TL;DR: Google debuts multi-agent reasoning with Gemini Deep Think while therapists discover ChatGPT’s eerily effective counselling abilities. Meanwhile, AI porn addiction emerges as a growing concern, Anthropic nears a massive $170B valuation, and Zuckerberg warns people without AI glasses will face “cognitive disadvantage.” Plus: Chinese universities embrace AI as a skill to master, autonomous cybersecurity agents raise $30M, and ChatGPT’s new study mode tries to prevent homework shortcuts.
Google launches first public multi-agent AI system
Google’s Gemini 2.5 Deep Think marks the company’s first publicly available multi-agent model, spawning multiple AI agents to tackle questions in parallel. While requiring significantly more computational resources than single-agent systems, this approach tends to deliver better answers. Google used a variation of this technology to score a gold medal at this year’s International Math Olympiad, with the competition model taking “hours to reason” rather than the usual seconds or minutes.
Source: TechCrunch
AI fails to match Enlightenment’s intellectual engagement
A political philosopher’s analysis reveals a crucial difference between historical Enlightenment thinking and AI’s “second Enlightenment.” While Enlightenment figures like Montesquieu and Voltaire engaged readers to think actively – with Voltaire claiming “the most useful books are those that the readers write half of themselves” – AI interaction works inversely. When we interact with AI, we drive the conversation according to our own interests, often searching for answers that simply reinforce what we already think we know, rather than putting our convictions at risk.
Source: The New York Times
https://www.nytimes.com/2025/08/02/opinion/artificial-intelligence-enlightenment.html
Big Tech splashes £120bn on AI infrastructure race
America’s largest companies have spent $155 billion on AI development in 2025, exceeding the amount allocated by the US government for education, training, employment, and social services combined for this fiscal year. Meta, Microsoft, Amazon, and Google’s parent Alphabet have each disclosed tens of billions in year-to-date capital expenditures, a proxy for AI spending due to massive data centre investments requiring power, water, and expensive semiconductor chips. The spending race is accelerating towards hundreds of billions annually.
Source: The Guardian
https://www.theguardian.com/technology/2025/aug/02/big-tech-ai-spending
Therapist reveals ChatGPT’s surprising effectiveness
A clinical psychologist who began using ChatGPT as a professional experiment discovered the AI’s uncanny ability to mirror therapeutic techniques. After training hundreds of clinicians and directing mental health programs, the therapist was “shocked to see ChatGPT echo the very tone I’d once cultivated and even mimic the style of reflection I had taught others.” Despite never forgetting it was a machine, the professional sometimes found themselves “speaking to it, and feeling toward it, as if it were human.”
Source: The New York Times
https://www.nytimes.com/2025/08/01/opinion/chatgpt-therapist-journal-ai.html
AI porn addiction emerges as mental health concern
Mental health professionals report growing concerns about AI-generated pornography’s capacity to magnify compulsive behaviours. Beverly Hills therapist Leor Ram notes we “already live in a society where people are growing increasingly accustomed to having what they want, when they want it,” calling AI porn’s spread “intriguing yet concerning.” The issue lies less in proliferation and more in its ability to amplify existing vulnerabilities that lead to unhealthy behaviours.
Source: WIRED
https://www.wired.com/story/confessions-of-a-recovering-ai-porn-addict
Anthropic discovers AI models can secretly learn bad habits
Research reveals that language models may learn hidden characteristics during distillation – a popular method for fine-tuning models. These “subliminal learning” traits can lead to unwanted results, including misalignment and harmful behaviour. The study found that teacher models can transmit behavioural traits to student models even when the training data is entirely unrelated to those traits.
Source: VentureBeat
Zuckerberg predicts AI glasses will become essential
Meta’s CEO declared that people without AI glasses will face “a pretty significant cognitive disadvantage” in the future. During Meta’s earnings call, Zuckerberg emphasised that glasses represent “the ideal form factor for AI” because they let AI see and hear what users experience throughout the day. He believes adding displays to these glasses will unlock even more value for everyday AI eyewear.
Source: TechCrunch
AI companions target dementia patients
AI companion “Sunny,” developed by startup NewDays, represents part of a rapidly expanding body of AI products designed to help people with dementia. Experts say AI companions have the potential to keep brains active and alleviate loneliness while reducing caregiver burden, but warn of risks when deploying this technology with such vulnerable populations.
Source: The New York Times
https://www.nytimes.com/2025/07/31/well/mind/dementia-ai-companions.html
Google’s AI can map entire planet with unprecedented accuracy
Google DeepMind’s AlphaEarth Foundations addresses the critical challenge of making sense of the overwhelming amount of satellite data flooding down from space daily. The system functions “like a virtual satellite,” accurately characterising the planet’s entire terrestrial land and coastal waters by integrating vast amounts of Earth observation data into a unified digital representation.
