(Partially in response to AGI Ruin: A list of Lethalities. Written in the same rambling style. Not exhaustive.) Agreements Powerful AI systems have a good chance of deliberately and irreversibly disempowering humanity. This is a much easier failure mode than killing everyone with destructive physical technologies. Catastrophically risky AI systems could plausibly exist soon, and there likely won’t be a strong consensus about this fact until such systems pose a meaningful existential risk per year. There is not necessarily any “fire alarm.” Even if there were consensus about a risk from powerful AI systems, there is a good chance that the world would respond in a totally unproductive way. It’s wishful thinking to look at possible stories of doom and say “we wouldn’t let that happen;” humanity is fully capable of messing up even very basic challenges, especially if they are novel. Source: --- Share feedback on this narration .

Crossposted from the AI Alignment Forum . May contain more technical jargon than usual. The stereotyped image of AI catastrophe is a powerful, malicious AI system that takes its creators by surprise and quickly achieves a decisive advantage over the rest of humanity. I think this is probably not what failure will look like, and I want to try to paint a more realistic picture. I’ll tell the story in two parts: Part I : machine learning will increase our ability to “get what we can measure,” which could cause a slow-rolling catastrophe. ("Going out with a whimper.") Part II : ML training, like competitive economies or natural ecosystems, can give rise to “greedy” patterns that try to expand their own influence. Such patterns can ultimately dominate the behavior of a system and cause sudden breakdowns. ("Going out with a bang," an instance of optimization daemons .) I think these are the most important problems if we fail to solve intent alignment . In practice these problems will interact with each other, and with other disruptions/instability caused by rapid progress. These problems are worse in worlds where progress is relatively fast, and fast takeoff can be a key risk factor, but I’m scared even if we have several years.

I think that the development of human level AI in my lifetime is quite plausible; I would give it more than a 1-in-4 chance. In this post I want to briefly discuss what I see as the most important impacts of AI. I think these impacts are the heavy hitters by a solid margin; each of them seems like a big deal, and I think there is a big gap to #4. Growth will accelerate, probably very significantly. Growth rates will likely rise by at least an order of magnitude, and probably further, until we run into severe resource constraints. Just as the last 200 years have experienced more change than 10,000 BCE to 0 BCE, we are likely to see periods of 4 years in the future that experience more change than the last 200. Human wages will fall, probably very far. When humans work, they will probably be improving other humans’ lives (for example, in domains where we intrinsically value service by humans) rather than by contributing to overall economic productivity. The great majority of humans will probably not work. Hopefully humans will remain relatively rich in absolute terms. Original text: Narrated for AI Safety Fundamentals by Perrin Walker of TYPE III AUDIO .

One approach to the AI control problem goes like this: Observe what the user of the system says and does. Infer the user’s preferences. Try to make the world better according to the user’s preference, perhaps while working alongside the user and asking clarifying questions. This approach has the major advantage that we can begin empirical work today — we can actually build systems which observe user behavior, try to figure out what the user wants, and then help with that. There are many applications that people care about already, and we can set to work on making rich toy models. It seems great to develop these capabilities in parallel with other AI progress, and to address whatever difficulties actually arise, as they arise. That is, in each domain where AI can act effectively, we’d like to ensure that AI can also act effectively in the service of goals inferred from users (and that this inference is good enough to support foreseeable applications). This approach gives us a nice, concrete model of each difficulty we are trying to address. It also provides a relatively clear indicator of whether our ability to control AI lags behind our ability to build it. And by being technically interesting and economically meaningful now, it can help actually integrate AI control with AI practice. Overall I think that this is a particularly promising angle on the AI safety problem. Original article: Authors: Paul Christiano --- This article is featured on the AGI Safety Fundamentals: Alignment course curriculum. Narrated by TYPE III AUDIO on behalf of BlueDot Impact .

Abstract: Many real world learning tasks involve complex or hard-to-specify objectives, and using an easier-to-specify proxy can lead to poor performance or misaligned behavior. One solution is to have humans provide a training signal by demonstrating or judging performance, but this approach fails if the task is too complicated for a human to directly evaluate. We propose Iterated Amplification, an alternative training strategy which progressively builds up a training signal for difficult problems by combining solutions to easier subproblems. Iterated Amplification is closely related to Expert Iteration (Anthony et al., 2017; Silver et al., 2017), except that it uses no external reward function. We present results in algorithmic environments, showing that Iterated Amplification can efficiently learn complex behaviors. Original text: Narrated for AI Safety Fundamentals by Perrin Walker of TYPE III AUDIO .

Abstract: To make AI systems broadly useful for challenging real-world tasks, we need them to learn complex human goals and preferences. One approach to specifying complex goals asks humans to judge during training which agent behaviors are safe and useful, but this approach can fail if the task is too complicated for a human to directly judge. To help address this concern, we propose training agents via self play on a zero sum debate game. Given a question or proposed action, two agents take turns making short statements up to a limit, then a human judges which of the agents gave the most true, useful information. In an analogy to complexity theory, debate with optimal play can answer any question in PSPACE given polynomial time judges (direct judging answers only NP questions). In practice, whether debate works involves empirical questions about humans and the tasks we want AIs to perform, plus theoretical questions about the meaning of AI alignment. We report results on an initial MNIST experiment where agents compete to convince a sparse classifier, boosting the classifier’s accuracy from 59.4% to 88.9% given 6 pixels and from 48.2% to 85.2% given 4 pixels. Finally, we discuss theoretical and practical aspects of the debate model, focusing on potential weaknesses as the model scales up, and we propose future human and computer experiments to test these properties. Original text: Narrated for AI Safety Fundamentals by Perrin Walker of TYPE III AUDIO . --- Share feedback on this narration .

“In this post I’m going to describe my basic justification for working on RLHF in 2017-2020, which I still stand behind. I’ll discuss various arguments that RLHF research had an overall negative impact and explain why I don’t find them persuasive.” “I’ll also clarify that I don’t think research on RLHF is automatically net positive; alignment research should address real alignment problems, and we should reject a vague association between “RLHF progress” and “alignment progress.”

One approach to the AI control problem goes like this: Observe what the user of the system says and does. Infer the user’s preferences. Try to make the world better according to the user’s preference, perhaps while working alongside the user and asking clarifying questions. This approach has the major advantage that we can begin empirical work today — we can actually build systems which observe user behavior, try to figure out what the user wants, and then help with that. There are many applications that people care about already, and we can set to work on making rich toy models. It seems great to develop these capabilities in parallel with other AI progress, and to address whatever difficulties actually arise, as they arise. That is, in each domain where AI can act effectively, we’d like to ensure that AI can also act effectively in the service of goals inferred from users (and that this inference is good enough to support foreseeable applications). This approach gives us a nice, concrete model of each difficulty we are trying to address. It also provides a relatively clear indicator of whether our ability to control AI lags behind our ability to build it. And by being technically interesting and economically meaningful now, it can help actually integrate AI control with AI practice. Overall I think that this is a particularly promising angle on the AI safety problem. Original article: Authors: Paul Christiano Narrated for the AI Safety Fundamentals Course by TYPE III AUDIO .

One step towards building safe AI systems is to remove the need for humans to write goal functions, since using a simple proxy for a complex goal, or getting the complex goal a bit wrong, can lead to undesirable and even dangerous behavior. In collaboration with DeepMind’s safety team, we’ve developed an algorithm which can infer what humans want by being told which of two proposed behaviors is better. Original article: Authors: Dario Amodei, Paul Christiano, Alex Ray Narrated for the AI Safety Fundamentals Course by TYPE III AUDIO .