How Does AI Work? A Brief Introduction.
Well done for getting here.
Onto to Part 3. Much shorter too!
AI’s learning includes 3 components:
Dataset – This is the information you give the AI to learn from.
Algorithm – This is the tool designed to mimic some method of human-learning – deep learning, machine learning, transformers, and convolutional neural networks, to name a few. All AI systems will use an algorithm to learn from the dataset.
Defined Objective – This is what we tell the AI to seek. If we want it to beat a video game, we generally set the objective to maximize points.
So, in human terms:
- Dataset = our human experiences
- Algorithm = our brain structure and genetics
- Defined Objective = our desires and instincts You don’t program into AI what an image of a cat looks like.
You “show it” (AI), hundreds of images of dogs and it learns what characteristics separate a dog from a cat. Its objective is optimized when it guesses right via a human telling it which pictures are of a dog and which are not.
Once the AI is trained on the dogs and cats’ dataset, you can save this into a model.
Another AI can then load that model – yes, exactly like Neo in The Matrix learned kung fu – and instantly have all the knowledge that the first AI developed.
Continue to Part 4 where “Its Not The Tech – Its The People.”