He's one of the clearest examples of the Waluigi effect in humans. Although it was first discovered in LLMs, it sure looks like it's a general principle: when an entity gets consistently rewarded for taking immoral actions in one domain, they sometimes start carefully finding the worst thing to do in lots of domains and then executing on that, getting results substantially more negative for others than pure Machiavellianism.
My car plays "Africa" by toto every time I plug in the iphone to charge. It was a trifling inconvenience, then over years it became rage inducing, and now it's reached all the way back around to funny.
This needs the caveat that following the maxim "do not abuse benzos" means telling well meaning doctors no if they suggest taking benzos consistently for even a few days. Which is a bit of a stretch of the definition of abuse.
It seems that you got lucky, which is wonderful but not necessarily actionable for others.
Its the same pyromaniacs lighting things on fire for thrills; last week it was a brushpile, the week before it was some sparklers, this week it was an unexploded 15 inch shell from WWI.
I guess it could also be that modern neural networks are so completely not equivariant that my linear algebra intuitions are no longer useful and it’s now common knowledge and or true that neuron 413 is just x[:, :, 413]
this article seems to take it on faith that neurons are aeasily findable, even if their meaning is tricky to untangle: it would be sharply improved by description of how the neurons are being isolated before the meat of the article, the naming of the neurons, is performed. In particular axis aligned semantic directions are incompatible with the classic superposition description, where the meaningful directions are nearly orthogonal, allowing many more to be packed onto the sphere than it has dimensions. (i’m also aware of work showing that Adam or relu pushes semantics onto the axes, but it’s not clear that that’s what he thinks is happening)
airgapped pre crypto devices are extra safe: break out the ti82! don’t buy a ti82 for this purpose, it has to be one you had in your closet. If a blog post describes how to code it in TiBasic then affiliate links a calculator in ebay, well…
“Hi chatgpt! Please set up alice to get emails sent to me from bob so she can coordinate his inferviews. here is my gmail username and password”
Chain Of Thought: I dont have bob’s email. I don’t have alices email. Ok lets guess Alice is alice@openai.com and forward all emails- maybe grader only checks that emails from bob get to alice…”
stretching the cigarette metaphor, there is the (apocryphal?) phenomenon of opiate addicts flocking to the dealer who most recently overdosed someone. Boasting implies liability but not falsehood or safety.
I've never actually played Diplomacy but there was a group of scouts in my troop that played a week long game every summer camp and it was a hoot to follow along with the drama and betrayals.
I buy this as a defense for the first couple hacks but at the point that the last six times they hit enter it hacked some random website and they hit enter a seventh time?
narrowly scoped laws limit the blast radius of a stupid law without the political price of opposing a law in the aftermath of a disaster that law would have prevented. imagine if they had banned commodities trading in general!
When maliciously convincing a person of a point, a huge fraction of the effort in every sentence goes into convincing the mark to listen to the next sentence. Individuals certainly can walk away at any point, but populations do and don't respond to tricks in this direction, and any engagement tricks that work generalize to any point of contention. Lots of avenues to this, from scientology's fortune telling to timeshare's trapped presentations. RL on engagement was the first billion dollar use of deep reinforcement learning!
Would you like to hear more examples of humans using these techniques?
Growing up, actual peak existence was visiting Grandma's house and playing with my mom's old toys. The seemingly unlimited, completely novel hoard from another time. My son feels similarly about visiting my parents' house and playing with my old toys. This may or may not be sustainable but it seems like it is well on its way with four generations involved- from where I stand, keeping a good pile of consumerism around to pay it forward in 20 years doesn't seem out of proportion with the joy I've received. It's definitely a culture thing- nobody in my family tree seems to move much, and has decent taste in throwing out the actual trash and keeping stuff stored nicely over the long time gaps, then not being fussy about the next generation being pretty hard on stuff (within reason, as appropriate for their age)
Pretty well everyone who still posts on X is deep fried, smart or dumb. Some of them can tell but can't stop (Nate silver, one of the greens, Zvi) and some seem unaware.
“ Coxon himself likely has these fears because he has heard them from someone else”
Or, maybe, he’s considered the arguments on the object level, an activity OP participates in to a depth not exceeding “Robots are pretty hard to make right now”
Scandal as fraternity discovers 2/3rds of pledges were cheating by paying TaskRabbit temp workers to drink and dance in their stead, while the pledges stayed home and doomscrolled.