OpenAI and its discontents: Why countries need laws to combat AI exploitation
content without permission? Or misattributing content to someone else and deflecting blame in the process? It is possible that both could result in equally disastrous consequences.
In a recent lawsuit, The New York Times accused Microsoft-owned ChatGPT of creating large-language-model (LLM) products out of copyrighted material derived from, and competing with, its authentic journalism, while also attributing incorrect and false information to the publication. The legal case involves violating copyright protection, and exploiting commercial value that is enabled by such protections.
An investigation revealed that of all the training datasets used in GPT-3, the most highly weighted dataset — 'common crawl', operated by an 'eponymous 501©(3) organisation run by wealthy venture capital investors' — had www.nytimes.com as the most highly represented proprietary source.
No matter how the courts in the US eventually rule in the case, there are two fundamental questions that need to be considered here:
- Should firms like OpenAI be allowed to train ChatGPT on third-party material while profiting from both copyrighted and unprotected content?
- Should firms like OpenAI be allowed to get away with compromising content integrity by attributing fictitious information, or adding misleading information to legitimate sources?
Nicking tunes One sector in India with questionable copyright and commercial protection is the arts and entertainment industry. So, while the music industry, for instance, has seen publishing revenues grow by 250% over the last three years, the copyright compliance rate remains 'abysmally low' at just 1.2%, according to an EY report.
The vulnerability of this sector is further exasperated by the predominance
