• British judges can now use AI tools for writing legal opinions, but not for case research or legal examinations – the British judiciary advises AI use for routine tasks but cautions against its use in new information analysis or detailed reasoning. - Reuters
• At Davos, Sam Altman admitted that his focus is OpenAI’s next big model after GPT-4, which “will be able to do a lot, lot more” – he also said that for OpenAI to thrive, they’re going to have to accept that some people or organizations will use their technology for things that don’t necessarily align with their morals. - Axios
• Security vulnerability in GPUs from Apple, Qualcomm, and AMD GPUs – nicknamed LeftoverLocals, it could enable data theft from GPU memory and leave AI systems vulnerable, exposing the need for improved GPU memory security. - Wired
• Salesforce CEO and Time magazine owner Marc Benioff accused AI companies of stealing intellectual property to train models – Benioff calls for standardizing payments to content creators. - Bloomberg
• Microsoft has announced new GenAI ad creation features in their Retail Media tool – it can create ads based on a product URL, automatically align with any retailer’s style guide, and enable further customization with more prompts.- Social Media Today
• AlphaGeometry produces human-readable proofs and solves 25 out of 30 olympiad-level geometry problems – geometry is one of the hardest areas to teach to LLMs, so this is a big step forward in making AI more useful for mathematics. - Nature
