Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is part four of: Forecasting TAI with biological anchors, published by Ajeya Cotra. Part 4: Timelines estimates and responses to objections This report emerged from discussions with our technical advisors Dario Amodei and Paul Christiano. However, it should not be treated as representative of either of their views; the project eventually broadened considerably, and my conclusions are my own. This is a work in progress and does not represent Open Philanthropy’s institutional view. We are making it public to make it easier to gather feedback, to help inform others’ thinking in the effective altruism community, and to allow for follow-on work outside of Open Phil. However, we may edit it substantially in the future as we gather feedback from a broader audience and investigate open questions. Accordingly we have not done an official publication or blog post, and would prefer for now that people not share it widely in a low-bandwidth way (e.g., just posting key graphics on Facebook or Twitter). The report has been divided into four Google docs to load faster. This is Part 4; the first part is here, the second part is here, and the third part is here. Additional materials (collected in this folder): Quantitative model: the Python notebook Biological anchor hypotheses for 2020 training computation requirements; a template spreadsheet When required computation may be affordable; and my best guess, conservative, and aggressive forecasts. Supplemental materials: a document containing various appendices; a folder of figures for the report; the spreadsheet Extrapolations of data and compute to train models; and the Python notebook Compute price trends, which draws on data in this folder. In Part 1, I provided an overview of the framework and estimates, provided definitions for key abstractions used in the model, and generated an estimate for the number of FLOP / subj sec of a transformative model. In Part 2, I reviewed theoretical and empirical evidence about training data requirements for a transformative model, introduced the concept of horizon length, and estimated how training data requirements may scale with parameter count for a transformative ML problem. In Part 3, I discussed in more detail the Neural Network hypotheses and other biological anchor hypotheses, and combined them into a 2020 training FLOP requirements distribution. In this part, I will: Explain my best guess, conservative, and aggressive forecasts for when the amount of computation required to train a transformative model may become affordable, incorporating forecasts for algorithmic progress, hardware prices, and spending on computation (more). Explain how I translate the outputs of this exercise into views on timelines, including a median estimate for TAI, probability of TAI in this century, and probability of TAI by 2036 (more). Address several high-level questions and objections to the framework (more). Briefly describe several open questions for further investigation (more). Timelines for when required computation is available Generating estimates for when the compute to train a transformative model may become affordable involves modeling three additional considerations besides the current compute requirements distribution: how compute requirements are likely to fall over time due to algorithmic progress, how the amount of computation available for a given price is likely to increase over time due to falling compute prices, and how the amount of money an AI project is willing to spend on compute to train a potentially transformative model would increase over time. This spreadsheet demonstrates how I model these three quantities to generate an estimate for when the amount of computation required to train a transformative model may become affordable. In this section, I will: Explain how the quantitative mode...