Elon Musk, Twitter’s CEO, vowed to sue Microsoft on Wednesday, saying that the software giant had stolen Twitter data in order to train an AI.
Musk’s tweet of the threat came after it was reported on Mashable and elsewhere that Microsoft will remove Twitter from its advertising platform, which helps ad buyers manage their accounts across many social media sites at once.
“They trained illegally using Twitter data,” A tweet from Musk.
They trained illegally using Twitter data. Lawsuit time.
— Elon Musk (@elonmusk) April 19, 2023
Musk, CEO of both Tesla and SpaceX, is notorious for tweeting about ideas that never materialize, although he has yet to face legal action for it. The Twitter press line didn’t provide any useful feedback, and Microsoft’s response was to stay mum.
The danger posed by Musk is the latest sign that data ownership is becoming a hotly contested arena in the race to commercialize generative AI. Cutting-edge artificial intelligence models are being developed by major tech firms like OpenAI’s GPT, while content owners want to halt or profit from their usage.
Microsoft creates its own LLMs (large language models) and also provides access to OpenAI’s LLMs for a fee. OpenAI received $10 billion from Microsoft in a strangely constructed agreement last year.
Prior to his resignation from the board in 2018, Musk was one of the company’s co-founders; he has lately voiced displeasure with OpenAI’s shift from a nonprofit model to a very valued corporation driven by Microsoft.
Terabytes of data, primarily gathered from social media and online communities like Reddit, Stack Overflow, and Twitter, are needed to train LLMs like GPT. Social network training data is useful since it records spontaneous interactions.
The original data owners are starting to make demands as cutting-edge AI models make the transition from academic laboratories to commercial enterprises.
For instance, Reddit said earlier this week that it will begin charging businesses to utilize its API to feed user discussions into AI development tools.
This week, Universal Music Group said that artist training through music would constitute a “significant” “both a breach of our agreements and a violation of copyright law” in reaction to a YouTube video where it is alleged that artificial intelligence sings a song in Drake’s style.
database of stock images Stable Diffusion is being sued by Getty Images on allegations that the latter used stolen materials to train an AI picture generator.
In December, Musk stated that Twitter will “pause” OpenAI’s database access. He has also stated that one of his firms, TruthGPT, will be developing its own massive language model.