Act like it is science

The author expresses frustration with the numerous doomsday predictions surrounding the recent OAI/HuggingFace incident and related stories. However, the author finds these concerns to be largely underargued. The author then offers a request to those who are worried and wish to persuade skeptics to employ the methods of science. The author proposes three specific actions to demonstrate the potential cybersecurity costs resulting from AI break-ins.

Firstly, the author asks for an outline of how, in two to three years, one might estimate the additional cybersecurity costs stemming from AI break-ins. While the author concedes that these costs may not be zero, they would like the proposed method to be clear and understandable. The author suggests that if available, an estimate of past cybersecurity costs, compiled by top economists, could serve as a reference point.

Secondly, the author requests a current numerical estimate of the potential cybersecurity costs that could arise from AI break-ins. The author also states that these estimates should be subject to testing against actual events over time. The author emphasizes that the exercise is not a substitute for preferred safety measures but rather a way to provide specific predictions and contribute to a more scientific discourse.

Lastly, the author requests a list of currently held short positions in stocks or other assets. The author suggests that if the estimated costs are sufficiently low, the response could be a zero, similar to the author's own approach. The author believes that the costs, while present, are not so high that they cannot be managed and still yield positive stock returns. The author views this exercise as worthwhile in order to enhance the scientific nature of the ongoing conversation and reduce the focus on personal anxiety.