FILE PHOTO: Former OpenAI employee Suchir Balaji was found dead in his San Francisco apartment on November 26th. | Photo credit: X

Suchir Balaji, 26, was found dead in his San Francisco apartment on November 26th. He was a researcher at OpenAI for four years before leaving the Microsoft-backed startup over disagreements over fair use and copyright policies.

His sudden demise has shocked the industry as the news comes three months after he spoke to Balaji The New York Times, in an extensive interview against OpenAI’s blatant use of copyrighted data and how it is illegal and why it could ultimately harm the Internet.

Balaji revealed that he quit in August because he could not support OpenAI’s actions. He shared that he intends to pursue his own ambitions in the AI ​​space.

While at OpenAI, Balaji also worked on the company’s industry-defining product, ChatGPT, for 1.5 years.

After graduating from Berkeley, Balaji interned at OpenAI and data labeling startup Scale AI. In 2020, he finally switched to OpenAI full-time. A few years later, he began collecting data to pre-train new flagship AI models at OpenAI, including GPT-4. As long as his research was limited to internal use, Balaji was fine with using digital data to train AI models.

But his mind changed after OpenAI generally released ChatGPT in November 2022. The AI ​​chatbot was well received by users and became a huge success, pushing the company to pursue a profitable path.

After it was eventually revealed that OpenAI had indiscriminately searched data, including prominent news publishers The New York Times himself filed copyright lawsuits.

Then, in October, Balaji reached out to the outlet to talk about his findings.

In a tweet, Balaji said that at first he “didn’t know much about copyright, fair use, etc. but became curious after seeing all the lawsuits filed against GenAI companies.” As I tried to understand the issue better, I came “finally concluded that fair use seems to be a fairly implausible defense for many generative AI products, for the main reason that they can create substitute products that compete with the trained data.”

Separately, he published a blog saying that companies like OpenAI and Microsoft argue that they can train their AI models using freely available data from the Internet because they violate the “fair use” requirement of U.S. copyright law fulfill.

According to his blog, while these AI models would not directly copy the training data, they would most likely contain recovered information mixed in the output. These large language models had to be trained multiple times and repeatedly for each data point, and in the end they usually tend to remember the text verbatim.

He also explained that in this way, the AI ​​chatbot simply attracted traffic from other communities such as Stack Overflow, a public question-and-answer site for computer programmers. Essentially, this would result in duplication of data across the internet.

He concluded by adding that this is not a problem exclusive to ChatGPT or OpenAI, but rather a problem that every other generative AI product faces.

The day before San Francisco police found Balaji’s body in his apartment, the researcher was named in a court filing in a copyright lawsuit against OpenAI. As a good faith compromise, OpenAI agreed to search Balaji’s custody file for documents related to the concerns he raised.

When Balaji’s interview came out, OpenAI was going through another period of turmoil. Investors had just deposited $6.6 billion in funds, and the company was facing a talent exodus. Senior executives, including Chief Technology Officer Mira Murati, Chief Research Officer Bob McCrew and Vice President of Research Barret Zoph, had all resigned.

A report from Wired shared that the departures were due to the company’s change in direction from research to product, which was a controversial change in mindset for many employees.

OpenAI has steadily moved away from its origins as a nonprofit organization. In 2019, it was converted to a “limited profit” entity, where the nonprofit organization was appointed the governing body of a for-profit subsidiary. Then, on September 26, it was reported that OpenAI was in the process of transforming into a fully for-profit company.