This Week in AI

Beijing to Introduce AI Education in Schools

Beijing will introduce AI courses in primary and secondary schools starting September 1, 2025, as part of China’s plan to strengthen its position as a global AI leader.

The Beijing Municipal Education Commission announced that students will receive at least eight hours of AI education per year, either as a standalone subject or embedded in existing courses like science and information technology.?

This initiative aligns with China’s broader push to lead AI innovation, following recent advancements by Chinese startups like DeepSeek, which developed an AI model comparable to those of major US companies.

Education Minister Huai Jinpeng highlighted AI’s role in education, with China set to release an AI white paper in 2025. By introducing AI early, Beijing aims to build future talent and reinforce its global leadership.

OpenAI Launches New Tools to Build Smarter AI Agents

On March 11, 2025, OpenAI rolled out the Responses API, a new set of tools aimed at helping developers build their own AI-powered agents.?

Released on March 11, 2025, the new Responses API allows companies to build AI-powered assistants capable of web searches, document analysis, and website navigation. This replaces the existing Assistants API, which OpenAI will phase out in 2026.

The Responses API includes OpenAI’s GPT-4o search and GPT-4o mini search, which achieve 90% and 88% factual accuracy on OpenAI’s SimpleQA benchmark. It also introduces a Computer-Using Agent (CUA) model, allowing AI to automate tasks like data entry and app workflows.

While AI agents have yet to prove their full potential, OpenAI aims to refine its tools and push AI-powered automation forward, positioning itself as a leader in the space.?

Celine Dion to AI: That’s Not My Voice

Imagine thinking you've stumbled upon a brand-new Celine Dion song—only to find out it’s not her at all. That’s exactly what’s happening, and Dion wants her fans to know the truth.

In an Instagram post, the legendary singer warned that AI-generated tracks falsely using her voice and likeness are making the rounds online. “These recordings are fake and not approved,” she stated, making it clear that they are not part of her official discography.

Dion isn’t alone in this fight. Last year, over 200 artists, including Billie Eilish and Jon Bon Jovi, signed an open letter condemning AI deepfakes and voice cloning, calling for stricter protections.

The singer hasn’t released new music since 2019. But when she did take the stage again—perched atop the Eiffel Tower at the 2024 Paris Olympics—it was unmistakably her. No AI required.

MTA Partners with Google to Improve Subway Maintenance

New York’s Metropolitan Transportation Authority (MTA) is expanding its AI-powered track monitoring system by partnering with Google Public Sector to detect and fix subway track issues before they lead to delays.?

Through a partnership with Google Public Sector, the MTA is equipping subway cars with Google Pixel smartphones that use sensors, microphones, and AI models to monitor track conditions in real time.

The program’s initial phase gathered 335 million sensor readings, 1.2 million GPS locations, and 1,200 hours of audio, helping inspectors pinpoint problem areas.

Following its success, the MTA is launching a new pilot program with Google to expand the initiative. With over 42,000 subway delays in December 2024 alone, the agency hopes this AI-driven approach will improve efficiency, reduce disruptions, and improve commuter safety.

Lila Raises $200M to Speed Up Scientific Discovery with AI

Lila Sciences, a Cambridge-based startup, has raised $200 million to advance AI-driven scientific research. After operating in stealth for two years, the company announced its work on March 10, 2025.?

Lila’s AI program is designed to analyze vast amounts of scientific data, including published research and experimental results, to generate new insights and discoveries. Unlike conventional AI models, Lila’s system is trained on the scientific process itself, enabling it to reason, design, and test experiments independently.

Early results include the development of novel antibodies for disease treatment and new materials for carbon capture – breakthroughs that would normally take years, completed in just a few months.

Backed by Flagship Pioneering, Lila aims to transform medicine, clean energy, and materials science by accelerating the hypothesis-experiment-validation cycle. While AI has great potential in scientific discovery, experts emphasize that human oversight remains essential to validate its findings.

Love this newsletter? Go deeper! ??

Join 150k+ readers who get AI insights, women in tech spotlights, polls, and more—3x a week. No jargon, just the best stories straight to your inbox.

Subscribe to Amplify with Dr. Ayesha Khanna now: https://drayeshakhanna.beehiiv.com/subscribe

Until next time! The future awaits.

Ayesha ??