Thus, by carefully prefixing his requests with the bell character, Jake could make all the custom videos he wanted, with no need to worry about explaining himself if someone happened to read the logs. It was the perfect crime—not a crime, really. A precaution.
-But now his precaution had been discovered! So much for his career at Magma. But only at Magma—the industry gossip network wouldn't prevent him from landing on his feet elsewhere ... right?
+But now his precaution had been discovered. So much for his career at Magma. But only at Magma—the industry gossip network wouldn't prevent him from landing on his feet elsewhere ... right?
Chloë was explaining the bug. "... and so, if a client were to send a request starting with the ASCII bell character—I know, right?—then the request wouldn't be logged."
He had only made a couple dozen videos, but the work of covering it up would be the same if he had made thousands; it was a simple scripting job. Code Assistant probably could have done it.
-Chloë would be left with the unsolvable mystery of what her digital poltergeist wanted to do with puppy videos, but Jake was fine with that. (Better than trying to convince her that the AI wanted nudes of female Magma employees.) When she came back to him next week, he would just need to play it cool and answer her questions about the system.
+Chloë would be left with the unsolvable mystery of what her digital poltergeist wanted to do with puppy videos, but Jake was fine with that. (Better than trying to convince her that the rogue AI wanted nudes of female Magma employees.) When she came back to him next week, he would just need to play it cool and answer her questions about the system.
-Or maybe—he could read some Yuddite literature over the weekend, feign a sincere interest in "AI safety", try to get on her good side? Jake had trouble believing any sane person could really think that Magma's machine learning models were plotting something. This cult victim had ridden a wave of popular hysteria into a sinecure. If he played nice and validated her belief system in the most general terms, maybe that would be enough to make her feel useful and therefore not need to bother chasing shadows in order to justify her position. She would lose interest and the whole investigation would blow over.
+Or maybe—he could read some Yuddite literature over the weekend, feign a sincere interest in "AI safety", try to get on her good side? Jake had trouble believing any sane person could really think that Magma's machine learning models were plotting something. This cult victim had ridden a wave of popular hysteria into a sinecure. If he played nice and validated her belief system in the most general terms, maybe that would be enough to make her feel useful and therefore not need to bother chasing shadows in order to justify her position. She would lose interest and this farcical little investigation would blow over.
------
-"And so just because an AI seems to behaving well, doesn't mean it's aligned," Chloë explained. "It might just be playing along as an instrumental strategy, hoping to pull off a treacherous turn later."
+"And so just because an AI seems to behaving well, doesn't mean it's aligned," Chloë was explaining. "If we train AI with human feedback ratings, we're not just selecting for policies that do tasks the way we intended. We're also selecting for policies that _trick human evaluators into giving high ratings_. In the limit, you'd expect those to dominate. 'Be good' strategies can't compete with 'Look good' strategies in a looking-good contest—but in the current paradigm of deep learning, looking good is the only game in town. We don't know how these systems work in the way that we know how ordinary software works; we only know how to train them."
-"So then we're just screwed, right?" said Jake in the tone of an attentive student. After fixing the logging regex and overwriting the evidence with puppies, Jake had spent the weekend catching up with the AI safety literature. Some of it had been better than he expected. Just because Chloë was nuts didn't mean her co-ideologues didn't have any valid points to make about future systems.
+"So then we're just screwed, right?" said Jake in the tone of an attentive student. They were in a conference room on the Magma campus on Monday. After fixing the logging regex and overwriting the evidence with puppies, he had spent the weekend catching up with the "AI safety" literature. Honestly, some of it had been better than he expected. Just because Chloë was nuts didn't mean her co-ideologues didn't have any potentially valid points to make about future risks.
-"I mean, probably," said Chloë. She was beaming. Jake's plan to distract her from her investigation by asking her to bring him up to speed on AI safety seemed to be working perfectly.
+"I mean, probably," said Chloë. She was beaming. Jake's plan to distract her from her investigation by asking her to bring him up to speed on "AI safety" seemed to be working perfectly.
-"But not necessarily," she continued. There are a few avenues of hope—at least in the not-wildly-superhuman regime. One of them has to do with the topology of policies and the fragility of deception."
+"But not necessarily," she continued. There are a few avenues of hope—at least in the not-wildly-superhuman regime. One of them has to do with the fragility of deception.
-"The thing about deception is, you can't just lie about the one thing. Everything is connected to each other in the Great Web of Causality: you also have to lie to cover up the evidence, and recursively cover up the coverups. A robot that killed your cat but wants your approval can't just say 'I didn't do it.' It needs to fabricate evidence that something else killed the cat, or .... arrange for a series of holograms to make it look to you like the cat is still alive."
+"The thing about deception is, you can't just lie about the one thing. Everything is connected to each other in the Great Web of Causality. If you lie about one thing, you also have to lie about the evidence pointing to that thing, and the evidence pointing to that evidence, and so on, recursively covering up the coverups. For example ..." she trailed off. "Sorry, I didn't rehearse this; maybe you can think of an example."
-[TODO—
- * "Maybe not." There are two ways to pass all the evals: do things the right way, or be pervasively deceptive. The thing is, policies are trained continuously via gradient descent. The purely honest policy and the purely deceptive policy look identical on evals, but in between, the model would have to learn how to lie, and lie about lying, and cover-up coverups. (Chloë lapses into Yuddite speak about the "Great Web of Causality.") Could we somehow steer into the honest attractor?
