Research
Relative Value Functions: A Flexible New Format for Value Estimation
We just published a blog post outlining the details of relative value functions. This was touched on in the presentation I sent out yesterday. https://forum.effectivealtruism.org/posts/EFEwBvuDrTLDndqCt/relative-value-functions-a-flexible-new-format-for-value The post has several code sections and tables, so I won’t attempt to copy that over here. Instead,
Presentation: Estimating Everything Everywhere Always
An overview of our plan at QURI
Conveniences in Thought and Communication
Communication often makes winners and losers. I propose we use the term "convenience" to help clarify what's going on.
Accuracy Agreements: A Flexible Alternative to Prediction Markets
A simple proposal to convert money into complex forecasts
Some estimation work in the horizon
Too much work for any one group
Estimation for sanity checks
I feel very warmly about using relatively quick estimates to carry out sanity checks, i.e., to quickly check whether something is clearly off, whether some decision is clearly overdetermined, or whether someone is just bullshitting. This is in contrast to Fermi estimates, which aim to arrive at an estimate
Use of “I’d bet” on the EA Forum is mostly metaphorical
Epistemic status: much ado about nothing.
Straightforwardly eliciting probabilities from GPT-3
I explain two straightforward strategies for eliciting probabilities from language models, and in particular for GPT-3, provide code, and give my thoughts on what I would do if I were being more hardcore about this. Straightforward strategies Look at the probability of yes/no completion Given a binary question, like
Six Challenges with Criticism & Evaluation Around EA
Continuing on "Who do EAs Feel Comfortable Critiquing?”
Eli Lifland, on Navigating the AI Alignment Landscape
Recently I had a conversation with Eli Lifland about the AI Alignment landscape. Eli Lifland has been a forecaster at Samotsvety and has been investigating said landscape. I’ve known Eli for the last 8 months or so, and have appreciated many of his takes on AI alignment strategy. This
Misha Yagudin and Ozzie Gooen Discuss LLMs and Effective Altruism
Cross-post on the EA Forum Misha and I recently recorded a short discussion about large language models and their uses for effective altruists. This was mostly a regular Zoom meeting, but we added some editing and text transcription. After we wrote up the transcript, both Misha and myself edited our
An in-progress experiment to test how Laplace’s rule of succession performs in practice.
Summary I compiled a dataset of 206 mathematical conjectures together with the years in which they were posited. Then in a few years, I intend to check whether the probabilities implied by Laplace’s rule—which only depends on the number of years passed since a conjecture was created—are