Artificially Intelligent

Any mimicry distinguishable from the original is insufficiently advanced.

  • Meta-EA Needs Models

    | 1274 words

    Thanks to Kuhan Jeyapragasan, Michael Byun, Sydney Von Arx, Thomas Kwa, Jack Ryan, Adam Křivka, and Buck Shlegeris for helpful comments and discussion. Epistemic status: a bunch of stuff Sometimes I have conversations with people that go like this: Me: Feels like all the top people in EA would have...

  • RSS Feed

    | 16 words

    Multiple people have requested an RSS feed for my blog. I have created one here.

  • Transparency Trichotomy

    | 2062 words

    Crossposted from the AI Alignment Forum. May contain more technical jargon than usual. Introduction In Relaxed Adversarial Training for Inner Alignment, Evan presents a trichotomy of ways to understand a model M: Transparency via inspection: use transparency tools to understand M via inspecting the trained model. Transparency via training: incentivize...

  • Intermittent Distillations #1

    | 3166 words

    This is my low-budget version of Rohin’s Alignment Newsletter. A critique of pure learning and what artificial neural networks can learn from animal brains (Anthony M. Zador) A critique of pure learning and what artificial neural networks can learn from animal brains Summary This paper points out that human learning...

  • Strong Evidence is Common

    | 289 words

    Portions of this are taken directly from Three Things I’ve Learned About Bayes’ Rule. One time, someone asked me what my name was. I said, “Mark Xu.” Afterward, they probably believed my name was “Mark Xu.” I’m guessing they would have happily accepted a bet at 20:1 odds that my...

  • Open Problems in Myopia

    | 2416 words

    Coauthored with Evan Hubinger. Crossposted from the AI Alignment Forum. May contain more technical jargon than usual. Thanks to Noa Nabeshima for helpful discussion and comments. Thanks to Noa Nabeshima for helpful discussion and comments. Introduction Certain types of myopic agents represent a possible way to construct safe AGI. We...

  • Towards a Mechanistic Understanding of Goal-Directedness

    | 1408 words

    Crossposted from the AI Alignment Forum. May contain more technical jargon than usual. This post is part of the research I have done at MIRI with mentorship and guidance from Evan Hubinger. Introduction Most discussion about goal-directed behavior has focused on a behavioral understanding, which can roughly be described as...

  • An Ode to My Parents

    | 696 words

    Sometimes I talk to my friends about their parents. I generally come away disoriented. I didn’t regard my parents as particularly exceptional; they were adequate at the task of raising me but not exemplary. However, what little I know about how parenting seems to work suggests that adequacy is exemplary....

  • Revenge of the Prediction Market

    | 642 words

    Recommended reading: Prediction Markets: Tales from the Election Suppose I wanted to know the probability of some future event. How might I do this? One way would be to pay forecasters from the Good Judgment Project to forecast the event. These forecasters are generally pretty good at what they do,...

  • Maslow First and the World Second

    | 652 words

    Saul McLeod: Maslow’s hierarchy of needs is a motivational theory in psychology comprising a five-tier model of human needs, often depicted as hierarchical levels within a pyramid. From the bottom of the hierarchy upwards, the needs are: physiological (food and clothing), safety (job security), love and belonging needs (friendship), esteem,...