AI news you can't miss this week

Microsoft and OpenAI forge restructuring deal, ByteDance launches Nano Banana challenger, Apple downplays AI on new iPhone release, MIT startup unveils brain-reading wearable device & OpenAI discovers why AI models hallucinate.

Best AI news of the week:

1️⃣ OpenAI and Microsoft forge restructuring deal

2️⃣ ByteDance launches Nano Banana challenger

3️⃣ Apple downplays AI on new iPhone release

  • MIT startup unveils brain-reading wearable device

  • OpenAI discovers why model hallucinates and how to reduce them.

WEEKLY AI RECAP

September 8th - 14th, 2025

🤖 OpenAI and Microsoft forge restructuring deal 🤖
It confirms ChatGPT maker will transform into for-profit while preserving nonprofit oversight.

The agreement signed by both parties allows OpenAI to transition into a for-profit structure while ensuring its nonprofit division retains control and secures an equity stake worth over $100 billion. The restructuring represents one of the most significant corporate transformations in AI history, as Microsoft commits to substantial investments in its own AI capabilities while maintaining its OpenAI alliance.

🎨 ByteDance launches NanoBanana challenger 🎨 
Chinese tech giant introduces Seedram 4 to rival Google's latest powerful image model NanoBanana.

ByteDance has unveiled Seedram 4, an advanced multimodal image generation system engineered to compete head-to-head with Google's Gemini 2.5 Flash (Nano Banana). This sophisticated platform merges image creation and editing functionalities, delivering 4K resolution outputs across various aspect ratios while accommodating up to six reference inputs. Seedram 4 seems to substantially surpasses Nano Banana's constraints of 1080p square-format outputs, providing superior performance, speed, and reliability at $30 per 1,000 generations. This launch underscores ByteDance's determined expansion into AI image generation, directly challenging Google's leadership in multimodal AI solutions.

🍎 Apple downplays AI on new iPhone release 🍎
Apple tunes down claims of AI integration into latest iPhones, emphasizing behind-the-scenes AI features.

Apple's iPhone 17 launch event marked a strategic departure from AI-centric marketing, with CEO Tim Cook highlighting how artificial intelligence enhances functionality behind the scenes rather than serving as a headline feature. The 75-minute presentation demonstrated neural engine improvements, on-device large language models for gaming, and neural processing units in GPU cores delivering "MacBook Pro-level performance in a smartphone." Contrasting with previous events that heavily promoted Apple Intelligence, the company concentrated on hardware advancements while showcasing how machine learning drives AirPods heart rate tracking and Apple Watch blood pressure monitoring. This strategic pivot emerges as Apple addresses criticism about lagging in the AI competition, following 10 documented departures from its AI research division.

INTERESTING TO KNOW

🧠 MIT startup unveils brain-reading wearable device 🧠 

MIT spin-off Alterego has revealed a revolutionary wearable technology that interprets neuromuscular signals, allowing users to interact in complete silence. CEO Arnav Kapur demonstrated the device's capability to capture notes and facilitate silent AI conversations without producing any audible sound. This breakthrough in brain-computer interface technology offers practical solutions for productivity enhancement in challenging acoustic environments or long-distance telepathic communications (similar to Ghost In The Shell and Cyberpunk sagas). The innovation holds life-changing potential for individuals with ALS and similar conditions that impair speech function.

🧠 OpenAI discovers why AI models hallucinate and how to reduce them 🧠

OpenAI has published groundbreaking research revealing that conventional training methodologies create an inherent contradiction where AI models develop a tendency to speculate rather than acknowledge uncertainty. Current assessment frameworks award maximum points for fortunate guesses while providing zero credit for honest admissions of ignorance, inadvertently encouraging overconfident speculation. The study discovered that sophisticated models like o3 produce inaccurate information 33% of the time—twice the rate of its predecessor o1. The proposed remedy involves restructuring evaluation systems to impose greater penalties on confident mistakes than on expressions of uncertainty, potentially dramatically enhancing AI dependability and credibility across diverse applications.

Club AI @ H-FARM Campus - September 18, 6:00 PM

H-FARM’s AI Club is a monthly community meetup that brings up to 500 people together to cut through the AI noise with real news, real cases, real debate.

On stage this edition:

  • Diego Pizzocaro (CEO, H-FARM AI): the latest AI news but mixed with our comments, perspectives and your questions.

  • Luca Pezzullo (President, Order of Psychologists of Veneto Region): “AI & Consciousness: Illusion, Reality, and Risks.”

  • Enterprise Voice Agents (CISCO): what really works in large organizations - architecture, rollout, ROI with AI Voice tech.

Language: English (full session)

📩 Have questions or feedback? Just reply to this email , we’d love to hear from you!

🔗 Stay connected: