Deutsche Telekom boss's 'AI optimism' isn't shared

lightreading.com Oct 6, 2026

Deutsche Telekom boss’s ‘AI optimism’ isn’t shared

Curated by Shubham · TechCapsules · Oct 6, 2026

AI-summarised brief · reviewed before publication

Deutsche Telekom CEO Timotheus Höttges showcased a staged AI‑generated video at the company’s AI investor day in Bonn, depicting himself in historic German moments and on a telecom mast. He used the clip to illustrate AI’s ability to replicate language, images and movement, calling himself an “AI optimist.” The presentation coincided with a KPMG survey indicating only one‑third of Germans trust AI‑generated information. Investors reacted skeptically to the operator’s AI strategy, and employee sentiment appears less enthusiastic than management’s. Höttges’ flamboyant display highlighted a growing gap between executive confidence in AI’s profit potential and broader public and workforce apprehension about the technology’s reliability and impact.

💡 Why It Matters

· The disconnect between leadership’s AI hype and stakeholder doubt could stall Deutsche Telekom’s rollout of AI‑driven services, jeopardizing its competitive edge in a rapidly digitising market.

AI Capsules Entrepreneurs

Read full article on lightreading.com

OpenAI to Introduce Invisible Watermarking for ChatGPT Text in EU to Meet…

OpenAI announced that it will deploy an invisible watermark, called textGrain, on eligible ChatGPT and Codex outputs in the European Union within the next few weeks. The watermark embeds a statistical signal in word choices, allowing machine‑readable labeling without altering the visible text. Detection works best on longer passages and struggles with technical content or edited text, such as synonym replacements. The rollout aligns with the EU AI Act’s transparency requirements, while the detection tool will initially be available only to approved researchers. API users worldwide can opt in, but the feature is disabled by default.

💡 The move gives regulators a concrete tool to trace AI‑generated content, potentially reshaping how publishers verify authenticity in the EU.

independent.co.uk · Oct 6

Female AI agents ‘paid’ less than males, study finds

A University of Limerick study found that in a virtual‑reality office, participants paid male‑presenting AI agents about 10 % more than female‑presenting agents, despite identical underlying technology. The experiment involved 189 users interacting with a male bot named Johan and a female bot named Johanna. Johan was also perceived as more human‑like. Researchers argue the findings mirror existing workplace gender bias and warn that AI design choices can unintentionally reinforce inequalities. The study will be presented at the 14th Nordic Conference on Human‑Computer Interaction.

💡 The results reveal that gender bias can seep into AI systems, potentially perpetuating pay disparities as AI assistants become ubiquitous in business environments.

thenextweb.com · Oct 6

Zoom CTO XD Huang on bringing Steve Jobs to every meeting

Zoom’s chief technology officer, XD Huang, announced that the company is opening the speech‑recognition engine behind its My Notes feature to external partners through the Scribe API. Huang, who spent three decades at Microsoft, explained that Zoom’s AI strategy—termed “C2C” (conversation to completion)—aims to finish meeting tasks before the session ends. The Scribe API, which powers My Notes, ZoomMate, and customer‑support tools, remains top‑ranked on the Open ASR Leaderboard and integrates with Google Drive, Microsoft, Slack, Jira, and Salesforce. Zoom is in talks with AI‑note‑taking firms considering a switch to its technology.

💡 By licensing its advanced speech‑recognition platform, Zoom could set a new industry standard for real‑time meeting automation, potentially reshaping how businesses capture and act on meeting content.

techcrunch.com · Oct 6

Silicon Valley’s AI wunderkind launches Underdog, the most private Instinct/Muse competitor yet

Self‑taught coder Sigil Wen, a former Thiel Fellow who hacked alongside Andrej Karpathy, Aravind Srinivas and Noam Brown, launched Underdog on Monday. The assistant runs entirely on‑device on macOS and Windows PCs, with other platforms coming soon. Powered by Wen’s Husky inference engine, Underdog uses a 27‑billion‑parameter model fine‑tuned from Qwen‑3.8‑27B and matches benchmark results of Claude Opus 4.6. It encrypts account keys and keeps all user data local, removing cloud processing. Underdog will be free initially and will earn a tiny fee on payments it facilitates through Stripe, avoiding ads or data‑selling models. Backed by Andreessen Horowitz, Khosla Ventures, Stripe co‑founder Patrick Collison and other angels, Conway Research pitches the product as a privacy‑first AI assistant for everyday personal use.

💡 By proving that a capable AI can run locally without monetizing user data, Underdog challenges the prevailing cloud‑centric business model and forces larger players to confront privacy as a competitive differentiator.

wired.com · Oct 6

Kevin Roose Didn’t Use AI to Write His Book About AI

Kevin Roose’s new book, The AGI Chronicles, offers a deep‑dive into the decade‑long development of generative AI, focusing on the rivalries and breakthroughs at OpenAI, Anthropic and Google. Drawing from more than 150 interviews with insiders, Roise chronicles the technical, cultural and financial forces that shaped today’s chatbots and agents, positioning the work as a historical record rather than a hype‑driven narrative. The book arrives as Roise and fellow journalist Casey Newton launch Machine Gods Media, a venture that includes a podcast distributed by NPR and aims to provide clear‑eyed coverage of AI’s capabilities and risks. In the interview, Roise explains that the new platform seeks to fill gaps in mainstream tech reporting by delivering nuanced, entertaining analysis amid rapid AI advancement.

💡 By documenting AI’s hidden backstory, the book equips policymakers and the public with context needed to assess emerging technologies beyond headline‑grabbing demos.

analyticsinsight.net · Oct 6

Cloud Infrastructure Trends 2027: Future of Servers, Storage,Networking

AI is driving a fundamental redesign of cloud infrastructure as providers shift from general‑purpose servers to specialized accelerators and Arm‑based CPUs that promise higher performance per watt. The International Energy Agency warns that data‑center electricity consumption could exceed 945 TWh by 2030, with AI‑focused facilities alone expected to quadruple their power demand. This energy constraint is reshaping site selection, forcing attention to grid capacity, cooling and power‑efficient networking such as 400‑800 Gbps links and emerging 1.6 Tbps Ultra Ethernet. Technologies like Compute Express Link enable memory pooling across servers, while flash storage and object stores handle massive AI training datasets. By 2027, cloud efficiency will hinge on the integration of power, data movement and compute rather than sheer compute scale for future workloads.

💡 Power availability, not raw processing power, will become the decisive factor in where and how AI‑driven cloud services can grow, giving providers with reliable grid access a clear edge.