Charlie . 11th Jan, 2025 12:13 PM
Does AI I use Human Created or Synthetic Data?
Excited to blog about human crated and synthetic data. Why, because I think both types data are important to the progression of artificial intelligence. Some are saying that all of human created data has been used to train LLMs. Now, companies are using synthetic data to do the training.
When I here we have used all of the human created data to train LLMs that is available. I am like that is not the case because when a LLM releases a new version for this blog purpose lets use Jan 17th 2025. The next time a LLM release a updated version will be Aug 8th 2025. The human data created between 1/17/25 and 8/8/ 25 can be used to train the LLM. My thought is the data is not large enough to increase the efficacy of the model. This is where synthetic data fills the gap.
Everyday we as in humans are creating data. At the big LLMs level they need a lot of data to train models on before they release another version. In my opinion this makes sense economically. Also, I wonder what happens to a model when its fed majority synthetic data between two or more versions. Will LLM models start giving us diminishing returns? And, yes AI uses both human created and synthetic data to train. Thanks for your time. Power-Up!
**disclaimer always do your own research on the information Help My Business Revenue Consulting Group is providing information. We do not endorse information accuracy or maintained links**
