#GPT-4

'An eight-minute conversation with a GPT-4 based chatbot reduced belief in conspiracy theories for many participants, with effects lasting at least two months'

'Australia relies on global LLMs while Kangaroo LLM aims to build a sovereign, Australian-tailored model, but remains in early data and governance phases as of August 2025.'

'Learn how to set up an Arena-as-a-Judge workflow to compare LLM outputs head-to-head using GPT-5 as an evaluator. The tutorial includes code, sample prompts, and interpretation of evaluation logs.'

SciSummary uses advanced AI to transform lengthy scientific papers into quick, clear summaries, podcasts, and slideshows, helping researchers and students save time and understand research faster.

Dimitri Masin, CEO of Gradient Labs, shares insights on building autonomous AI agents for regulated industries, focusing on quality, compliance, and future trends in customer support.

New research from Microsoft and Salesforce shows that large language models experience a 39% performance drop when handling real multi-turn conversations with incomplete instructions, highlighting a key challenge in conversational AI.

OpenAI has launched HealthBench, an open-source framework to rigorously evaluate large language models in healthcare using expert-validated multi-turn clinical conversations.

Google researchers introduced REFVNLI, a new cost-effective metric that jointly evaluates textual alignment and subject consistency in subject-driven text-to-image generation without relying on expensive APIs.

A recent study reveals that being polite to AI does not enhance the quality of its answers, as AI output deterioration depends on content tokens rather than courteous language.

Geoffrey Hinton proposes that human thought relies more on analogies than logic, reshaping our understanding of cognition and driving new directions in AI development.