A training session is organizing the difference between traditional machine learning, which performs prediction and classification, and generative AI. Compared with traditional machine learning, which is the MOST fundamental difference that characterizes generative AI? Choose by focusing on the nature of the output.
Choose the fundamental difference between traditional ML and generative AI.
It generates new content such as text, images, and code
Correct. Whereas traditional machine learning mainly performs prediction and classification (producing an answer from existing options), the fundamental difference is that generative AI generates new content such as text, images, and code.
It can be trained without using any labeled data at all
Pretraining of generative AI is mainly done with unlabeled data, but fine-tuning and RLHF also use labeled data.
In addition, the training method is an insufficient description of the fundamental difference, which is the generation of new content, so this is incorrect.
It runs with less computation than traditional ML
Foundation models are huge, and their computation is generally larger than traditional ML, not smaller.
This is factually wrong, so it is incorrect.
The basis of a prediction can be fully explained
Large generative AI is internally complex, and explainability is actually a challenge.
This describes the property in reverse, so it is incorrect.
Whereas traditional machine learning mainly performs prediction and classification (producing an answer from existing options or labels), the fundamental difference is that generative AI generates new content such as text, images, and code. 'It uses no data,' 'it does no computation,' and 'it always uses less power' are all factually wrong and are not the fundamental difference either.