Will AI replace humans?
theoretical limit of being able to perform any human tasks - present or future -> humans become redundant and resort to leisure activities -> humans lose sense of self-worth without a job or meaning to life -> parts of humanity become insane and the population dwindles with possibly some self- inflicted extinction event. Although this is quite far-fetched and has only been touched upon in dystopian Sci-Fi movies, there are already existing indicators that we are giving up a larger proportion of control to AI. Another approach to the question is to look at the rate and feasibility of AI development and where this will lead us. We can then extrapolate the state of AI in the future and if it breaks the intelligence threshold. The first work recognized as AI was done by Warren McCulloch and Walter Pitts in 1943, and since then AI has grown at a rapid rate owing to an increase in funding, interest and results. Nevertheless, there have been setbacks such as the ‘AI winter’ in which many companies stagnated as they failed to deliver on extravagant promises. 6 Looking at present-day AI the current market for cognitive and artificial intelligence is $24 billion in revenue. 7 This shows that there is a demand for research into AI and that companies believe it is economically viable. If this trend continues it is likely that more automated and AI-driven systems will be developed (as there is no need to pay employees), resulting in the same consequence as outlined in the previous paragraph. It might also be useful to consider the potential of exponential advancement within the AI industry. It is quite probable that growth in AI won’t be linear and we already have examples of this: Google recently announced its AutoML project which is designed to create other AI more effectively than humans. 8 Along with this, increases in efficiency in engineering AI and computers will only shorten production time and bring us closer to artificial superintelligence (a god-like level of intelligence surpassing that of humans). The ability of AI to replace humans also relies on the ability of humans to write programs in which AI can self-learn over time (either through machine learning or other means) creating a feed-back loop in which the primary goal is blurred and instead the AI becomes intent on destroying humanity. This may be the case with reinforcement learning in which maximizing reward or some other parameter is the end goal. As is often hypothesized a poorly built algorithm could evolve into deciding that removing humans is the optimal way tomaximize X. Once theAI has reached this stage it may be unable to retrace its steps and so this is a feasible claim for human replacement. On the other hand, there is room for optimism: many computer scientists and visionaries see the potential of an AI revolution in which AI and humans collaborate, utilizing each other’s skills, with AI well regulated by humans. In this outcome humans will still have a place in society, providing creativity and dealing with the ethics (inherently what humans are good at) around AI. It is unclear how long this parallel relationship would continue but potentially long enough so that humans discover some higher form of intelligence, thus reclosing the gap between AI and human intelligence and avoiding AI domination.
6 Russel and Norvig 2010: 24. 7 Columbus 2019. 8 The Friendly Brain 2019.
121
Made with FlippingBook - Online catalogs