The Daily AI Briefing - 16/06/2025

The Daily AI Briefing

Welcome to The Daily AI Briefing! I'm your host, bringing you today's most significant developments in artificial intelligence. From self-improving AI systems to surprising insights on how children interact with this technology, we're covering the breakthroughs and trends shaping our digital future. Let's dive into today's stories that matter. In today's episode, we'll explore MIT's breakthrough in self-improving AI, examine how AI is developing human-like understanding, look at a concerning digital divide in children's AI access, and touch on other significant industry developments including notable statements from tech leaders. First up, MIT researchers have achieved a remarkable breakthrough with their Self-Adapting LLMs framework, or SEAL. This system enables large language models to essentially teach themselves by creating their own training data and instructions for self-updates. What makes this particularly impressive is that in knowledge tasks, the AI learned more effectively from its own notes than from materials generated by the much larger GPT-4.1. Even more striking, the system improved its puzzle-solving abilities from 0% with standard methods to 72.5% after learning to train itself. This development is particularly noteworthy as self-improving AI is often considered a potential pathway to superintelligence. In related news, Chinese scientists have discovered that AI models are spontaneously developing internal 'maps' of the world that mirror human conceptual understanding. Their research involved testing AI models on 4.7 million "odd-one-out" decisions across nearly 2,000 common objects. The results showed that AI naturally developed 66 core ways of categorizing objects, closely matching human mental categorization patterns. The AI's conceptual map showed strong alignment with human brain activity patterns, suggesting these systems are building genuine internal concepts rather than simply memorizing patterns. Shifting focus to a concerning social trend, a new study from The Alan Turing Institute has revealed a significant digital divide in children's AI usage. While 22% of UK children aged 8-12 are already using AI, private school students are nearly three times more likely to have access than their state school peers. Private school children showed 52% usage rates compared to just 18% in state schools. Interestingly, some children reported refusing to use AI after learning about its environmental impact, specifically its energy and water consumption. The research also found that children primarily use AI for creativity and learning, with many reporting that the tools help them communicate better. In industry news, Nvidia CEO Jensen Huang made waves by stating he "pretty much disagrees with almost everything" Anthropic CEO Dario Amodei has said regarding AI and job automation. Meanwhile, OpenAI has updated its Projects feature with new support for deep research and voice mode, and AstraZeneca has signed a $5.3 billion AI research deal with China's CSPC aimed at developing new oral medications for chronic diseases. As we conclude today's briefing, it's clear that AI continues to develop at a breathtaking pace. From systems that can teach themselves to improving conceptual understanding that mirrors human cognition, we're witnessing fundamental advances in the field. Yet the digital divide revealed in the UK study reminds us that access to these technologies remains uneven, potentially creating new forms of inequality. Thank you for joining us on The Daily AI Briefing. Stay informed, stay curious, and we'll see you tomorrow for more developments from the cutting edge of artificial intelligence.