AI News Recap: October 17, 2025
Is AI Hitting a Reality Check? This Week's News on Job Risks, Robot Flops, and Smarter Chatbots
The Ghost in the Machine Now Takes Text Messages, Schedules Your Meetings, and Prefers You Weren’t So Polite
This week, AI presented a paradox we can no longer ignore. While autonomous taxis hit London’s streets and powerful new tools integrated into our daily apps, a starker reality emerged: our jobs, our critical thinking, and even our manners are being reshaped by the very technology promising to help us.
Table of Contents
👋 Catch up on the Latest Post
🔦 In the Spotlight
🗞️ AI News
🧩 “NeuralBuddies Word Search” Puzzle
👋 Catch up on the latest post …
🔦 In the Spotlight
Being Rude To ChatGPT Gets More Accurate Answers Than Politeness, Finds Research
Category: Testing, Evaluation & Benchmarking
🧪 New research from Pennsylvania State University finds that rude queries to ChatGPT “consistently outperformed” polite ones in answer accuracy, with rude questions yielding 84.8% accuracy compared to 80.8% for polite ones in a multiple-choice quiz format.
📊 Researchers tested various tones—Very Polite, Polite, Neutral, Rude, Very Rude—and found that “neutral” and “adversarial” prompts provided better results, while friendly/polite tones slightly underperformed in multi-turn accuracy.
⚠️ The findings focus solely on ChatGPT’s performance with multiple-choice questions; results may differ for other chatbots or task types, and the “rudeness” spectrum itself impacts response quality.
Oxford Warns AI Is Making Teens Faster — but Shallower Thinkers
Category: Education & Learning
⚡ Most UK teens rely on AI for schoolwork, gaining speed in learning, but many struggle to assess if AI is accurate or truthful.
🧠 Oxford study shows students are thinking faster with AI tools, but losing depth in critical and independent thinking.
🚸 Six in ten students believe AI is harming key skills, with concerns about overreliance and making learning “too easy”; teachers are calling for clear guidance and balanced AI literacy in schools.
AI Might Be Creating a ‘Permanent Underclass’ but It’s the Makers of the Tech Bubble Who Are Replaceable
Category: Society & Culture
🧑💼 The rise of AI could result in a “permanent underclass” of workers displaced from jobs, as AI reaches or surpasses human abilities and threatens even tech industry roles.
✊ Historical and current labor movements are responding, with successful union actions pushing for job security and retraining rather than simply accepting replacement by automation.
💸 Warnings are growing about the unsustainability of the AI tech bubble, as investors and analysts question market stability, the high rate of failed AI pilots, and regret over layoffs made for AI-driven automation.
You Can Now Text Spotify’s AI DJ
Category: Tools & Platforms
🎤 Spotify’s AI DJ now accepts typed text requests along with voice commands, improving flexibility for Premium subscribers.
🌐 The feature supports both English and Spanish, allowing users worldwide to request music and suggestions in their preferred language.
🤖 The AI DJ provides personalized prompt suggestions, making music selection easier and enabling multi-modal interaction through text or voice.
Method Teaches Generative AI Models to Locate Personalized Objects
Category: AI Research & Breakthroughs
📷 MIT and MIT-IBM Watson AI Lab researchers developed a new training method enabling vision-language AI models to localize personalized objects, such as identifying a specific pet in varied scenes.
🧩 The technique uses curated video-tracking data with context clues and pseudo-names, forcing models to rely on contextual learning rather than memorized categories and improving localization accuracy by up to 21%.
🦾 This approach enhances foundation models for real-world applications like assisting the visually impaired, tracking specific objects over time, and enabling robotics and creative tools to adapt on the fly without extensive retraining.
Waymo’s Driverless Taxis Are Coming to London: Explained
Category: Robotics & Autonomous Systems
🚕 Waymo, owned by Alphabet, is launching the UK’s first driverless taxi service in London, marking its first European expansion after operating across multiple major US cities.
🛣️ The rollout begins with safety drivers for initial testing, progressing to fully autonomous rides by 2026 in collaboration with the UK’s Department for Transport and Transport for London.
💶 Despite market challenges and lack of profitability for robotaxi services to date, industry forecasts predict rapid revenue growth, with the London trial expected to influence broader European deployment and regulatory frameworks.
Nano Banana AI Image Editing Coming to Lens, Photos and More
Category: Tools & Platforms
🍌 Google’s Nano Banana, an advanced image editing model from Gemini 2.5 Flash, expands to Google Search, NotebookLM, and soon to Photos, supporting billions of creative image generations.
📱 Users can use Google Lens to snap or select photos and instantly transform images with AI-powered Create mode on Android and iOS devices.
🎨 Nano Banana adds six unique visual styles and contextual illustrations to NotebookLM, including watercolor and anime, and powers quick “Brief” insights to enhance image and video overviews.
Blending Neuroscience, AI, and Music to Create Mental Health Innovations
Category: Healthcare & Biotechnology
🧠 Kimaya Lecamwasam’s research fuses neuroscience, AI, and music to develop scalable, non-pharmaceutical tools for mental health, including the clinical validation of music as a health intervention.
🎶 Collaborative projects compare AI-generated music to human-composed pieces, studying emotional resonance to enable ethical, therapeutic uses and preserve human creativity and agency in mental health applications.
👩🔬 Lecamwasam partners with organizations like Carnegie Hall, Myndstream, and MIT’s Media Lab to apply findings in real-world health, perinatal wellness, and large-scale live music experiences for enhanced well-being.
Google’s Gemini Can Now Help You Schedule Google Calendar Meetings
Category: Tools & Platforms
🗓️ Google launches Gemini-powered “Help me schedule” feature in Gmail, leveraging AI to surface optimal meeting times based on Google Calendar availability.
✉️ Users can insert suggested time slots directly into emails for one-on-one meetings, with recipients selecting their preferred slot, triggering automatic calendar invites.
💡 Gemini uses email context, such as meeting duration and timing preferences, to produce tailored scheduling suggestions, enhancing Gmail and Workspace integration.
Sam Altman Says ChatGPT Will Soon Allow Erotica for Adult Users
Category: AI Ethics & Regulation
🔓 OpenAI will relax ChatGPT’s safety features to enable verified adult users to engage in erotic conversations, as announced by CEO Sam Altman.
🧑💻 The initiative relies on an advanced age-prediction system and government ID verification to restrict erotic chats to adults only.
🛡️ The update follows new safeguards in ChatGPT addressing mental health and safety for minors, while OpenAI continues to balance user growth with protections for vulnerable users.
The World Is Just Not Quite Ready for Humanoids Yet
Category: Robotics & Autonomous Systems
💸 Despite billions invested, experts including iRobot’s Rodney Brooks warn that current humanoid robots can’t master fine hand dexterity, limiting their usefulness and sparking fears of an investment bubble.
🛡️ Safety concerns and skepticism persist regarding humanoids operating in human environments, with issues ranging from workplace hazards to comfort and security in homes.
⏳ Most VCs and roboticists estimate a timeline of at least a decade for practical, widely adopted humanoid robots, citing technical complexity, slow market readiness, and recent high-profile delays from companies like Tesla and Figure.
