Tesla’s (TSLA) CEO Elon Musk stated in a recent livestreamed conversation with Stagwell chairman Mark Penn that the available real-world data for training AI models is largely exhausted, according to a TechCrunch report.
Stay Ahead of the Market:
- Discover outperforming stocks and invest smarter with Top Smart Score Stocks
- Filter, analyze, and streamline your search for investment opportunities using Tipranks' Stock Screener
“We’ve now exhausted basically the cumulative sum of human knowledge in AI training,” Musk explained, adding that this milestone was reached just last year.
This sentiment echoed what former OpenAI chief scientist Ilya Sutskever discussed at the NeurIPS machine learning conference in December. Sutskever pointed out that the AI industry had hit “peak data,” warning that the lack of new, real-world data would drive a shift in the methods used to train AI models.
Musk Champions Synthetic Data to Overcome AI Limitations
In his conversation with Mark Penn, Musk agreed with this, proposing that synthetic data, that is data generated by AI models themselves, would be the key to overcoming this limitation. He pointed that with synthetic data, AI could engage in a form of self-learning, where it would create its own training data to continue evolving.
This shift toward synthetic data has already gained traction among major players in the tech industry, including Microsoft (MSFT), Meta (META), OpenAI, and Anthropic. In fact, according to the TechCrunch report, citing Gartner estimates by 2024, 60% of the data used for AI and analytics projects will be synthetically generated.
What Are the Benefits of Synthetic Data?
There are notable benefits to this approach, such as significant cost savings. For example, the TechCrunch report stated that AI startup Writer developed its Palmyra X 004 model using almost entirely synthetic data for just $700,000, a fraction of the estimated $4.6 million required for an OpenAI model of similar size.
However, synthetic data is not without its challenges. Some research has raised concerns that training on synthetic data can lead to model collapse, where AI outputs become biased and less creative. If the training data itself is flawed, the resulting model’s performance will be similarly compromised, limiting its usefulness and functionality.
What Is the Best AI Stock to Buy?
For investors interested in investing in the artificial intelligence sector, we have rounded up the best AI stocks to buy that Wall Street analysts are bullish or cautiously optimistic about, using the TipRanks Stock Comparison tool.
