I’m an newbie to AI and started my learning path with “Machine Learning Specialization” course, I remember at the beginning of the course, it has said that machine learning is that the machine can learn itself and perform without being explicitly programmed, but after several weeks of learning, I feel that all of the popular AI terms(Neural network, ChatGPT, etc…) are running with a pre-trained model, so I think these AIs still haven’t achieved self-learning yet, the model determines their behavior so they are still running with being programmed.
Yeah I know that the training is the learning process, but since training is actually intervened by human so it’s still not self-learning right?
1 Like
The key concept is that a machine learning system takes in observations and known outputs and creates a model of their relationship. This allows you to use the model to make predictions on new data.
You don’t have to program this specific relationship yourself by creating individual rules.
2 Likes
yeah that makes sense, I also have rethought it and was suddenly human brain works in a similar way too, human brain learns and forms a “model”, when new problems coming the brain uses its “model” to solve the problem.
One difference is that the brain is learning(training) its model all the time, while AI gets trained less frequently, like ChatGPT3.5’s train data stops at Sep 2021.
1 Like
It means the machine is learning from what, from data being fed. That data is always of the past, future is impossible, present is also impossible, so “it has read a book and learned something from it”.
What do you mean with “self” ?
The only one that can guaranty the existence of self is you; talking to your-self.
The rest is only illusion, a side effect of your neural machinery.