Artificial intelligence is not just about destruction

Maybe because it sells, but artificial intelligence is not just about catastrophes. Long before ChatGPT became famous, people already used AI. For instance, the face recognition on your smartphone is AI, and it is the latest advance, it was a revolution. When you use a noise reduction, say on your computer, this is AI. It is much more than the public is discussing. Even though it is good that after ChatGPT, people started to realize AI exists, it also created a biased discussion. Not all problems from chatGPT can be migrated to other models. Large Language Models have their own issues, as so have computer vision, as example.

I am going to mention two examples, that I know closely, where AI is unique. They create what can be called transfer learning (this is not the technique called transfer learning, I need a better name). Basically, different from humans, once AI is trained, it can be shared. chatGPT is a pretrained model, that is, it knowledge is shared easily, of course, for fee. Still, it is cheaper than human experts. That is why Sam Altman said on an interview that in the future, intelligence will drop drastically.

The first example is from my paper SnakeFace: a transfer learning based app for snake classification. As I was studying snakes, the model learnt much faster than I did. My job was to gather annotation images, and train the model, which learnt in seconds.

The second example is from my paper Robodoc: a conversational-AI based app for medical conversations. Basically, I have trained a model to detect pneumonia, I am not a medical doctor (I am a scientific doctor though!).

Those are example where knowledge is extracted from annotation images, without the programmer actually have to know much about the images. This is something to celebrate that those models can actually be used to accelerate human intelligence.