Hello Reddit community
I am hearing so much about these terms around AI, but don’t understand the difference. I understand the basic premise of machine learning with fitting the line.
But can someone please explain the different between these two terms?
In: Technology
A predictive AI predicts things like weather or traffic.
A generative AI generates new content, like images or text.
You could phrase it in a way that mixes them; “can you predict what a kangaroo cooking pizza would look like?”
And some use cases are bit of a combination, eg AI rendering a red circle around a fracture in a X-ray.
A possible difference is in how they are trained. For predictive AI, you usually feed past data with a known outcome, and then the network is modified with what is called backpropagation to change the network so that its output matches more closely with the known correct answer.
Generative AI is usually taught unsupervised, meaning that it is fed data without knowing what the output should be. There is usually also a supervised learning step, for example, the AI might generate two outputs and humans then label one as better than the other.
Especially in image generation, an adversial network can be used in training. Here, the adversial network tries to determine how good the output of the generative network is, and the two networks essentially compete.
But, lot of techniques can really be shared between predictive and generative AI, so it’s more about their purpose really.
Latest Answers