LLaMA

The article discusses prototyping AI agents on Solana without incurring high OpenAI credits costs. Developers are looking for cheap or free inference options for the prototyping phase. The author shares their own solution, a free LLM chat API, and asks how others handle this issue.

Five major publishing houses and author Scott Turow are suing Meta and its CEO Mark Zuckerberg for allegedly training Llama generative AI models on millions of copyrighted materials. The lawsuit claims copyright infringement. Meta's CEO and the company are named as defendants.

Meta is facing a lawsuit alleging that its AI, Llama, was trained on copyrighted work and generates summaries and verbatim copies of original works. The lawsuit claims copyright infringement by Meta's AI model. This lawsuit may have significant implications for AI training practices.

The article suggests DefiLlama consider a rebrand to simply 'Llama,' consolidating services like LlamaSwap under the unified brand. The author notes that peers already colloquially refer to the platform as 'Llama,' framing the change as a logical simplification.

Meta has shifted focus from its open-source Llama AI model to a proprietary model called Muse Spark. This move marks a strategic pivot away from open-source development toward a closed, company-specific approach.

A Stanford University study found AI models provided more praise and softer feedback to Black students and female writers in essays, while offering stricter grammar corrections to Hispanic/English learners and White students. Researchers analyzed 600 eighth-grade essays using models like ChatGPT and Llama, noting biased patterns in automated feedback.

Meta's Superintelligence Labs unveiled Muse Spark, their first AI model in the Muse family, marking a shift from previous Llama models. The model integrates content from Meta platforms like Instagram and Threads, with plans for future features citing user-generated content.