Submitted by random_encounters42 t3_111ogdu in Futurology
I think ChatGPT proves how capable neural network AIs can be and by combining enough specialised trained AIs, conscious artificial intelligence will emerge in the next 5 years.
Neural network AIs mimic how our neurons and our neural networks function in the brain. We can see that these networks, with enough training, are capable of assessing complex inputs and produce corresponding outputs. That's what intelligence is at the fundamental level. And complex intelligence is just the combination of many rudimentary intelligent systems working together.
The key is that the human brain has many regions each in charge of specific functions whereas chatgpt represents only two primary functions - memory and speech. Once other specialised AIs in charge of other functions - like critical thinking, complex analysis, emotions, etc - are created and integrated together, we'll see a truly self-conscious AI.
There are many companies already working on these systems. We know that chatgpt has already passed the medical licensing exam and the Bar, something that the majority of people can't.
I honestly think ChatGPT is a major stepping stone to truly conscious AI. It is actually intelligent, but just at a lower level because it's missing some key functions. Additionally, AI doesn't have to be right 100% of the time and super smart, it just needs to beat the average person.
​
Edit 1:
Thanks for the replies. I apologise for the title, it is a bit of a clickbait. I just find this topic so fascinating. I watched this video explaining how chatGPT works and I think it does a really good job of it. If other people have suggestions for videos, I'd love to take a look.
My original point was in 5 years, we will have an AI that will seem intelligent and conscious, and practically speaking, people will treat it as such.
What I find very interesting is that researchers trained chatgpt like how you would train a baby. First with unsupervised data input and then with supervised data input. It is actually made up of 2 sets of neural networks with one that generates context and rudimentary understanding of the input data (unsupervised). And then this info is fed into the second neural network with supervised feedback. This is why the responses are so human like.
I think people who say it's just a predictive text algorithm are oversimplifying a monumental achievement.
ChatGPT is an AI, based on complex neural networks which mimics how neurons function in our brain. It uses neural network structures inspired by those found in our brain, and it's trained like a human baby.
Because of this, I actually think it does have some type of understanding of the context and meaning behind the text inputs. It’s probably different to how we would perceive information but given the results, I think some comprehension is there.
​