Will AI replace humans?
Joshua Soyke-Pinon
Artificial Intelligence (AI) is the pursuit for the best agent programme on a given architecture – an agent being something that can perceive its environment through sensors and consequently act upon that environment through actuators. 1 Note that there is no constraint on the AI having to be weak (the assertion that machines act as if they were intelligent, by simulating thought) or strong (that machines that do so are actually thinking). 2 In terms of the title, the idea of replacement is that in which AI replaces humans as the dominant form of intelligence on earth, either through extermination of the human race or replacing human’s societal role, rendering them useless. This need not be a physical upheaval of the human race through robots but a cyber-takeover. I have left the title quite open to allow for amultitude of factors and ideas to support and conflict with one another. The format of the question is also that of a yes/no but I will be supporting arguments on both sides. I have also not included a time- frame but ammore interested in AI over the next decade rather than some arbitrary time until the end of the universe. To summarize, I will be looking at the ethics, predictions, precautions and development surrounding artificial intelligence and its ability to replace humanity. In som e ways part of the question is already answered: the world’s most powerful supercomputer – IBM Summit, USA – has a memory of 2,801,664 GB and can perform calculations at a rate of 148.6 petaflops (10 17.2 operations per second). 3 In contrast the brain is estimated to have a memory ranging from 1- 2,500,000 GB and able to perform 10 17 operations per second. 4 (Ironically, the power requirement for Summit is 10,000kW compared to around 20W for the human brain indicating the brain is still much more efficient!) Therefore, modern supercomputers, capable of running AI programs, are already outperforming human brains in this metric. However, the catch lies in recreating the brain’s level of intelligence using a computer of this magnitude. Until this is understood, the continued increase in processor power as indicated by Moore’s law does not directly threaten humanity. One of the more prominent fears is not current levels of AI replacing humans, but the possibility of more advanced forms called AGI or artificial general intelligence. Most existing AI systems can be classified as ‘narrow AI’ in that they are trained to complete a specific task to a high level of accuracy and consistency but are unable to venture out of their comfort zone. The upcoming materialization of ‘general’ AI would be able to performmultiple tasks adapting in real -time to environments and be of a much better use to society – on a par with humans. Looking at the modern categories of AI: machine learning, natural language processing, expert systems, vision, speech, planning and robotics, 5 AGI would be able to enhance all these fields resulting in job losses and an increased reliance on AI for a wide range of tasks. Taking this further we can map out a potential path for humanity: AI reaches some
1 Russel and Norvig 2010: 34. 2 Russel and Norvig 2010: 102/ 3 Strohmaier and others 2019. 4 Russeland Norvig 2010: 12. 5 Kumar 2018.
120
Made with FlippingBook - Online catalogs