Eliezer Shlomo Yudkovsky is an American specialist in artificial intelligence, who studies the problems of technological singularity and advocates the creation of Friendly AI. He is the author of several science fiction stories in which he illustrates some topics related to cognitive science and rationality.
Eliezer Yudkowski: biography
Eliezer Shlomo Yudkovsky is one of the most active rationalists of modern America, a computer researcher and popularizer of the idea of "friendly artificial intelligence."
Born September 11, 1979. Co-founder and researcher at the non-governmental research organization Institute for Research on Machine Intelligence, which publishes his books. Eliezer Yudkovsky is the author of the philosophical novel Harry Potter and the Methods of Rationalism, published in parts on the Internet between 2010 and 2015. In it, Harry grows up in the family of an Oxford scientist and, before traveling to Hogwarts, learns from him the methods of rational thinking.
As far as is known, it has no connection with the circle of “speculative realists”. His name is often mentioned in connection with Robin Hanson, for several years (from 2006 to 2009) they were the two main authors of the blog "Overcoming the Prejudice", which existed on the money of the Institute for Future Humanity in Oxford.
Sometimes the name of Yudkovsky can be heard in connection with Raymond Kurzweil. He is a philosophizing representative of the techno-geek community, thus inaccessible and incomprehensible to most of the humanitarian public, to which almost all readers of speculative realists belong. It is not surprising, because he often turns to the logical and mathematical language, probabilistic and statistical conclusions, exposed using formulas and tables.
Scientific interests
Yudkovsky is a co-founder and researcher at the Singularity Institute for Artificial Intelligence Institute (SIAI). He made a great contribution to the development of the institute. He is the author of the book "Creating Friendly AI" (2001), articles "Levels of Organization in General Intelligence" (2002), "Coherent Extrapolated Volition" ("Coherent Extrapolated Volition", 2004) and The Timeless Decision Theory (2010). His latest scientific publications are two articles in the collection “Risks of a global catastrophe” (2008) edited by Nick Bostrom, namely, “Artificial Intelligence as a Positive and Negative Global Risk Factor” and “Cognitive Distortion in Assessing Global Risks”. Yudkovsky did not study at universities and is an auto-didact without formal education in the field of AI.
Yudkovsky explores AI designs that are capable of self-understanding, self-modification, and recursive self-improvement (Seed AI), as well as AI architectures that will have a stable and positive motivation structure (Friendly Artificial Intelligence). In addition to his research work, Yudkovsky is known for his explanations of complex models in a non-academic language, accessible to a wide circle of readers, for example, see his article “An Intuitive Explanation of Bayes' Theorem”.
Yudkovsky, along with Robin Hanson, was one of the main authors of the Overcoming Bias blog (overcoming prejudice). In early 2009, he participated in the organization of the blog Less Wrong, aimed at "developing human rationality and overcoming cognitive distortions." After that, Overcoming Bias became Hanson's personal blog. The material presented on these blogs was organized as chains of posts that attracted thousands of readers - see, for example, the entertainment theory chain.
Yudkovsky is the author of several science fiction stories in which he illustrates some topics related to cognitive science and rationality.
Career
- 2000. Yudkovsky founds the Singularity Institute (later renamed MIRI).
- 2006. Yudkovsky joins the Overcoming Bias collective blog and begins writing texts that later became Chains.
- 2009. Based on Yudkovsky’s entries on Overcoming Bias, the famous collective blog LessWrong.com is created.
- 2010-2015. Yudkovsky writes GPiMRM.
- 2013. Yudkovsky publishes the latest posts on LessWrong.com and stops writing on the site. As of 2017, Yudkovsky publishes most of the publicly available texts on Facebook and Arbital
Thought experiment
Imagine an incredibly smart artificial superintelligence locked in a virtual world - say, just in a snuffbox. You do not know if he will be mean, friendly or neutral. All you know is that he wants to get out of the box and that you can interact with him through a text interface. If AI is really superintelligent, can you talk to him for five hours and not succumb to his persuasion and manipulation - not to open the snuffbox?
This thought experiment was proposed by Eliezer Yudkowsky, a researcher at the Engineering Research Institute of Engineering (MIRI). MIRI has many scientists who are exploring the risks of developing artificial superintelligence; although it hasn’t even appeared yet, it is already attracting attention and fueling a debate.
Yudkovsky argues that artificial superintelligence can say everything it can to convince you: careful reasoning, threats, deceit, building rapport, subconscious suggestion and so on. With the speed of light, the AI builds a plot, probes weaknesses and determines how it is most easy to convince you. As the existential threat theorist Nick Bostrom put it, "we must believe that superintelligence can achieve everything that it sets as its goal."
The snuffbox AI experiment casts doubt on our ability to control what we could create. It also makes us comprehend the rather bizarre possibilities of what we do not know about our own reality.