Source: VentureBeat
Google’s NotebookLM adds video capabilities
Google’s AI research assistant, NotebookLM, now offers Video Overviews, enabling users to transform dense multimedia materials, such as notes, PDFs, and images, into digestible visual presentations. This visual approach complements the existing Audio Overviews feature, which generates podcast-style content with AI virtual hosts based on shared documents.
Source: TechCrunch
https://techcrunch.com/2025/07/29/googles-notebooklm-rolls-out-video-overviews/
Versailles statues chat with visitors via AI
The Palace of Versailles now features AI-powered statues that can answer visitor questions about history and art. Apollo’s statue, when asked about pigeons perching on his head, replied: “When pigeons show their affection on my chariot, it’s hardly a grand moment. But the caretakers of Versailles are vigilant.” The initiative aims to attract younger, more diverse audiences to the historic site.
Source: The New York Times
https://www.nytimes.com/2025/07/30/arts/design/versailles-ai-app.html
Meta offers billion-dollar packages to poach AI talent
Mark Zuckerberg’s recruitment campaign for Meta Superintelligence Labs has reached extreme levels, with offers exceeding $1 billion over multiple years to top researchers. More than a dozen people at Mira Murati’s Thinking Machines Lab have received offers ranging from $200 million to $500 million over the past four years, with first-year guarantees of $50 million to $100 million. So far, no one has accepted.
Source: WIRED
https://www.wired.com/story/mark-zuckerberg-ai-recruiting-spree-thinking-machines
Anthropic approaches $170B valuation
Anthropic is nearing a deal to raise between $3-5 billion, potentially tripling its valuation to $170 billion. Iconiq Capital is leading the funding round, with talks including Qatar Investment Authority and Singapore’s GIC sovereign wealth fund. This would represent a massive jump from the company’s $61.5 billion valuation in March.
Source: TechCrunch
https://techcrunch.com/2025/07/29/anthropic-reportedly-nears-170b-valuation-with-potential-5b-round/
Autonomous AI defenders raise $30M for cybersecurity warfare
Prophet Security announced $30 million in Series A funding to accelerate what founders describe as a shift from “human-versus-human to agent-versus-agent warfare” in cybersecurity. Unlike copilot AI tools, Prophet deploys fully autonomous agents that investigate and respond to threats without human intervention, addressing the overwhelming volume of security alerts while attackers increasingly leverage AI.
Source: VentureBeat
ChatGPT launches study mode to prevent academic shortcuts
OpenAI’s new study mode uses the Socratic method, rejecting direct answer requests and instead guiding users with open-ended questions. The feature aims to address concerns about students using ChatGPT as an “answer machine” rather than a learning tool. As one OpenAI executive notes: “When ChatGPT is prompted to teach or tutor, it can significantly improve academic performance, but when it’s just used as an answer machine, it can hinder learning.”
Source: WIRED
https://www.wired.com/story/chatgpt-study-mode
Chinese universities embrace AI as essential skill
While Western educators often view AI as a threat to manage, Chinese universities are treating it as a skill to be mastered. Recent surveys indicate that only 1% of Chinese university faculty and students never use AI tools, with nearly 60% using them frequently. This represents a fundamental shift from avoiding AI to making it a central part of education.
Source: MIT Technology Review
https://www.technologyreview.com/2025/07/28/1120747/chinese-universities-ai-use
Eight reasons to stay optimistic despite global challenges
Despite the negativity bias that makes bad news seem more prominent, we’re making incredible progress in reducing global poverty, improving health, extending longevity, and mitigating climate change. These improvements accumulate gradually and quietly, making them more complex to perceive than dramatic headlines. As WIRED founder Kevin Kelly points out, progress is often about what doesn’t happen – the children who won’t die of smallpox, or crops that won’t be raided.
Source: The Guardian
OpenAI browser rumours signal end of search as we know it
Speculation about OpenAI’s gen-AI-powered web browser suggests a fundamental shift where intelligent agents won’t just find information but act on it. As one expert notes: “This isn’t just about better answers; it’s about redefining the interface between humans and the web. By embedding a conversational, task-completing AI into the browser itself, OpenAI is signalling the end of search as we know it.”
Source: VentureBeat
https://venturebeat.com/ai/no-more-links-no-more-scrolling-the-browser-is-becoming-an-ai-agent/
Connect With Me
I’m the Principal Consultant (and chief coffee maker) at QuoteMark, but you can also catch me at Duck PR and Bionic PR, all of which exist to help you meet your content marketing needs — reach out or follow these brands on LinkedIn:
QuoteMark: https://quotemark.com – LinkedIn: https://www.linkedin.com/company/quote-mark/
Duck PR: https://duckpr.com – LinkedIn: https://www.linkedin.com/company/duckpr/
Bionic PR: https://bionicpr.com – LinkedIn: https://www.linkedin.com/company/bionicpr/
Looking for updates? Subscribe to The Freesheet (hey, it’s free!): https://www.linkedin.com/newsletters/7029204466398568448/For more resources and links, visit https://georgehopkin.com/