Artificial Intelligence News - The Intelligent Times

Google Agentspace: AI-powered search for enterprises I have been with a few companies throughout my career and have seen many internal enterprise search systems. Most of them sucked, tbh. We have had an internal version of Agentspace here at Google for a quite a while now, and I have to say this is by far the best enterprise search I have experienced. And it is not just search: In addition to information discovery it also comes with NotebookLM Plus, which brings an enterprise-ready version of the popular NotebookLM enhanced with security and privacy features for work. Deloitte was one of the first customers for Agentspace and their testimonials speak for themselves. Enterprise search might actually, eventually get good 🙃

If you're diving into Generative AI, make sure to understand the transformer architecture - it is the backbone of all modern AI models. These resources are amazing and free: - Attention Is All You Need: arxiv.org/abs/1706.03762 - The Illustrated Transformer: https://lnkd.in/e7v9Fz9n - Transformers (how LLMs work) explained visually: https://lnkd.in/e-QwFB_7 - Large Language Models explained briefly: https://lnkd.in/efnnU8XF - How might LLMs store facts: https://lnkd.in/e2DdkRxZ

https://lnkd.in/eHH9Q9ea

Big day at SC24!! Excited to announce El Capitan, powered by AMD is now the world's fastest supercomputer at 1.742 exaflops and Frontier, our first exaflop machine is now #2 on the Top500 list at 1.35 exaflops! AMD EPYC and Instinct power 5 of the top 10 and 21 of the top 50 supercomputers in the world. Thank you to Hewlett Packard Enterprise, Lawrence Livermore National Laboratory, Oak Ridge National Laboratory, U.S. Department of Energy (DOE) for their partnership in pushing the boundaries of innovation for HPC workloads!

✨ More exciting news, this time from TechCrunch! ✨ D-ID is announcing the next generation of our #AI #Avatars! Today we're launching D-ID's new Marketing Suite, featuring our new state-of-the-art hyper-realistic Avatars that are practically #indistinguishable from real people. After being recognized yesterday as one of the most innovative products in 2024 by TIME, we're proud to be leading the way in AI-powered content creation.  Kudos to the #1 team in the world - D-ID! 🏆 Thanks, Ivan Mehta and #TechCrunch, for the coverage! https://lnkd.in/e66GbVqc Sella Blondheim Eliran Kuta  #DID #MarketingSuite #AI #DigitalHumans #Innovation

BREAKING: OpenAI just dropped its long-rumored next-gen model "o1" (aka 🍓). What's new? ChatGPT can now actually "think". Early benchmarks show it's on par with Ph.D. students. But there's more. Here's what happened: OpenAI's new strawberry model allows it to take time to think through a problem. Imagine switching between System 1 thinking (fast, intuitive, error-prone – GPT4o) and System 2 thinking (slow, deliberate, more reliable – Strawberry o1). I simple words: Instead of answering right away, ChatGPT can now take a while and come back to you with a much better answer after a while. For some tasks (like creative writing) this doesn't make a big difference, but for others, where you need complex reasoning, it's a complete game changer. Think about programming or complex data analysis. But it doesn't stop here. The real deal is that OpenAI seems to have figured out a way to get better answers the longer the model thinks. This is insane. Right now, Strawberry can think for seconds. But OpenAI plans bigger. They are looking for models that can think for hours, days, or even weeks to solve more complex problems like mathematical proofs, drug discovery or solving tough legal cases. This is completely changing the way we interact with LLMs. I'm excited to dive deeper into it as it's being rolled out. Folks, buckle up. These are exciting times ahead! Find some cool demos in the comments 👇

OpenAI unveils smaller, cheaper #AI model: GPT-4o mini OpenAI has launched a smaller, more affordable version of its AI model, GPT-4o mini. The new model is designed to be less energy-intensive and is priced significantly lower than its predecessor. Despite its size, GPT-4o mini demonstrates impressive performance on various benchmarks. This move aims to expand access to OpenAI's technology and compete with rivals like Meta and Google. more information: https://lnkd.in/e_wwkaqa #artificialintelligencenews #chatgpt

OpenAI's secret weapon 🍓 "Strawberry" seeks to crack the code of #AI reasoning OpenAI, the company behind the popular ChatGPT language model, is reportedly making significant strides in artificial intelligence reasoning with a project codenamed "Strawberry." This project comes as various tech giants race to develop AI models with advanced reasoning capabilities. Details about Strawberry are scarce, but internal documents reviewed by Reuters shed some light. The project revolves around a specialized post-training process applied to OpenAI's existing AI models. This post-training, potentially similar to Stanford's STaR method, focuses on enhancing the models' ability to conduct "deep research." Deep research, according to the documents, entails the model's capacity to autonomously navigate the internet and gather information to complete complex tasks. This "long-horizon task" (LHT) capability has been a major hurdle for AI models, often leading to inaccurate information and illogical solutions. Some experts, like Stanford's Noah Goodman (creator of STaR), believe such advancements could push AI beyond human-level reasoning, potentially leading to unforeseen consequences. However, there's disagreement within the AI research community. Yann LeCun of Meta, for example, doubts the ability of large language models (LLMs) to achieve true human-like reasoning. OpenAI reportedly hopes Strawberry will empower its AI models to conduct independent research, potentially leading to breakthroughs in science and software development. The company has privately hinted to developers about this upcoming technology with significantly enhanced reasoning capabilities. While Strawberry represents a significant step forward, many questions remain. The exact workings of the project are shrouded in secrecy, even within OpenAI. The nature of the "deep-research" dataset and the timeframe for extended periods are unknown. Additionally, the project's public availability remains uncertain. #artificialintelligence #deeplearning #innovation

A consortium of current and former employees from leading AI companies, including OpenAI and Google DeepMind, issued a warning about the potential risks associated with #AI technology. In an open letter, they expressed concerns about the financial motives of AI companies hindering effective oversight, highlighting risks from unregulated AI, such as misinformation spread and the deepening of existing inequalities. The letter also underscored instances of image generators producing disinformation despite company policies against such content. The group urged AI firms to establish channels for employees to raise risk-related concerns and refrain from enforcing confidentiality agreements that stifle criticism. For further insights, check this excellent article from the New York Times: https://lnkd.in/dnhDyjvW #artificialintelligence #ainews #regulation

NASA - National Aeronautics and Space Administration Appoints Inaugural Chief Artificial Intelligence Officer NASA has named David Salvagnini as its new Chief AI Officer. Salvagnini, already serving as Chief Data Officer, will now lead NASA's responsible use of AI technology. This move aligns with President Biden's AI Executive Order. Salvagnini brings over 20 years of tech leadership experience, including in the intelligence community and the Air Force. He will spearhead AI innovation to benefit humanity both on Earth and in space. Source NASA. More info: https://lnkd.in/ef5JmKby #nasa #ai