Research
The Value of Existential Risk Mitigation (2023)
Abstract
This report investigates the value of existential risk mitigation by extending the model developed by Ord, Thorstad and Adamczewski. The work here uses more realistic assumptions, like sophisticated risk structures, variable persistence and new cases of value growth. By enriching the base model, we are able to perform sensitivity analyses and can better evaluate when existential risk mitigation should, in expectation, be a global priority. [Full Report] [Abridged Report] [Jupyter Notebook] [Post and Summary]
Simulations for financial management to study the observed effects of unconditional cash transfers (2023)
Abstract
This work is part of Jaroszewicz et al. (2022). To better elucidate the empirical findings of the underlying mechanism, we propose a relatively simple discrete time model that captures how an agent allocates scarce resources while managing their finances. The basic structure of the model focuses on the fact that the act of [re]optimizing one’s financial choices in the face of new information incurs its own costs, and hence that there is a tradeoff between ‘passively’ following a predetermined plan versus ‘actively’ engaging with an entire financial portfolio. [Code] [Relevant Paper’s Appendix]
General Parallelisability in ‘Existential Risk and Growth’ (2023)
Abstract
‘Existential Risk and Growth’ (Aschenbrenner, 2020) assumes equal λ, parallelisability parameters, across the safety and consumption sectors. This work relaxes this assumption, and it observes any differences in the results that follow. [In progress.][Available upon request.]
Examples of Ethical Consumerism (2022)
Abstract
Based on Philip Trammell’s paper in progress ‘Ethical Consumerism’ (2022), this document presents a detailed mathematical example of the results found in the paper. In this context, consumers have preferences over both their own consumption and, out of their ethical concerns, over the global supply of products, like chicken or solar panels. The code can be found here. The reader can edit it to taste to toy with a similar class of examples. [In progress.*]
Strategic Verifiable Communication and the Deployment of Advanced Artificial Intelligence (2022)
Abstract
This paper explores verifiable communication scenarios between artificial intelligence (AI) developers. These actors can opt to disclose the safety and progress of the risky technology they have built, and, having sometimes learnt these facts about their competitor, they decide whether or not to deploy their AI. In contrast to previous research, we find that societal welfare, measured in terms of minimising the risk of an AI catastrophe, can be improved by AI developers sharing and hiding information at the appropriate times, and we outline such cases. [In progress.] [Available upon request.]
Dialetheism and On Escaping the Strengthened Liar’s Paradox (2021)
Abstract
The options explored in this essay seem, at least in the first instance, to be the different available strategies that the dialetheist could adopt, and they all have considerable problems. As it stands the dialetheist does not have any satisfying way of answering the challenge posed by '“the strengthened liar’s” and escape the semantic paradox headache that it was designed to avoid.
Other Writings
[Being updated…]