AI News Recap: August 15, 2025
Hinton's Warning, Perplexity's Gambit, and GPT-5's Stumble: AI News Roundup
Warnings, Wagers, and a $34 Billion Browser Bid
The AI world delivered drama this week with Geoffrey Hinton sounding alarm bells about superhuman AI arriving sooner than expected, while Perplexity made headlines with an audacious $34.5 billion offer to buy Chrome from Google. From OpenAI's GPT-5 stumbling out of the gate to Apple plotting an animated Siri takeover of your smart home, here's everything that happened in AI's latest rollercoaster week.
⚠️ Note:
Good morning! That's right, it's Friday! Hope you are well.
2 things I wanted to update you on ...
1. Each week, I will be including a Word Find puzzle at the end of the post. In the future, I will be adding more. Eventually, when I get https://www.neuralbuddies.ai up-and-running, these games will be interactive. For the time being, you can edit the image on your phone and use the highlighter feature or pen to circle your words.
2. All news posts in the past were results of an automated script I had written. This was more of a personal project itch I needed to scratch. While it did work and the results were ok, I have decided to start hand-picking articles to give you more variety and relevancy. Have a great weekend and tune in on Sunday for our next post!
Table of Contents
👋 Catch up on the Latest Post
🔦 In the Spotlight
🗞️ AI News
🧩 Weekly “Word Find” Puzzle
👋 Catch up on the latest post …
🔦 In the Spotlight
AI ‘Godfather’ Geoffrey Hinton Warns Machines May Outthink Humans in Years
Category: AI Safety & Policy
⚠️ Geoffrey Hinton warns that artificial general intelligence could arrive within a few years and surpass human intelligence, accelerating beyond prior 30–50 year estimates.
🧠 He urges building “maternal instincts” into advanced AI so systems care for and protect humans, advocating international cooperation on safety-focused alignment.
🏥 While cautioning about risks, Hinton notes potential benefits in healthcare, including faster diagnoses, targeted drug development, and personalized treatments.
Apple’s Plan for AI Could Make Siri the Animated Center of Your Smart Home
Category: Consumer Devices
🤖 Apple is developing a tabletop robot with a movable arm and a more visual, conversational Siri powered by LLMs, targeting a 2027 launch.
🏠 A smart home display is planned for mid-2026 to control the home, play media, take notes, and do video calls, potentially featuring the new animated Siri and multi-user personalization via face scanning.
🔒 Apple is building a new lineup of home-security hardware, including multiple types of security cameras integrated with its smart home ecosystem.
GPT-5: Overdue, Overhyped and Underwhelming. And That’s Not the Worst of It.
Category: Policy and Critique
🧪 GPT-5’s debut is portrayed as a major letdown marked by persistent errors, subpar benchmark results, and a swift drop in perceived leadership, contradicting claims of “PhD-level” expertise.
📉 The article argues OpenAI’s technical edge and valuation narrative are eroding amid competition, staff departures, price cuts, and fraying partnerships, with GPT-5 offering only incremental gains.
🧠 A new Arizona State University study is cited to claim LLM “chain-of-thought” reasoning is brittle under distribution shift, reinforcing the case that pure scaling won’t reach AGI and advocating neurosymbolic approaches with explicit world models.
Sam Altman, OpenAI Will Reportedly Back a Startup That Takes On Musk’s Neuralink
Category: Neurotech & Brain-Computer Interfaces
🧠 Sam Altman is co-founding Merge Labs, a brain-computer interface startup reportedly seeking funding, potentially from OpenAI’s ventures arm; talks are early and not yet finalized.
🧩 Merge Labs would compete with Elon Musk’s Neuralink, which is in human trials for paralysis patients and recently raised a large Series E at a multibillion valuation.
🤝 The effort reportedly involves collaboration with Alex Blania of Tools for Humanity, linking the project to broader identity and human-tech integration themes previously discussed by Altman as “The Merge.”
Perplexity Offers to Buy Chrome for Billions More Than It’s Raised
Category: AI Strategy
💰 Perplexity made an unsolicited $34.5B cash offer to acquire Google’s Chrome, pledging $3B in investments to keep Chromium open source and maintain the project’s development.
