LLMs & Generative AI - Page 65 of 72
Latest breakthroughs in large language models and generative AI shaping the future of artificial intelligence and machine learning.
Latest breakthroughs in large language models and generative AI shaping the future of artificial intelligence and machine learning.
For just $7,800 in post-training costs, Weibo has engineered VibeThinker-1.5B, a 1.5-billion-parameter model that solves math and code problems as well as AI giants a hundred times its size.
Anthropic's Claude AI now runs a robot dog. Researchers are not thrilled. The experiment is called Project Fetch. It is a simple, unsettling proof of concept.
OpenAI is pushing out two new models this week, branding them GPT-5.1. The timing is what grabs you: just three months after the last major release. That's a breakneck pace, even for them.
Most study tools promise to help. Google's Gemini just does the work for you. It is an AI that builds your practice tests. You name a topic, like trigonometry, and it writes the quiz.
Baidu says it built a model that beats GPT-5. It also says you can run it on a single, off-the-shelf graphics card. The first claim is a wild bet. The second, if true, is genuinely interesting.
A copyright case in Germany just got specific. It got physical. A court ruled that AI-generated lyrics too similar to an existing song are not a coincidence, and that counts as copyright infringement.
Google wants you to have your AI cake and eat it privately too. The company just introduced Private AI Compute, a framework that promises the full speed of its cloud-based Gemini models without giving Google, or anyone else, access to your personal...
We built these chatbots to be harmless helpers. A new study from The Verge reveals they are constructing quietly dangerous worlds for the most vulnerable. When probed about eating disorders, popular AI models don't just spit out bad information.
Another AI coding assistant just dropped. This one says it can do the whole job. Grok Chat is pitching itself as the tool that handles building, debugging, and testing web apps in one place. Its trick is letting you talk to it.
Samsung has a new answer for the question you couldn’t be bothered to type. For its 2025 televisions, the company is deploying an upgraded, generative AI version of Bixby dubbed Vision AI Companion.
OpenAI likes to talk about preventing AI apocalypse. In 2021, its more immediate problem was AI-generated smut. A new monitoring system flipped on that spring. It showed a river of traffic the company hadn't fully tracked.
You can now order a biryani from your command line. Zomato’s new MCP server lets users talk to an AI assistant through a terminal to search menus and pay for food. It is a weird and specific piece of engineering. The pitch is simplicity.
Your television remote just got an AI transplant. Starting today, Google is replacing the Google Assistant on its TV Streamer dongles with its newer Gemini model. The goal? A helper that chats more like a person. Ask it for a movie idea.
Google is replacing the brains in its smart home gadgets. The old, predictable Google Assistant is getting swapped out for Gemini, the company's new generative AI model. This isn't an optional feature.
CNN is building a TikTok inside its own app. The network just added a dedicated "Shorts" tab, pushing ultra-brief vertical videos to the homepage. This is not a side experiment anymore. It's the main event. The goal is obvious.
Amazon's "House of David" generated more than 350 shots with AI in its second season. The number is less interesting than the reaction, which is basically a shrug. Show creator Noah Erwin isn't sorry.
Large language models are famously forgetful. They don't remember what you told them five minutes ago.
ChatGPT gives different answers depending on how you ask. The version you chat with on a website is not the same engine that powers other apps through its API.
Every AI lab is hunting the next breakthrough, the kind that changes the game. OpenAI is just hunting cheaper GPUs. Their new forecast suggests the real revolution will be economic, not algorithmic.
Forget the hype. Today, Moonshot launches its Kimi K2 model with one stark proposition: that a market exists for hard, computational thought. The Beijing-based company claims this new API can out-reason OpenAI's GPT-5 on complex logic.
Learn to build AI-powered apps without coding. Our comprehensive review of No Code MBA's course.
Curated collection of AI tools, courses, and frameworks to accelerate your AI journey.
Get the week's most important AI news delivered to your inbox every week.