A.I. Tracker 🤖

Quantization, a technique to make AI models more efficient, has limits as shown by a study. Training large models and then quantizing them may not always be beneficial as smaller models could perform better in some cases. Lower precision formats can make models more robust, but models have finite capacity. Inference costs cannot be reduced indefinitely without impacting model quality. New architectures focusing on low precision training may be important in the future.

Motorola Solutions utilizes AI in their Vesta NXT software to streamline 911 emergency calls, consolidate data, and improve response times for call handlers. The software also assists in managing nonemergency calls, translating languages, and providing location details to first responders.

AI on smartphones, touted as revolutionary, still falls short of expectations. Major phone makers like Samsung, Google, and Apple offer AI features that are more gimmicky than genuinely useful. The promise of true AI smartphones remains unfulfilled.

The article discusses the phenomenon of people placing unwarranted trust in AI systems like ChatGPT, despite their lack of true understanding or reliability, as they seek quick and easy answers in a world filled with complexity and doubt.

Meta (formerly Facebook) is focusing on AI, investing billions for infrastructure and generative AI tools. Its Meta AI assistant is seeing growing popularity, especially in India. Advertisers are using its AI tools for more efficient ad creation. The company is also aiming to expand into enterprise technology with the hiring of Clara Shih from Salesforce. Critics note the confusion and challenges in using Meta's AI tools, while analysts predict continued success for Meta in 2025.

Agentic AI, the next wave of AI technology, is being developed by major tech companies like Salesforce, Google, and OpenAI to autonomously complete complex tasks for humans in various fields. Salesforce plans to have over 1 billion AI agents in use by next year.

Elon Musk's AI company xAI is testing a standalone iOS app for its chatbot Grok, offering generative AI features like text rewriting and image generation. The app is available in beta in Australia and other countries. Grok was previously only accessible to paying subscribers but is now free for all users.

Sriram Krishnan, former a16z partner, appointed as senior policy advisor for AI at the White House. Will work with David Sacks on AI policy and continue American leadership in technology.

OpenAI introduces new AI reasoning models, o3, with improved alignment to company's safety policy using deliberative alignment method during inference phase, aiming for safer and more ethical responses.

Founders Cristian Ponce and Théo Schäfer met at a Halloween party in 2023, bonding over lab tech struggles. They founded Tetsuwan Scientific to modify lab robots with AI capabilities, aiming to automate the scientific method. They hope to create independent AI scientists for hyperbolic growth in scientific discovery.

Aditya Challapally teaches a Stanford Online course on generative AI for tech professionals, emphasizing the importance of technical skills and domain expertise. He recommends using tools like ChatGPT for better understanding AI.

Former OpenAI engineer and whistleblower, Suchir Balaji, who helped train ChatGPT AI systems, has died at 26. He raised copyright concerns regarding AI technology and its use of data. His death was ruled a suicide. Colleagues and family mourn his loss.

The Acer Swift 16 AI offers a sharp OLED display, good speakers, and decent performance for multitasking, but falls short in battery life and keyboard comfort. Priced at $1,199.99, it may be worth considering if on sale for $850 or less.

OpenAI's development of GPT-5, code-named Orion, is reportedly behind schedule and not justifying costs. Despite improved performance, it's not advancing enough, leading to new strategies like creating fresh data.

TCL released five new AI-generated short films for its streaming platform, ranging from tolerable to disturbing. From a futuristic concept to a Kafkaesque transformation into a slug, the films struggle with animation and storytelling.

Mondelez, the company behind Oreo cookies, has developed a machine learning tool to create new flavors for its snacks. The AI has been used in over 70 products, generating unique flavors like "Gluten Free Golden Oreo" and updated Chips Ahoy recipes. Human taste testers ensure quality control.

Databricks raises record $10 billion, focusing on buying back employee stock for AI talent. Naveen Rao discusses the AI talent war and the company's strategy.

OpenAI's ChatGPT has had a significant impact since its launch, with updates like the GPT-4o model and various controversies, including legal issues and misuse cases. Despite challenges, OpenAI is continuously evolving the chatbot's capabilities and expanding its reach.

OpenAI announced safety testing of new frontier model o3, skipping o2. It excels in coding, math competitions, and reasoning challenges. Also, new research on deliberative alignment for AI safety.

AI companies raised $19 billion in Q3 of 2024, representing 28% of all venture funding. Several U.S.-based AI companies secured $100 million or more in funding, with multiple billion-dollar rounds, showing continued investor interest in the AI sector.