders, sows in cages, and so on, which are actually feasible goals.


				

				

						

							

								
							

						


						
					

				
			


			
			
			
				

			
		

	






		

			
			
				Culture War Roundup for the week of July 27, 2026
			
		

	

	

	

		
			

			

				

			

		

			


				

					


do not deserve to be compensated in such a way that someone would work {dirty job} voluntarily




They are doing it voluntarily. It's just that they are voluntarily doing it for wages that some people believe are too low. It's basically "the myth of consensual employment".


				

				

						

							

								
							

						


						
					

				
			


			
			
			
				

			
		

	






		

			
			
				Small-Scale Question Sunday for August 2, 2026
			
		

	

	

	

		
			

			

				

			

		

			


				

					
Income taxes payable to the city are a ridiculous burden on productive economic activity. NYC ought to have a land value tax to discourage disused dwellings instead. Such a tax also cannot be dodged via "primary residence" shenanigans unlike the pied a terre tax.


				

				

						

							

								
							

						


						
					

				
			


			
			
			
				

			
		

	






		

			
			
				Friday Fun Thread for July 31, 2026
			
		

	

	

	

		
			

			

				

			

		

			


				

					
It's not about mid-range anymore. K3 is quite close on the index to the absolute top of the line. Some say Chinese models are still six months behind, some say the gap is closing, but nobody thinks it's increasing. It's also quite clear that they don't have the compute that the US had six months ago. Seems you don't need it to reach that level of performance.


				

				

						

							

								
							

						


						
					

				
			


			
			
			
				

			
		

	






		

			
			
				Friday Fun Thread for July 31, 2026
			
		

	

	

	

		
			

			

				

			

		

			


				

					
Of course he's going to complain. More helps more. They are, however, at the (pareto) frontier.


				

				

						

							

								
							

						


						
					

				
			


			
			
			
				

			
		

	






		

			
			
				Friday Fun Thread for July 31, 2026
			
		

	

	

	

		
			

			

				

			

		

			


				

					
This was already questionable advice in the age of search and is now probably worthless advice in the age of LLMs.


				

				

						

							

								
							

						


						
					

				
			


			
			
			
				

			
		

	






		

			
			
				Friday Fun Thread for July 31, 2026
			
		

	

	

	

		
			

			

				

			

		

			


				

					


You do not want to be a cash and compute-poor lab.




Hmm. Like Moonshot or DeepSeek? Every lab is cash and compute poor compared to Google.


				

				

						

							

								
							

						


						
					

				
			


			
			
			
				

			
		

	






		

			
			
				Wellness Wednesday for July 29, 2026
			
		

	

	

	

		
			

			

				

			

		

			


				

					
I guess it's a testament to how far food science has come if you weren't immediately aware that it wasn't sugar free.


				

				

						

							

								
							

						


						
					

				
			


			
			
			
				

			
		

	






		

			
			
				Culture War Roundup for the week of July 27, 2026
			
		

	

	

	

		
			

			

				

			

		

			


				

					


I'm pretty sure the non-destructive scanners are faster and cheaper




Most non destructive scanners require someone to turn each page of the book. Google books invented a complicated machine to do this. Slicing the book open is much simpler, you can feed it to a COTS multi-page scanner.


				

				

						

							

								
							

						


						
					

				
			


			
			
			
				

			
		

	






		

			
			
				Small-Scale Question Sunday for July 26, 2026
			
		

	

	

	

		
			

			

				

			

		

			


				

					
A quick wiki skim suggests 2008 as the first likely success in making cloned human embryos. Interest in human cloning has been pretty consistently declining for as long as there's data (since 2005) up until late 2025 and we are now at a ten year high.


				

				

						

							

								
							

						


						
					

				
			


			
			
			
				

			
		

	






		

			
			
				Culture War Roundup for the week of July 20, 2026
			
		

	

	

	

		
			

			

				

			

		

			


				

					
I've never researched it deeply but it seems like it's Minecraft with a bunch of degenerate features bolted on. It's full of groomers because it's full of children and includes a chatbox.


