Understanding artificial intelligence and generative artificial intelligence
Generative AI is at the center of all discussions. However, it can sometimes be challenging to understand AI clearly.
I. What is the difference between AI and Generative AI?
Did you know that the learning method used by DeepMind, which contributed to its fame, relies on the same principle as the artificial intelligence behind Tesla’s self-driving cars and represents the latest step in improving the ChatGPT model? This method is called Reinforcement Learning, a category of learning similar to supervised learning, unsupervised learning, or distillation learning.
Reinforcement Learning works with agents that repeatedly perform actions within a given environment. When an agent achieves its goal, it receives a reward. Based on the state, which is the representation of the environment, the AI learns from its interactions. It’s important to note that these agents are not the same as the agents used in Large Language Models (LLMs) like GPTs. In the case of LLMs, agents are tools designed to automate processes.
Not so simple, is it? However, understanding these distinctions is crucial to fully appreciating the capabilities and applications of different approaches in artificial intelligence.
II. Use cases
To clarify things, here is a brief table of use cases for generative AI and cases that do not fall under generative AI.
The table provides an overview of use cases by industry and their associated data formats. These use cases offer a clear representation of what each technology is used for.
At Giris, we analyze briefs in terms of technologies and architectures to determine whether it involves AI or GAI. Achieving a goal may require combining multiple technologies. The table includes use cases along with the recommended technologies and architectures for each objective. All the AIs listed have either been developed by Giris or are similar to those we have created, with the exception of autonomous vehicles.
Want to know which architecture is best suited to achieve your goals? Feel free to get in touch!
Tesla also uses annotated data to train their models, involving supervised learning, and has explored techniques such as imitation learning and semi-supervised learning.