About

The AI Progress Essay Contest is closed. Thank you to everyone who contributed to this project! Contest results can be found here.

Are you new to Metaculus, forecasting, or Fortified Essays? Here's a short FAQ to get you started. Otherwise, read on to learn more about the AI Progress Essay Contest.

In the AI Progress Essay Contest, we open the floor for investigations of predictions regarding the future of AI, especially as it relates to the prospects, timing and impacts of potentially transformative advanced AI systems. Our aim is for systematic thinkers to engage with the wealth of AI forecasts that Metaculus has built up over the years.

Analyses we would like to see are those that synthesize Metaculus' AI forecasts, that draw out their key implications, and assemble these into a coherent view of the future of AI.

The Dreyfus prize is to be awarded to five authors of the best analyses of the future of AI. The prize is named after Hubert Dreyfus, an American philosopher, well-known for his incisive and prescient analyses of the prospects and limits of AI.

Some of the goals of the contest are as follows:

To encourage and support rigorous and nuanced thinking about the future of AI that engages with quantitative probabilistic forecasts
To Identify and reward the top thinkers and forecasters focused on the future of AI
To provide an arena for discussion and exchange of ideas regarding the future of AI, and especially the likelihood, timing, and impacts of transformative AI

Fortified Essay writers will have their work reviewed by a distinguished panel of judges:

Jack Clark is a co-founder of Anthropic, the co-chair of the AI Index, an expert member of the Global Partnership on Artificial Intelligence, the co-chair of the OECD's working group on classifying and defining AI systems, and a non-resident research fellow at the Center for Security and Emerging Technology. He also writes the well-known ImportAI, a weekly newsletter about artificial intelligence.

Owain Evans is Research Associate in Artificial Intelligence, University of Oxford. Owain has a broad interest in AI Safety and the future of AI. My current focus is truthful/honest AI (post, paper,). Owain is a research associate at the Future of Humanity Institute, Oxford University and got his PhD at MIT. He previously worked at Ought, and currently serves on Ought's Board of Directors.

Tamay Besiroglu is a Visiting Research Scientist at the Neil Thompson Lab at MIT where he works on the Economics of Computing and trends in AI and ML. He got his MPhil from the University of Cambridge. Tamay previously led Strategy and Operations for Metaculus.

Evaluation criteria

An expert panel of judges will be instructed to rate the entries by the degree to which they are relevant and interesting, as more specifically described below, with ⅔ weight given to essay quality, and ⅓ weight given to relevance.

Quality. High-quality submissions will be those that:

Original and Creative: the intellectual content of the essay must push forward understanding of the future of AI in a fresh way or with a new perspective
Synthesize at least some of the predictions made in the Forecasting AI Progress tournament, draw out their key implications, and assemble these into a coherent view of the future of AI
Demonstrate a solid understanding of both forecasting and relevant aspects of Artificial Intelligence and Machine Learning
Are well and clearly written, so that it is comprehensible and enjoyable to read
Are accessible to a diverse, well-educated but non-specialist audience, aiming in the range between the level of Scientific American and a review article in Science or Nature

Relevance. The theme for this contest is: Forecasting the medium and long-term future of AI. The contest asks participants to explore and elucidate some aspects of the future of AI, especially as it relates to the likelihood, timing, and impacts of Transformative AI. Moreover, analyses should draw on, amongst other things, the questions and predictions that formed part of the Forecasting AI Progress tournament. Additional predictions may be brought in also.

We're looking for short-to-medium length essays (somewhere between 700 and 2000 words).

Essay Topics

Here is a non-exhaustive list of topics that could make for worthwhile essays:

Will the scaling up of compute be sufficient to get human-level performance on a wide-range of important domains, such as language modelling and computer vision? If so, when can expect training runs to be sufficiently compute intensive to hit important milestones? If not, why not? Justify your views with, amongst other things, reference to quantitative predictions.
How much progress in AI algorithms and architectures has there been in the past decade, and how much should we expect in the next decade? Justify your views in reference to quantitative predictions, such as those involving performance on some common ML benchmarks.
Can we map the disagreements between notable figures in the AI risk community, as presented in the Late 2021 MIRI Conversations, onto disagreements about quantifiable predictions? Which views do we expect to imply predictions that we think are likely to fare better, and why?
When will AI-enabled automation constitute a negligible fraction of US GDP? Justify the view that this may or may not happen this century.
Is Deep Learning sufficient for achieving Transformative AI (roughly, AI that precipitates a transition comparable to the agricultural or industrial revolution), if not, how many new substantial insights might be needed?
How might the trajectory of the field of AI look like if Moore’s Law were seriously stunted in the next decade? How might the rate of progress be affected, and where would the field look for new sources of improvements?
What fraction of the field of AI will be dedicated to working on AI Safety, broadly defined? Will the field have substantial influence on how AI systems are designed, tested and/or deployed? Justify your views with, amongst other things, reference to quantitative predictions.

Prizes

A total of $6,500 will be awarded for the contest. The prize allocation by rank is as follows:

$2,250.00
$1,750.00
$1,250.00
$750.00
$500.00

With each author’s permission, all 5 winning essays will be published on the EA forum under Metaculus’s account. Metaculus will also make a post announcing the winners and linking to their essays on the Metaculus site, the EA Forum, and LessWrong.

Submission Rules

Submissions must be made through Metaculus’s Notebook feature between February 8, 2022 and April 16th, 2022 (click + at the top of this page).
Submissions must include at least 3 forecasts from the Related Forecasts. You can find them here.
An entry should differ substantially from any previously published piece by the author.
Submissions should not exceed 3000 words (excluding bibliography and footnotes, which are allotted another 500 words).
All Forecasting AI Progress rules that are stated here apply.

After an essay is submitted and accepted by moderators, it will appear below for anyone to read, though the author’s name will not be displayed. We encourage essay writers to consider and respond to each other’s work!

Click here for a video walking you through the process of drafting a Fortified Essay, using the editor's functionality, and submitting your essay to the AI Progress Essay Contest. Note that essays can be formatted in a text editor such as Google Docs and then pasted into the essay editor where they will retain formatting. Essays can also be written in markdown and pasted into the editor where they will be automatically formatted.