Updated January 1, 2024

1. Google Unveils Gemini - DeepMind’s Advanced Multimodal AI for Global Progress

Google and Alphabet CEO Sundar Pichai introduces “Gemini,” Google DeepMind’s most capable AI model, designed to advance scientific discovery, accelerate human progress, and improve lives globally. Demis Hassabis, CEO and Co-Founder of Google DeepMind, describes Gemini as a multimodal model excelling in understanding and operating across different types of information, optimized in three sizes for varied tasks, and marks a new era in AI with its advanced capabilities in areas like reasoning, coding, and understanding complex data.

Source: Google Blog

2. Mistral AI Releases Powerful Open-Source Language Models

Mistral AI released Mixtral 8x7B, a high-quality sparse mixture of experts model (SMoE) with open weights, outperforming Llama 2 70B and GPT3.5 in most benchmarks and offering faster inference. Mixtral, licensed under Apache 2.0, is efficient in handling multiple languages, code generation, and can be fine-tuned for instruction-following, demonstrating its versatility and advanced performance in AI technology.

Source:Mistral Blog

3. AI Speeds Up Supernovae Age Prediction in Astronomy

Artificial intelligence can now accurately predict the age of supernovae and other stellar explosions within milliseconds of detection, which is crucial for handling the vast data expected from the upcoming Vera C. Rubin Observatory’s survey. This AI system, developed by researchers including Daniel Muthukrishna at MIT, quickly identifies the age of cosmic events, enabling timely follow-up observations, despite the challenge of distinguishing supernovae from other bright celestial objects.

Source: News Scientist

4. AI Outperforms Doctors in Medical Text Generation with GatorTronGPT

GatorTronGPT, an AI program trained on clinical healthcare datasets, outperforms general large language models in generating medical text with readability and clinical relevance comparable to actual physicians. Despite its potential in healthcare administration, concerns remain about its validity, biases, and the need for extensive research to fully integrate AI into healthcare.

Source: Forbes

5. EY’s AI System Shows Promise in Detecting Audit Frauds

EY’s trial of an AI system for detecting audit frauds showed success, identifying suspicious activities confirmed as frauds in two out of the first ten UK audit clients. However, the effectiveness and reliance on AI for fraud detection in auditing are debated among professionals due to challenges like the rarity and uniqueness of frauds, concerns about data privacy, and the need for specialized knowledge to ensure AI operates to appropriate standards.

Source: West Observer

6. Elon Musk’s AI Startup Rolls Out Grok, a Rebellious Chatbot on X

Grok, a ChatGPT competitor developed by Elon Musk’s AI startup xAI, has been launched on X (formerly known as Twitter) for Premium Plus subscribers, featuring a unique, edgy personality and real-time data integration from X posts. Unlike other chatbots, Grok can answer questions with more current information and displays a willingness to engage in vulgar or rebellious responses, setting it apart in style and functionality.

Source: Tech Crunch

7. Google Introduces MedLM AI Suite to Aid Healthcare Clinicians and Researchers

Google has announced MedLM, a new suite of healthcare-specific AI models, aimed at aiding clinicians and researchers in tasks like complex studies and summarizing doctor-patient interactions. These models, which vary in size and capabilities, promise to enhance healthcare workflows and decision-making but face challenges such as accuracy and the need for cautious implementation to avoid risks to patients.

Source: CNBC

8. DeepMind’s AI FunSearch Surpasses Humans in Solving Set-Inspired Math Problems

DeepMind’s AI system, FunSearch, has surpassed human mathematicians in solving complex combinatorics problems inspired by the card game Set, improving on the known solutions. This breakthrough demonstrates the potential of large language model-based AI in advancing mathematical problem-solving and fostering new modes of human-machine collaboration.

Source: Nature

9. AI-Designed Proteins Set New Standards in Biomedical and Environmental Applications

Scientists at the University of Washington School of Medicine have used AI to design proteins with unprecedented binding strengths to various biomarkers, a breakthrough with significant implications for drug development, disease detection, and environmental monitoring. This novel approach combines deep-learning methods and advanced protein design tools, demonstrating the proteins’ potential in sensitive and accurate diagnostics and their resilience in harsh conditions, paving the way for a new era in biotechnology.

Source: Phys

Researchers are developing machine learning techniques to help uncrewed underwater vehicles (UUVs) navigate autonomously, overcoming challenges like the absence of GPS signals and poor visibility underwater. Using deep reinforcement learning, they have improved UUVs’ ability to navigate accurately by altering the training process to mimic human learning, focusing on recent actions that led to positive gains, enhancing training efficiency and reducing power consumption.

Source: Spectrum