Anthropic: Just 250 Poisoned Docs Can Backdoor an LLM
When Anthropic teamed up with the UK’s AI Security Institute and the Alan Turing Institute, they ran a set of tests that kind of surprised me.
Latest breakthroughs in large language models and generative AI shaping the future of artificial intelligence and machine learning.
When Anthropic teamed up with the UK’s AI Security Institute and the Alan Turing Institute, they ran a set of tests that kind of surprised me.
I still remember the buzz when OpenAI dropped ChatGPT in late 2022 - a chatbot that could write, code and answer questions with an uncanny fluency.
Ten-minute grocery drops feel like a trick, yet Zepto, the Indian quick-commerce startup, runs it with a lot of data work. They handle more than 4,000 orders a day in each dark store - a number that would crumble without a solid system.
These days most AI chatbots get their smarts in two passes: first they crunch huge text corpora to guess the next word, then they get a second round of fine-tuning with reinforcement learning so they follow prompts a bit better.
The push to make AI inference quicker just got a bit more clever. Together AI rolled out ATLAS today - a speculative-execution engine that actually learns from the jobs you feed it, and it can crank inference up to four times faster, roughly a 400 %...
These days the tools we use to write code feel almost like teammates - they suggest completions, catch bugs, even draft whole functions. The flip side? They tend to lock us into whatever features the vendor built in.
When we ask a large language model a question, the exact wording can nudge the answer in subtle ways.
When I tried Google’s newest AI agent, I actually saw it move the mouse, type, and scroll through a page all by itself. The tool is free, built by Google DeepMind, and runs on the Gemini 2.5 Pro model.
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.