Musk’s latest legal action against OpenAI, AI self-improvement with DeepMind’s Boundless Socratic Learning, and video sound post-processing with MultiFoley from Adobe — the top 3 AI news stories of the week

#1 — Enough is enough: Musk sues OpenAI

Elon Musk has filed a legal action in federal court in Northern California against OpenAI, Sam Altman, and others, including Microsoft. The suit seeks to halt OpenAI’s transition from a non-profit organization to a for-profit commercial entity and accuses OpenAI and Microsoft executives of anti-competitive behavior and arbitrary resource management. Musk alleges that OpenAI’s actions violate its original commitments to the public and to investors — the retention of non-profit status and a focus on openness and AI safety.  The recently filed suit is just the latest in a series of legal actions by Musk against OpenAI and an assortment of other defendants.

#2 — Boundless Socratic Learning brings AI closer to perfection

Researchers at Google DeepMind introduced the Boundless Socratic Learning framework in a recent scientific paper. BSL could allow AI systems to self-improve using “structured language-based” interactions — without external data or human feedback. The approach is based on “language games” — interactions in which AI agents develop and modify their own learning environments. The system generates data, designs tasks, and evaluates its own effectiveness using game metrics and rewards. Researchers highlight three levels of AI self-improvement:

• Basic training
• Input/output training
• Game choice and potential self-directed code modification

An interesting approach, and one that is not without challenges.

#3 — MultiFoley by Adobe will ease video sound processing

Adobe has introduced the AI system MultiFoley that can automatically generate synchronized post-production sound effects for videos — from prompts, reference audio, or existing sound clips. MultiFoley was trained on a combined dataset of videos from the network and professional sound effect libraries. The system can transform sounds — for example, turning a cat’s meow into a lion’s roar — and maintain synchronization with the video. But, so far, there is only a scientific paper. We are waiting for the release of this tool, currently missing from PremierPro.