This Week in Generative AI – 15th to 21st December

Google limits election-related AI use, including restricting election-related queries with Bard

• Google is implementing measures to prevent misuse of its platform during the 2024 elections.
• It will limit election-related queries on its AI products, Bard and Search Generative Experience (SGE), require disclosure of AI-generated content in ads, and introduce labels for synthetic content on YouTube.
• Google will also enhance cybersecurity for high-risk individuals during elections.

Read the Article

Google Gemini heads to developers amid lukewarm reception

• Google’s Gemini system, designed to outperform OpenAI’s GPT-4, has received mixed reviews.
• While Google claims superior performance, the margin is slim and its multimodal functionality is not yet fully available.
• Despite this, experts believe the system’s ability to interchange between voice, text, and imagery could revolutionize AI applications.

Read the Article

New language model breakthrough on Nature: Google DeepMind’s FunSearch addresses a longstanding mathematics challenge

• A collaborative research team has developed FunSearch, a novel approach that combines a Large Language Model (LLM) with an evaluator to solve complex mathematical problems.
• The method evolves low-scoring programs into high-scoring ones, leading to the discovery of new algorithms.
• FunSearch has successfully addressed the longstanding cap set problem, outperforming current computational solvers.

Read the Article

Google unveils MedLM generative AI models for healthcare with HCA, Augmedix and BenchSci as early testers

• Google has launched MedLM, a set of AI models for healthcare applications, available through Google Cloud.
• The models, developed from the Med-PaLM 2 platform, are being tested by HCA, Augmedix, and BenchSci.
• The models are designed to assist with tasks such as medical note creation, drug discovery, and improving patient experiences.

Read the Article

Google unveils VideoPoet, a new LLM for video generation

• Google has introduced VideoPoet, a large language model (LLM) for video generation. It performs tasks like text-to-video conversion, video stylisation, and video-to-audio conversion.
• VideoPoet stands out for its ability to create coherent large motions in videos and its integration of multiple video generation capabilities.
• However, its real-world application and effectiveness are still under debate.

Read the Article