The release and publicization of DeepSeek’s R1 model roiled the technology sector of the stock market. In this video, Bryan Sterba discusses the many questions that remain unanswered weeks after R1’s release.
Speakers:
Bryan Sterba, Partner, IP & Tech Transactions, Lowenstein AI, and Emerging Companies & Venture Capital
Lowenstein AI: A-I Didn’t Know That
Partner Bryan Sterba examines the myriad legal and business implications of this revolutionary technology. Bryan and guests from across firm practices discuss the many ways that businesses can leverage AI to increase productivity and efficiency, as well as the challenges presented as companies, legislatures, and regulatory agencies weigh differing approaches to managing intellectual property rights, privacy and data security, and national security interests, amongst other concerns.
READ THE TRANSCRIPT
Bryan Sterba: Hi, I'm Bryan Sterba, partner at Lowenstein Sandler’s AI practice, here with another update for our "AI Didn't Know That" series.
This one is on the release and fallout from the R1 model from DeepSeek. For those who don't know, DeepSeek is an artificial intelligence company owned by and growing out of a Chinese hedge fund. It released an open-source version of its powerful R1 model in January of this year, along with claims the model was trained on a $6 million budget.
On Monday, January 27, public companies heavily invested in the so-called AI gold rush lost over a combined trillion dollars in market cap, with Nvidia at one point shedding nearly $700 billion on the belief that DeepSeek could prove the cost of building and training powerful models could really be done at a fraction of the assumed spend. In the weeks since, tech stocks have largely recovered, but mostly based on skepticism of DeepSeek’s training and development claims. There are still important questions, though, that investors are grappling with:
- Will cheaper-to-train models soon be available?
- Will their availability lead to increases or decreases in sales for companies like Nvidia?
- Will this heighten the market for computational power and further grow the data center sector?
- Will all of that give rise to higher energy demand?
And while investors focus on that, the AI developing market incumbents must answer their own question: how do we stay ahead of constantly improving open-source models? If open-source models are able to keep pace with their closed-source cousins, foundation models will quickly become commoditized and shift the market's focus to use case specifics and the incredible value of proprietary data for fine-tuning.
Ultimately, the quantity and quality of data may be the most decisive factor for those building and utilizing AI models and differentiating the winners and losers in the AI ecosystem. Those who possess vast stores of world-class data with high-value use cases will derive great rewards in the coming years from their ability to license proprietary nonpublic content.
And with all that, the biggest open legal question concerning AI and model development still remains open: the question of whether or not the training of generative AI on copyrighted data can be deemed a "fair use."
We'll be sure to keep an eye on DeepSeek and all related developments.
Thank you for watching, and tune in next time to "AI Didn't Know That."
