the-decoder3d
Perplexity uses Deepseek-R1 to offer Deep Research 10 times cheaper than OpenAI
Perplexity says the new tool automatically conducts comprehensive research, performing dozens of searches and analyzing hundreds of sources to produce detailed reports in one to two minutes - a ...
the-decoder3d
AI assistants fail basic fact-checking in BBC news study
The study, conducted by the BBC, tested ChatGPT, Microsoft Copilot, Google Gemini, and Perplexity on their ability to accurately report current events. In December 2024, 45 BBC journalists evaluated ...
the-decoder3d
Claude Jailbreak results are in, and the hackers won
The results from Anthropic's Claude jailbreaking challenge are in. After five intense days of probing - involving over 300,000 messages and what Anthropic estimates was 3,700 hours of collective ...
the-decoder5d
Anthropic prepares new Claude hybrid LLMs with reasoning capability
Anthropic is getting ready to release a new AI model that combines traditional language model capabilities with advanced reasoning functions. According to The Information, the model will be available ...
the-decoder6d
Adobe's Firefly AI video generator is now in public beta, turning text and images into short clips
Adobe's latest addition to Creative Cloud lets users turn text and images into short videos, bringing AI video generation directly into familiar creative tools. After limited testing last year, Adobe ...
the-decoder3d
Language models tend to favor other LLMs that make mistakes similar to their own
A new study of how language models evaluate each other has uncovered a troubling pattern: as these systems become more sophisticated, they're increasingly likely to share the same blind spots.
the-decoder6d
OpenAI plans GPT-4.5 release within weeks, sets stage for GPT-5
OpenAI is getting ready to release GPT-4.5, developed internally as "Orion," in the coming weeks. CEO Sam Altman says it will be the company's final traditional language model before a major shift in ...
the-decoder6d
Perplexity AI launches new ultra-fast AI search model Sonar
The latest version of Perplexity AI's search model Sonar has arrived, powered by Meta's Llama 3.3 70B and some specialized hardware. In internal tests, the company says Sonar performs better than ...
the-decoder4d
'Boomer prompts' don't work with its new reasoning models, OpenAI says
The o-series models, which OpenAI calls "the planners," specialize in tasks related to mathematics, engineering, law and finance. The GPT models serve as "workhorses" for quick, basic tasks. "One ...
the-decoder5d
OpenAI is thinking about a "grown-up mode" and wants ChatGPT to be less sycophantic
OpenAI has significantly updated its Model Spec, expanding the document that outlines how its AI models should behave from its initial release in May 2024. The new 63-page guidelines focus on three ...
the-decoder3d
Meta joins the race to build humanoid robots
Meta has begun developing both hardware and software for AI-powered humanoid robots, according to a new report from Bloomberg. The company's Reality Labs division has assembled a dedicated team ...
the-decoder6d
AI language models struggle to connect the dots in long texts, study finds
The latest generation of AI language models hits its limits when connecting information across long texts and drawing conclusions, according to new research from LMU Munich, the Munich Center for ...