In other AI news, under the Trump administration, major AI companies have shifted from supporting regulation to lobbying for fewer restrictions on AI advancements. Meta has been accused of using pirated books from Library Genesis to train its Llama 3 model, raising legal concerns. Meanwhile, DeepSeek, a Chinese AI startup, has released an upgraded language model and faces potential bans in the U.S. due to national security concerns. OpenAI is enhancing ChatGPT with new image generation capabilities, while AI models continue to struggle with chemistry tasks, failing to accurately depict chemical structures.
Key takeaways:
- Perplexity, an AI search startup, is interested in acquiring TikTok to merge its search capabilities with TikTok's video recommendation algorithm and make the For You feed more transparent by open sourcing it.
- Under the Trump administration, major AI companies like Google, Meta, OpenAI, and Microsoft are lobbying for fewer restrictions on AI advancements and arguing that their use of copyrighted works for training models is legal under fair use.
- Meta reportedly used pirated books from Library Genesis to train its Llama 3 large language model, raising legal concerns as part of a copyright lawsuit filed by authors.
- OpenAI's ChatGPT is gaining new image generation capabilities, with the GPT-4o model improving in following complex prompts and producing images, while also releasing research on the emotional well-being effects of using ChatGPT.