AI News Update | Google Expands AI-Powered Accessibility Features!

Hello from the world of AI! 🚀In today's newsletter, we'll cover the following key topics.・Google expands AI-powered accessibility features: image descriptions, on-screen content read-aloud, and more・OpenAI rolls out GPT-4.1 and GPT-4.1 mini in ChatGPT — faster processing and improved coding performance・Perplexity partners with PayPal — to offer AI-integrated shopping experiences in the US

GOOGLE

Google expands AI-powered accessibility features: image descriptions, on-screen content read-aloud, and more

Google announced new AI accessibility features for Android and Chrome, integrating its AI assistant "Gemini" into "TalkBack," the screen reader for visually impaired users, so users can ask direct questions about what's on screen.TalkBack's 2024 image description feature could already describe images lacking alt text. Now users can ask follow-up questions about an image — for example, a guitar photo's brand, color, or other details.Users can also ask questions about any content on their smartphone screen, not just images — useful, for instance, when checking a product's material or sale details in a shopping app.

Source: Google

Real-time emotion captions "Expressive Captions" get an upgrade: now capturing vocal nuance

"Expressive Captions," Android's real-time captioning feature, now visualizes emotional expression alongside what a speaker says.The AI recognizes speech traits like drawn-out vowels or elongated word endings and reflects them in the captions — for example, "amaaaazing shot" or "nooooo" is transcribed exactly as spoken. Non-verbal sounds such as whistling or throat-clearing are now also recognized and shown as labels.The feature is currently available on Android 15+ devices in four countries: the US, UK, Canada, and Australia.

Source: Google

Accessibility improvements for PDF and Chrome too: better scanned-document reading and zoom features

Optical character recognition (OCR) has been added to desktop Chrome, giving access to text inside scanned PDFs that screen readers previously couldn't handle. Users can now select, copy, and search PDF text just like a regular web page.Chrome for Android's page zoom has also been revised so enlarging text no longer disrupts the page layout, with zoom levels set individually — applied to all sites at once or to specific pages only. It's accessible from Chrome's top-right three-dot menu.

OPEN AI

OpenAI rolls out GPT-4.1 and GPT-4.1 mini in ChatGPT — faster processing and improved coding performance

On Wednesday, OpenAI announced two new ChatGPT models, "GPT-4.1" and "GPT-4.1 mini," via its official X (formerly Twitter) account.Per OpenAI spokesperson Shaokyi Amdo, GPT-4.1 is especially useful for software engineers using ChatGPT to write and debug code — faster than the previous "o series," with better instruction comprehension and coding performance.GPT-4.1 is available to ChatGPT's paid plans (Plus, Pro, Team), while the lightweight GPT-4.1 mini is available to all users, including free users. GPT-4.0 mini will be discontinued from ChatGPT as a result.GPT-4.1 and GPT-4.1 mini had already been available as a developer API preview since April. At that time, some AI researchers criticized the lack of a safety report. OpenAI responded that while GPT-4.1 beats GPT-4o on performance and speed, it isn't a "frontier model," so it doesn't need a report of the same level.

Source: Interconnects.ai

Johannes Heidecke, OpenAI's head of safety systems, explained on X that "GPT-4.1 doesn't bring new interaction methods or intelligence, and isn't smarter than o3 — that's why its safety positioning differs."The same day, OpenAI launched a "Safety Evaluations Hub" to regularly publish safety evaluations of its models, signaling a commitment to greater transparency.This update comes amid growing interest in AI coding tools. OpenAI is reportedly in talks to acquire the AI coding platform "Windsurf" at a valuation around $3 billion.Also that day, Google updated its "Gemini" chatbot to connect more easily with GitHub projects.

PERPLEXITY

Perplexity partners with PayPal — to offer AI-integrated shopping experiences in the US

Perplexity, the AI chatbot and search tool, announced a PayPal partnership on Wednesday. Starting this summer, Perplexity Pro users in the US can search for and buy products via the chat feature, paying with PayPal or Venmo.Per PayPal's announcement, this fits the concept of "agentic commerce," where AI handles the actual transaction, not just purchase assistance. Beyond payment processing, PayPal will also handle shipping, order tracking, and invoicing.Perplexity isn't first with AI shopping — Visa and Mastercard already use AI to analyze purchasing trends and personalize experiences with consumer consent.

Source: Interconnects.ai

That said, handling credit card and personal data online requires caution. Amid concerns over data protection and breach risk, PayPal says it will implement fraud detection systems and data security protocols.Visa and Mastercard are currently advancing card tokenization via one-time passwords (OTP) to boost payment security, though it's unclear whether PayPal will adopt a similar mechanism.Whether you stick with familiar shopping habits or try new technology, basic online security matters: use a VPN to protect your IP address and personal information, and check your bank account regularly to catch suspicious transactions early.

In Closing

We hope today's content helps with your AI adoption efforts and information gathering. If you found it useful, please feel free to share it with interested colleagues or friends. We'll keep curating the most noteworthy topics for you going forward. Look forward to our next

news update

. Thank you for your continued support.