Artificial Intelligence: Recent Advancements and Implications

    Artificial Intelligence (AI) is a field that combines computer science and robust data sets to enable problem-solving (IBM). It is the science and engineering of making intelligent machines. These problem-solving models can be used for a wide variety of tasks. These tasks include speech recognition, customer service, computer vision, recommendation engines, and automated stock trading (IBM). These are just as of now, and the wide range of uses seems to keep on growing as time goes on. 

    In 1950, Alan Turing published Computing Machinery and Intelligence. A paper is widely known as the start of the idea of artificial intelligence. Turing became famous for being the person to decode the Nazi’s ENIGMA code; he proposed the question “can machines think?” The paper goes into depth on this new test Turing has created, now known as the “Turing Test.” The Turing test is essentially a test of AI intelligence; it says that if a computer can convince a human that it is not a computer, and believe that they are speaking to a human, the computer passes the Turing test, and therefore the computer must have some form of intelligence. The validity of this test has been debated ever since. 

    In his talk, Dr. Vukov explained recent developments and thoughts on whether or not Artificial Intelligence is truly conscious. He explained the Turing test, different philosophical perspectives, the Chinese Room, and weak AI versus strong AI. Weak AI, also known as Narrow AI, is most of the AI that surrounds us today. Examples of Weak AI are Amazon’s Alexa, IBM’s Watson, and automated driving vehicles. These interfaces have a clear input and output and are not directed by anything except for user input. Strong AI is a theoretical form of intelligence in which a machine would have intelligence comparative or equal to humans (IBM); it would have self-aware consciousness that can solve problems and think of the future. Following this, Dr. Vukov gave us the example of the Chinese Room, which holds the argument that current Artificial Intelligence is not conscious.

    The Chinese Room is a comprehensive test of whether or not something has a true understanding of what it is doing. In this example and story a man or woman is placed in a room with symbols placed across the room. They are then instructed to put together a series of symbols. Turns out these symbols are Chinese characters, and the person made a real sentence that makes sense to a fluent speaker. However, this person was not fluent and did not know what they were doing, but it made sense. This argument translates to Artificial Intelligence such as ChatGPT not fully understanding what it is saying, or the true context that it holds, but it makes sense to us humans. This brings to question and more or less shows that ChatGPT and other forms of current artificial intelligence are not Strong AI, and they are more or less ‘fancy machines’ that will do as told. 

    Recently, researchers at Amazon have trained a new Large Language Model (LLM) for text-to-speech that they are claiming exhibits “emergent abilities,” (AI News). They call it ‘Base TTS’, it is a 980-million parameter model and is the largest text-to-speech model created yet. Text-to-speech models are used in the development of voice assistants such as the Weak AI models previously mentioned. What sets BaseTTS apart isn’t just its amount of parameters, but the extensive training it has received. BaseTTS was trained on over one hundred thousand hours of recorded speech from public sites across many languages, allowing it to navigate pronunciations and translations with ease. Amazon’s researchers found that when their model had upwards of one hundred fifty thousand parameters, it suddenly became increasingly better at understanding and speaking language. They claim that “it can use complex words, show emotions, use punctuation correctly, and cleverly ask questions,” (Sejal, 2024). 

    Now this brings some questions to the forefront… Regarding Dr. Vukov’s talk, a sentient AI would have to be able to not only show emotions but also understand them and give a response that makes sense to the system internally, such as we humans do. Although BaseTTS does not do this (making it not sentient nor any other AI) and holding the Chinese Room to be true; it does bring us other factors into the conversation such as the expression of emotion. Not only does it make us revisit the conversation, but it brings us one step closer to AI completely fooling us that it is indeed another human.

References

   Sharma, Sejal. “Amazon Develops World’s Largest Text-to-Speech Model with ‘emergent’ Qualities.” BASE TTS: Amazon’s Largest Ever Text-to-Speech Model Promises Natural Speech, Interesting Engineering, 19 Feb. 2024, interestingengineering.com/innovation/amazon-develops-worlds-largest-text-to-speech-model.

    

    “What Is Sentient Ai?” Built In, builtin.com/artificial-intelligence/sentient-ai. Accessed 19 Feb. 2024.

    “What Is Artificial Intelligence (AI) ?” IBM, www.ibm.com/topics/artificial-intelligence. Accessed 19 Feb. 2024.