Chipmaker Nvidia relies more on the production of tools for generative AI developers with the acquisition of the Greetel synthetic data company for more than $ 320 million, according to a Wired report Wednesday.
This decision comes as IA generating companies find it difficult to find enough data to train and improve their models, increasing the need to generate data.
According to the report, Gretel employees will be folded in Nvidia. Gretel, which produces synthetic or simulated data for the formation of the AI model, will strengthen Nvidia’s offers for AI developers.
A spokesperson for Nvidia refused to comment on the report.
Why is synthetic data important
The formation of generative AI models like OpenAi Chatgpt, a large language model, requires a lot of data. Real data can cause problems for AI developers – namely, it can be noisy and there is not enough.
AI companies come up against the limit of training data available to them for free, resulting in conflicts as to whether they can use content protected by copyright. Hundreds of actors, writers and directors submitted an open letter to the Office of Sciences and Policy of the Trump Administration to raise their concerns concerning the use of data protected by copyright. Currently, Openai requests the government to Authorize better access to material protected by copyright To train AI models, otherwise American companies will be left by China.
Look at this: Look at the NVIDIA GTC 2025 Keynote: all the strengths in 16 minutes
Synthetic data also has a value for the protection of private information. Gretel says his synthetic data Can be used to train models and tools without exposing sensitive or personal information – for example, health care data that does not identify individual people and potentially violate confidentiality laws.
There is concerns about the use of this data in the formation of models. Excessive dependence on information that is not actually rooted can increase the probability that a model is mistaken. If the problem becomes bad enough, it can cause a problem called collapse of the model, when the model becomes so inaccurate that it becomes useless.