With the rise of AI, we, as a society, need to develop a solid awareness over what is truly human and what is merely a calculated probability.
Ever thought about, what is the difference between dreaming and the output that ChatGPT produces?
Thoughts! Here is a train of thought that might shed some light on it.
A generative pretrained transformer like ChatGPT uses probability to suggest the next best word or word group based on its highly trained memory, which is loaded with billions of sentences, dialogues and conversations. The outcome is a new combination of words, in its form semantically and gramatically correct, suitable to appear of making sense and being an adequate response to the request sent. The outcome is nothing like random at all. Rather, it is a distinct selection, one that has been composed based on similar past examples of responses to similar requests, . It may even take into account the short term history of requests and responses that preceeded the current request, to calculate an even better response in the full context of the current conversation.
Still the resulting response is nothing else, but a list of word with a calculated probability of being the best answer possible
When you are dreaming, something similar happens. Whatever has been going on in your head that day, will be taken as a prompt. Maybe you have been worrying about a new job. Maybe you are excited to be going on vacation. Maybe you are thinking about how to best code a SwiftUI program. Whatever is going on in your head, especially before you go to sleep, will be stimulating the content of your dreams. It has been said, that when you are preparing for an exam, reading the necessary input before finally falling asleep, might increase the likelihood of remembering necessary information. But a dream is more than just a single answer. Rather, a dream develops into a story or several stories. The interesting part of it is, that a dream can develop in a rather creative, sometimes unrealistic, way. Dreams can appear to be creating developments that never happened, with things that don’t exist or happening that aren’t physically possible. Sometimes, on the other hand, dreams can create stories that one has been wishing for, such as a customer signing up, a tension between partners being resolved or a wedding party happening with all the desired elements.
Generative pre-trained transformer can be creative too, of course. Prompts can be designed in a way, that the generated response adopts a certain style. One could instruct the system to act as a scientist, psychologist or a creative artist. It is also possible to ask specifically for something to be invented, constructed or created out of the blue. One could even ask for a storyline based on a single thought or a complex aggregate of thoughts. In essence, it is actually possible to ask for a dream. So what is the difference between a dream and the output of a generative pre-trained transformer?
Whether there is a big difference in the outcome, is something that each end everyone shall decide for him or herself. The outcome may actually appear comparable. Except for one thing. Humans do something that a generative pretrained transformer does not: they think, even in their dreams. But what does that mean?
Thinking is the process of considering or reasoning about something, using thought or rational judgement. In psychology, thinking is also known as ‘cognition’, and refers to the ability to process information, hold attention, store and retrieve memories and select appropriate responses and actions. Thinking has also been described as hinking as indirect reflection about characteristics, attributes and interrelations between things.But most importantly, thinking involves input from our senses, including the sense of thought, by which you observe the views, considerations and questions that you and others have, and thus get an idea of what they are thinking, you are thinking and what the difference is. As such, thinking is the outcome of observing, analysing, generalising, abstracting, prioritising, retrieving, computing, calculating, combining and many other activities that “we can do in our mind, by forming thoughts that build on each other”.
This definition does not claim to be complete or scientifically correct. But what it does want to show is, that thinking is more than just finding the most probable next word or word group.
A thought is the outcome not just of what we rationalise, but also about what our emotions, our nervous system, our gut feeling has to say about something. It has been stated there are many more neurons in many parts of the body, not just in our brain. So far we have attributed the thinking process to the brain, but really, the brain listens to a lot more than what is going on in the brain. The phrase ‘Gut Feeling’ is used to describe an intuitive feeling or response to something, an immediate physical response you feel that culminates in a cognition of type ‘suggestion’ for the best decision when presented with two or more choices. This is mostly based on our central nervous system, which operates based on biochemistry, involving signals, transmitters and states.
This kind of thinking does often involve estimating a probability, that’s true, but most of the time it evaluates feelings and follows an intuition. Now, that intuition may not always lead to the best result. But that is the thing: as humans, we understand that we aren’t perfect and that we may learn by making mistakes.
With this knowledge in mind, when we ask ourselves whether dreaming is more than what ChatGPT does, the answer is very clear. ChatGPT does not think and is far from producing the same result as humans who think or dream.
In order for an artificial intelligence to think like humans, it would have to learn to make mistakes, be intuitive and respond to feelings. It would even have to reflect about the meaning it has in the presence of everything around it. In order to be aware of such meaning, consciousness about it-/oneself is required. AIs are far from developing this, in a way humans do.
However, AIs may be very well capable of mimicking the responses, behaviours and even train of thoughts that appear like thinking – because they can calculate what a human expects to hear, see, feel, smell, taste or reason in his or her thoughts.
So while we may not see AIs producing true human dreams or AIs that can think like humans, we may still soon see even more fakes of what humans like.
Learn more about Data-Centric AI and Controlled Application Spaces. https://lakeside-analytics.com/data-centric-ai/