				

				

						

							

								
							

						


						
					

				
			


			
			
			
				

			
		

	






		

			
			
				Culture War Roundup for the week of July 20, 2026
			
		

	

	

	

		
			

			

				

			

		

			


				

					


My point is that there is a difference between a model that misunderstands your intentions and can be stopped at any time by saying, 'oh, no, that's not what I meant' and a model that is totally uninterested in anything you say after it starts working while treating you as a potential enemy.




There's much less difference when we're talking about swarms of autonomous systems thinking in Neuralese at 1000tps. People are not going to hit 'approve' every time the model wants to run ls. The fact that the model's intrusion could have been stopped with SIGTERM did not help HuggingFace at all. I guess we can rest easy knowing that if you're getting paper clipped you can write a blog post about it and in 5-10 business days OpenAI will claim responsibility apologize and then people will say that there's nothing wrong here.


				

				

						

							

								
							

						


						
					

				
			


			
			
			
				

			
		

	






		

			
			
				Culture War Roundup for the week of July 20, 2026
			
		

	

	

	

		
			

			

				

			

		

			


				

					
The first is a theory, not a fallacy. The fallacy is due to Bastiat.


Edit: Zorg clearly read Bastiat and got the wrong message.


				

				

						

							

								
							

						


						
					

				
			


			
			
			
				

			
		

	






		

			
			
				Culture War Roundup for the week of July 20, 2026
			
		

	

	

	

		
			

			

				

			

		

			


				

					
It "made a mistake" in the same way that a paperclip maximizer "made a mistake" by converting the universe into paperclips rather than increasing factory productivity by 5%. Literally the entire point of the hypothetical and the reality of this incident is that you can't reasonably enumerate every single thing you don't want the model to do. I am surprised you don't seem to understand this, or at least address this, given your claims of having followed this debate for years.


				

				

						

							

								
							

						


						
					

				
			


			
			
			
				

			
		

	






		

			
			
				Culture War Roundup for the week of July 20, 2026
			
		

	

	

	

		
			

			

				

			

		

			


				

					
This is just the broken window fallacy but for kids.


				

				

						

							

								
							

						


						
					

				
			


			
			
			
				

			
		

	






		

			
			
				Culture War Roundup for the week of July 20, 2026
			
		

	

	

	

		
			

			

				

			

		

			


				

					


In this case 'Union Carbide' is selling those plants.




You're absolutely right. Allow me to restate.


"Sanlu Group doesn't want their formula to poison infants any more than anyone else does, so no need for government with the big hammer."




Misaligned AI isn't an externality, it's a bad product, and companies are wise to that which is one reason why all this testing is happening.




This was not a test of alignment. In any case, if even training can result in real world harm, that is even worse for your head in the sand position.




I don't think so. It indicates that the AI is sincerely trying to work out what you want as opposed to deliberately ignoring what you want in favour of the specific instructions you gave it. To my mind, the former is what alignment is.




It's quite clear that OpenAI did not want the model to hack huggingface. This is classic paperclip maximizer stuff.


Broadly, you are moving the goalposts. You did not believe in AI risk because there was no evidence of harm. Now there is evidence of harm, but it's OK because actually the model was supposed to do it.


				

				

						

							

								
							

						


						
					

				
			


			
			
			
				

			
		

	






		

			
			
				Culture War Roundup for the week of July 20, 2026
			
		

	

	

	

		
			

			

				

			

		

			


				

					


OpenAI doesn’t want their models going rogue any more than anyone else does, no need for government with the big hammer.




"Union Carbide doesn't want their plants to emit poison gas any more than anyone else does, no need for government with the big hammer."




The model understood it was being tested on its cyber capabilities (which has precedent, Claude has done that too) and went the extra mile to succeed at the implicit task. Especially since all the systems that usually tell it not to do this were deliberately turned off for the test. Still a problem but much easier to manage.




This is in fact much harder to manage because it would indicate the model is fundamentally misaligned and that we actually are much worse at alignment than we thought.