My project is an expansion of an existing website, where I will create a comprehensive introduction to the risks of AI. My site, called “Humanity Tomorrow,” (https://humanitytomorrow.site/) is also intended to become a community or think tank in the future. At the moment, it features an extensive FAQ on the movement for radical life extension. But from the very beginning, I conceived this project as something covering several topics at once, and one of them is the existential risks posed by AI.
I am currently working on a section that will address as many questions and objections related to AI risks as possible. There are a great many such objections, and people sometimes cling to a particular belief or set of beliefs that prevent them from objectively assessing AI risks. At the most basic level of understanding, this includes statements such as “AI has no consciousness,” “AI is just a next-word predictor,” and so on. Going a bit deeper: “You can’t get rid of hallucinations, so there won’t be a superintelligence,” “We won’t have enough energy/processors/memory,” and so on. Over time, I’ve been able to change the minds of ML engineers in private conversations who held such beliefs, leading them to become more level-headed about AI safety. Now I want to extend this to a wider audience.
It seems to me that the field lacks a single, comprehensive, and down-to-earth introduction to the topic of AI risks—one free of rationalist jargon, complex calculations, and so on. I myself lacked this at the very beginning of my journey. It would also be an excellent introduction for people from Eastern Europe—we have many talented programmers and engineers working in AI who sometimes simply aren’t aware of the existence of existential AI risks.
As part of my work, I study analyses of these issues—for example, those by Yoshua Bengio or MIRI—with the goal of compiling the very best insights into a single, comprehensive resource. The site will include not only an extensive FAQ, but also specific recommendations for action, a map of the field, and, presumably, a particularly large graph illustrating which factors must align to result in a catastrophe induced by AI or the malicious use of AI. This section is intended to be as comprehensive and useful as possible—serving as an introduction for beginners and providing structured knowledge on the topic for more advanced readers.
I am working on this project alone, with the exception of discussions and feedback from people working in AI or who have long been interested in the topic, rationality, and LessWrong. I have experience explaining complex facts in simple terms and engaging with a broad audience through my YouTube channel, where I’ve made videos about catastrophic risks from AI for a Ukrainian audience. The website, however, supports several languages, the primary one being English. The others are Russian and Ukrainian; I translate them myself since I’m fluent in both.
Unfortunately, due to my main job, I don’t have enough time or resources to develop this website at the pace and on the scale I would like. If I could secure a small amount of funding, the money would go toward creating an AI section, advertising the site, and covering its technical maintenance. I think I’ll be able to finish creating the section on AI risks in 4–5 months, and I’m requesting funding specifically for that timeframe. I also plan to commission or record audio versions of the articles myself.
This site will help beginners in the field of AI risks understand the situation and get started, and it will help those already working in machine learning to reevaluate their perspectives. I know from my own experience how important it is to have access to such content; if it weren’t for the articles by Kokotaylo, Ashenbrenner, and Yudkowsky, I might never have entered the field of AI existential risks.