I just started to learn about AI, I have a quick question.
how do you teach a machine? how do a machine learn?
Since Machine does not have a brain, the machine is not really learning, is there another way to explain what the machine is doing, without using the word, “learn”?
Just curious, trying to understand better what machine learning is really all about.
1 Like
“Learning” here means creating a mathematical model that, given a set of examples, can be used to make predictions for future examples.
3 Likes
I would say more like memory patterns than “learning”, it memorizes patterns and rules, and can combine patterns based on the rules it memorized 
4 Likes
Thank you for the answer, I really appreciated.
I understand the idea of memory patterns, but I am still confused with “Memorization and rules”.
How does the machine memorize the patterns and rules and combine the patterns based on the rules it memorized?
I just started to learn; I have a long way to go to fully understand it.
if you had to replace the word memorization, which word would you use?
Thank you for your answer, I appreciated.
If I understand correctly, based on the mathematical model the machine used the set of examples to make predictions for future examples.
I just start learning python, I am taking the AI python short course.
when you write the code print (“Hello World”), you see Hello world on the screen.
Correct me if I am wrong, under the hood there still programming involves training the machine on how to use the set of examples to make predictions.
I don’t think “rules” is a good description.
The model is a mathematical relationship between the inputs (the examples) and the outputs (the truth values for each example).
The model consists of weight values for each feature, and a bias value for each decision.
In the simplest form, it’s essentially the same equation for a straight line from algebra (where the ‘weight’ is the slope and the ‘bias’ is the y-intercept, ‘x’ is the feature value, and ‘y’ is the truth value.
In algebra, it’s:
y = m*x + b
We’re essentially learning ‘m’ and ‘b’.
For more complex models the relationship gets more complicated than a simple linear approximation, but the behavior is still controlled by the weight and bias values.
1 Like
Thank you again for the answer
Well, at this point you should start going through our specializations here or some other source, and see for your yourself how this works!
1 Like
Agree,
Thanks for taking the time to answer.
Rather than ‘learn’, I like to think of it as ‘pattern fitting’. We are fitting a pattern to a formula on information. It is only the marketers (or those that do not quite understand) that do or want to try and convince us this conforms to ‘knowledge’ or ‘learning’.
I think this can also easy be seen in that otherwise, ‘learning’ consists of being able to take a second or third additional action (or, basically inference beyond the original scope of what it was trained on). One cannot do such a thing with the models we are presently building without additional intervention (i.e. reinforcement learning, etc).
2 Likes
The machine Learning specialization is on coursera.org, is it possible to give me some resources to learn about Machine Learning? I like Andrew Ng teaching style, but unable to afford coursera.org what now.
Thank you for your answer.
Francois Chollet book on Keras came into my mind and also Prof Andrew there is a free book on AI, try to search in google. Also coursera can offer financial aid, you have to apply for it!
4 Likes
great,
if I have more questions while learning I will ask.
I am glad i find that website.
Thanks
1 Like
Is Ai-900 microsoft azure certification worth it to start with AI?
What about IT specialist AI certification?
My library hold finally came in. If, later, I can fill in more details I will.