The question of when the machines surpass human intellect is discussed for more than one decade. However, with the development of machine learning and program algorithms that can easily undergo turing, this issue is becoming more and more relevant. How long does it remain until the creation of artificial general intelligence (Agi) and what risks can arise in the process?
Contemporary artificial intelligence (AI) is mainly based on large language models (LLM), and textual AI do not really think about their answer and do not conduct research-they operate with probabilities and statistics. Using the teaching data, they calculate the most probable word (sometimes the letter), which should follow the previous. This can give reasonable results, but also leads to errors and even dangerous conclusions, with rare funny answers , which a person would never give. These cars do not think.
These models specialize in the implementation of narrow tasks requiring specific training data. However, the possibility of AGI is already actively discussing in this area. These algorithms will be able to perform many tasks at the human level. And although the artificial consciousness can be still Far , the development of Agi is considered as an important stage. Some sectors argue that only a few years .
“In the next ten or two decades, probably a separate computer will reach the computing power of the human brain by 2029 or 2030. And after another 10-15 years, one computer can have the computational power of the whole society,” said Ben Gertzel, founder Singularitynet, which seeks to create “decentralized, democratic and inclusive agi” at the conference Beneficial Agi Summit 2024. Details about this < a href=”https://youtu.be/cw4jkuof9fe?t=652″ Target = “_ Blank”> in his speech .
Two important questions arise from this. The first – how accurate are these grades? Critics argue that talking about imminent agi is a way to warm up interest in the current technologies and explosive growth their popularity before an inevitable decline occurs. Recently Nobel Prize laureate and one of the founders of AI Jeffrey Hinton believes that less than 20 years remain before the appearance of Agi. Yoshua Benjio, who in 2018 divided the Turing Prize with Hinton, claims that The exact deadlines cannot be predicted