2 advantages of Evolutionary Algorithms compared to other types of Generative AI
Generative AI is usually used to describe very advanced systems used to produce text, images and other intellectual work and art. They typically do this by combining vast data collected from the internet and indirectly from human knowledge, preferences and creations stored there. ChatGPT is likely the best known example right now and most of us have seen it or heard about it. To sum it up, these Generative AIs use existing human creations and behavior to extrapolate what is asked of it; without such data they are powerless and they don't create anything entirely original.
Let's now look at Evolutionary Algorithms (EAs), which can be considered a niche type of generative AI. They do indeed generate new solutions such as industrial designs or plans; they are not currently used for art but that is mostly due to their novelty. They strictly speaking do not belong to AI (since they do not emulate decision-making and neurons) but few people besides academics care about that. What EAs actually do however is create solutions from the scratch: historical data can be used to understand the business case and to evaluate the quality of the solution, but in the end this data is not used anywhere in the solution itself. This has 2 significant implications:
- Evolutionary Algorithms can create new solutions never seen before and not derived in any way from current knowledge, and
- The created solutions should be unburdened by any intellectual and copyright issues.
Thanks to these differentiators, for the use cases where Evolutionary Algorithms are a practical approach, they will present a welcome alternative to other types of generative AIs.
Milos & Ladislav
Center for Genetic Programming
#generativeai #evolutionaryalgorithms #geneticprogramming