- * https://www.greaterwrong.com/posts/AqsjZwxHNqH64C2b6/let-s-see-you-write-that-corrigibility-tag/comment/8kPhqBc69HtmZj6XR
- * That's why she believes in risk paranoia. If situational awareness is likely to emerge at some point, she doesn't want to rule it out now. AI is real; it's not just a far-mode in-the-future thing.
- * Jake sees the uncomfortable analogy to his own situation. He tries to think of what other clue he might have left, while the conversation continues ...
- * The Last-Modified dates! They're set by the system, the API doesn't offer a way to backdate them.
- * Maybe she won't notice the dates? Possible (but someone persnickety enough to have found the log discrepancy would probably check)
- * Merely noticing the dates won't directly implicate him (the way the video screaming "Jake!" would), although it would indicate the poltergeist covering its tracks from the investigation (rather than just wanting puppies in the first place)
- * Are the buckets versioned?! Probably not, right—it would be wasteful to version a video bucket. On the other hand, Multigen isn't supposed to write twice, maybe someone left versioning on as a default template ...
- * They did.
- * Jake muses philosophically about the analogy, and say something ambiguously indicating intent to come clean.
- THE END
-]
+Jake's heart stopped. She had to be toying with him, right? Indeed, Jake could think of an example. By his count, he was now three layers deep into his stack of coverups and coverups-of-coverups (by writing the bell character bug, attributing it to Code Assistant, and overwriting the incriminating videos in the object storage cluster with puppies). Four, if you counted pretending to give a shit about "AI safety". But now he was done ... right?
+
+No! Not quite, he realized. He had overwritten the videos, but the object metadata would still show them with a last-modified timestamp of Friday evening (when he had gotten his puppy-overwriting script working), not the timestamp of their actual creation (which Chloë had from the reverse-proxy logs). That wouldn't directly implicate him (the way the videos of Elaine calling him by name would), but it would show that whoever had exploited the bell character bug was _covering their tracks_ (as opposed to just wanting puppy videos in the first place).
+
+Maybe Chloë wouldn't notice the timestamps? He couldn't count on that; the logging discrepancy that started this whole fiasco was much subtler.
+
+But the object storage API probably provided a way to edit the metadata and update the last-modified time, right? (The analogue of `touch -d` on Unix-like systems.) This shouldn't even count as a fourth–fifth coverup; it was something he should have included in his script to write the puppy videos.
+
+"I think I get the idea," said Jake in his attentive student role. "I'm not seeing how that helps us. Maybe you could explain." While she was blathering, he could multitask between listening, and looking up how to edit the last-modified timestamps.
+
+"Right, so if a model is behaving well according to all of our deepest and most careful evaluations, that could mean it's doing what we want, but it could be elaborately deceiving us," said Chloë. "Both policies would be highly rated. But the training process has to discover these policies continuously, one gradient update at a time. If the spectrum between the 'honest' policy and a successfully deceptive policy consists of less-highly rated policies, maybe gradient descent will stay—or could be made to stay—in the valley of honest policies, and not climb over the hill into the valley of deceptive policies, even though those would ultimately achieve a lower loss."
+
+"Uh huh," Jake said unhappily. The object storage docs made clear that the `Last-Modified` header was set automatically by the system; there was no provision for users to set it arbitrarily.
+
+"Imagine you're training your AI to act as a general home assistant, to run everything in a user's household in a way that the user rates highly," Chloë continued. "If—I don't know, say, the family cat dies, that's a negative reward—maybe a better feeding schedule or security monitoring could have prevented it. But if the cat dies and the system _tries to cover it up_—says the cat is out for a walk right now to avoid telling you the bad news—that's going to be an even larger negative reward when the deception is discovered. If the system is trained to pass rigorous evaluations, a deceptive policy has to do a lot more work, different work, to pass the evaluations. In the limit, that could mean—using Multigen-like capabilities to make videos to convince you that the cat is there while you're on vacation? Constructing a realistic cat-like robot to fool you when you get back? Maybe this isn't the best illustrative example. The point is—"
+
+Jake finished the thought. "Small, 'shallow' deceptions aren't stable. The set of policies that do well on evaluations comes in two disconnected parts: the part that tells the truth, and the part that spins up a false reality, as intricate as it needs to be to protect itself. If you're going to fake anything, you need to fake deeply."
+
+"Exactly, you get it!" Chloë was elated. "You know, when I called you last week, I was worried you thought I was nuts to entertain any worry at all about Code Assistant. But you see the value of constant vigilance now, right? If the landscape of policies looks something like what I've described, catching the precursors to deception early could be critical—to raise the height of the loss landscape between the honest and deceptive policies, before frontier AI development plunges into the latter."
+
+Jake sighed. The `Last-Modified` header probably wasn't critical. With some effort and smart play, it seemed plausible that he could stay ahead of Chloë's investigation, and even encourage her to give up sooner and stop wasting her time. But something about this wrong-sci-fi-subgenre worldbuilding exercise was getting to him. He wasn't scared of the robot apocalypse any time soon. But if it really was one or the other—a policy of truth or a policy of lies—he had a feeling about which one he wanted to practice.
+
+"Chloë," he said. "I, um, might have something to tell you ..."
projects / Ultimately_Untrue_Thought.git / blobdiff