Artificial intelligence as intelligent
characters back out under the door. This then leads to those outside to mistakenly suppose that there is a Chinese speaker in the room. Ultimately, the conclusion of this argument is that computers merely use syntactic rules to manipulate symbol strings but have no deeper understanding of meaning or semantics. As a result, not only does this display the inadequacy of the Turing Test, but it also displays there is no practical way to truly prove whether or not an AI is truly intelligent, or whether it is simply feigning intelligence. Despite this, are there any theoretical methods that could result in AI developing consciousness? At least in the human context, consciousness must have developed in brains through millions of years of random evolutionary mechanisms. A strong AI algorithm could potentially evolve by simulating random mutations in itself. But that reveals a deeper problem: computers are incapable of true randomness, 8 as they’re inherently deterministic, meaning if you ask the same question you’ll receive the same answer. The m achine can be programmed to generate ‘random’ numbers (known as ‘pseudo - random numbers’), but as it is ruled by a carefully defined and consistently repeated algorithm, the numbers it produces are not truly random. While genetic algorithms can still be used, the evolutionary mechanisms simulations will not be truly random, and therefore it is effectively useless for attempting to mimic the development of human consciousness. Genuine evolutionary change cannot be replicated without true randomness, and as a machine can never access real random numbers to do so, it is unlikely that a strong AI algorithm will ever develop through this method. It is worth mentioning, however, that research has shown that the brain uses a number of complex and deterministic algorithms for various unconscious operations, such as breathing or blinking. 9 This therefore implies that part of our brain is, or used to be, acting like a computer programme on our biological hardware. The step between non-conscious computer to conscious brain could be hidden in these random evolutionary changes. This could have granted us a way of ‘looking outside’ existing dead-ends in problems, a form of meta-concept, or self-awareness. From this, our thinking eventually became more sophisticated, and we eventually reached our current level of intelligence. If computers ever became able to generate truly random numbers, then the evolution simulations could provide a breakthrough in developing the first strong AI. In conclusion, AI cannot be considered intelligent at this point in time, and it is incredibly unlikely that any will achieve intelligence in the coming decades. While weak AI is useful for the specific tasks that each algorithm is designed for, it is unable to think or be conscious, and therefore cannot be considered intelligent. Similarly, machine learning, and more importantly deep learning, is capable of learning and applying what it has learnt, but the knowledge cannot be applied to tasks outside the AI’s function. The best chance for a strong AI to be invented is by simulating evolution to result in artificial consciousness: however, this is limited by the fact that no current computer has the ability to generate genuine random numbers, and therefore cannot simulate random mutations. As long as this remains a fact, it is extremely unlikely that any AI will ever achieve consciousness, and therefore no AI will never become truly intelligent.
8 Rubin 2022. 9 Contente 2021.
46
Made with FlippingBook - Online catalogs