Mark Zuckerberg, CEO of Meta has a searing take on Big Tech’s race for AI training data: It’s not about the data.
“What I think will be more valuable are feedback loops rather than any kind of initial corpus,” Zuckerberg said in a statement. interview with the command linea newsletter about the technology industry.
Feedback loops are used to retrain and improve AI models over time based on their previous results. For example, these algorithms allow AI models to know when they are making an error and provide them with data to adjust their future performance.
“The fact that a lot of people are using it, and then seeing how people are using it and being able to improve from there will actually be a more differentiating thing over time,” he said.
Finding new data for their insatiable AI models – which will theoretically make them smarter – is a daunting task. is now an obsession for companies racing to dominate AI.
Companies like OpenAI, Google, Amazon, Meta and others have considered some crazy solutions. Meta, for example, needed data so badly at one point that it considered buying the publishing house Simon & Schuster and even weighed the risk of copyright lawsuits for more material, The New York Times reported.
Another solution to the problem of limited data is to simply create new data, what big tech calls “synthetic data.” Synthetic data is artificially generated and designed to mimic data generated by real-world events. Zuckerberg is in it.
“I think there will be a lot of synthetic data, where models will try to look at different problems and see what pathways end up working, and then use that to reinforce,” he said.
Anthropic, the creator of the Claude chatbot, also powered internally generated data in its models. And ChatGPT Maker OpenAI is considering it, although CEO Sam Altman said at a conference last May that the key was having a model “smart enough to produce good synthetic data.”
And while Zuckerberg sees feedback loops as the key to creating powerful AI models, there are also risks to relying on them. They could reinforce some of their mistakes, limitations and biases if they are not trained on the “right data” from the start.
businessinsider