🔧 The proposal commits to preserving Chrome’s existing user defaults, including keeping Google as the default search engine rather than switching to Perplexity’s AI search.
⚖️ The bid follows DOJ proposals that Google divest Chrome after antitrust rulings on search and adtech; Google has not agreed to sell and is contesting the cases.
Anthropic Takes Aim at OpenAI, Offers Claude to ‘All Three Branches of Government’ for $1
Category: Government & Enterprise AI
🏛️ Anthropic will offer Claude access to all three branches of the U.S. government—executive, legislative, and judiciary—for $1 for one year, expanding beyond OpenAI’s executive-branch-only offer.
🔐 The offering includes Claude for Government with FedRAMP High support and multicloud access via AWS, Google Cloud, and Palantir to handle sensitive unclassified workloads.
🤝 Anthropic will provide technical support to help agencies integrate AI into workflows, citing existing deployments at Lawrence Livermore National Laboratory and the DC Department of Health.
Apple’s New Siri May Allow Users to Operate Apps Just Using Voice
Category: Tools & Platforms
🗣️ Apple is testing a Siri upgrade that can take actions across third-party apps via voice commands, powered by a new version of App Intents integrating with systems like Search and Shortcuts.
📱 Reported capabilities include multi-step tasks such as finding a photo, editing it, and sending it; posting comments; and logging into services.
🤝 Apple is piloting the feature with apps including Uber, AllTrails, Threads, Temu, Amazon, YouTube, Facebook, and WhatsApp, with a broader Siri overhaul targeted for spring 2026.
Alan Turing Institute: Humanities Are Key to the Future of AI
Category: AI Ethics & Regulation
🧭 The Alan Turing Institute and partners launch “Doing AI Differently,” advocating a human‑centred, humanities-informed approach to AI development focused on interpretive depth.
🧩 The report warns of a “homogenisation problem” in current AI architectures and calls for alternative designs and Interpretive AI that handle ambiguity and multiple perspectives.
🤝 The initiative promotes human‑AI ensembles and announces a UK–Canada funding call to build safer, context-aware systems aligned with societal values.
Suvianna Grecu, AI for Change: Without Rules, AI Risks ‘Trust Crisis’
Category: AI Ethics & Regulation
⚠️ Suvianna Grecu warns that rapid AI deployment without strong governance could trigger a “trust crisis,” enabling “automating harm at scale” across critical sectors.
🧭 She calls for concrete accountability measures—pre‑deployment risk assessments, design checklists, and cross‑functional review boards—to embed ethics into daily development workflows.
🤝 Grecu advocates joint responsibility between government and industry: regulation to set minimum standards and industry to innovate safeguards, with values like human rights and transparency built into AI by design.
Anthropic’s Claude Chatbot Can Now Remember Your Past Conversations
Category: Tools & Platforms
🧠 Anthropic launched an opt-in memory feature for Claude that can search and summarize your past chats on request, without building a persistent user profile.
💻 The feature works across web, desktop, and mobile, supports separating projects/workspaces, and is rolling out to Max, Team, and Enterprise now, with broader access coming soon.
🔧 Users can enable it via Settings > Profile > “Search and reference chats”; Claude retrieves prior conversations only when asked, unlike always-on memory in some competitors.
The GPT-5 Fallout, Explained...
Category: Business & Market Trends
🔄 OpenAI’s sudden sunsetting of GPT-4o at GPT-5 launch triggered widespread user backlash on Reddit and X, prompting a rapid reversal to restore 4o access for Plus subscribers while free users remained excluded.
🧩 Mixed performance reactions stemmed from GPT-5 being a family of models with variable behavior; OpenAI acknowledged underestimating user attachment to 4o’s “warm” personality and plans to make GPT-5 warmer with per-user customization.
🏁 The episode highlights intense competitive parity across top labs (OpenAI, Anthropic, Google, xAI), with models converging in capabilities; speculation also arose that the move could boost paid conversions ahead of potential revenue goals.
