AI Daily Briefing — 2026-05-23

At Google I/O, DeepMind CEO Demis Hassabis declared humanity stands at the “foothills of the singularity,” a moment when AI surpasses human intelligence. Yet the context—a segment on scientific AI showcasing WeatherNext, which provided early warnings for Hurricane Melissa in Jamaica—highlighted a gap between grand promises and practical achievements. While WeatherNext may have saved lives, it remains a specialized tool, not evidence of an imminent singularity, according to MIT Technology Review AI.

The keynote underscored a growing divide in AI for science. One path focuses on targeted tools like WeatherNext or AlphaFold, designed for specific problems. The other envisions agentic, LLM-based systems that could conduct research autonomously. Pushmeet Kohli of Google Cloud recently wrote that AI is moving from facilitating science to doing it. This shift makes it harder to justify massive investments in specialized tools, even Nobel-winning ones, as agentic systems promise a future where humans and AI collaborate as peers.

Google isn’t abandoning specialized tools—AlphaGenome and AlphaEarth were released last summer, and WeatherNext updated in November. AlphaFold remains widely used by over three million researchers. However, signs of realignment are emerging. John Jumper, AlphaFold’s Nobel-winning Google fellow, now works on AI coding, a field where Google has lagged behind rivals like Anthropic and OpenAI. This move may reflect a prioritization of agentic science, as coding abilities are critical for autonomous research systems, MIT Technology Review AI notes.

Across the industry, agentic systems show promise. OpenAI recently announced a breakthrough with an AI that autonomously performed research. As resources shift toward these systems, the path for AI-driven science is evolving, balancing specialized tools with the potential for fully autonomous discovery.

As political support for climate initiatives wanes under the second Trump administration, many climate tech companies are pivoting to critical minerals to stay afloat. Instead of solely pursuing decarbonization, these firms are now highlighting their roles in producing metals like niobium and tantalum, which are in demand for defense and high-tech industries. This strategy allows them to attract investment and generate revenue while navigating a challenging regulatory landscape.

Boston Metal, known for its low-emission steel production, recently raised $75 million to expand its critical metals business. The company’s molten oxide electrolysis technology can produce metals such as chromium and vanadium, which are deemed critical by the U.S. government. CEO Tadeu Carneiro explained that focusing on these metals enables faster deployment and generates resources to eventually tackle steel decarbonization.

Similarly, cement startup Brimstone is now emphasizing its production of alumina and supplementary cementitious materials after losing federal funding. The company’s website now highlights critical minerals alongside cement. Other firms, including carbon removal startups, are also targeting the mining industry, offering efficiency improvements or cleanup services for abandoned sites.

This broader messaging shift reflects a trend where climate goals are downplayed in favor of politically popular topics like energy abundance and critical minerals. While this approach may help companies survive, some worry it could dilute their climate missions. For now, the pivot offers a lifeline, but its long-term impact on emissions remains uncertain.

Google's recent overhaul of its search engine, which prioritizes AI-generated summaries over traditional links, has led to an unexpected glitch. Typing the word "disregard" into the search bar now produces a large empty space where the AI response should be, followed by a single link to Merriam-Webster buried far below. The AI summary offers no useful information, leaving users with a broken experience. Social media users have criticized the change, highlighting how the new layout can fail even for simple queries.

In contrast, a search for the same term on Bing, which has been less aggressive with AI integration, provides more helpful results. While not perfect, Bing's layout includes relevant dictionary entries and contextual information without the wasted space. This marks a rare instance where Bing outperforms Google, a reversal noted by a veteran tech journalist with nearly fifteen years of experience.

The incident underscores the challenges Google faces as it rolls out AI summaries across its massive search platform. While the company aims to enhance user experience, edge cases like this reveal gaps in the system. As Google continues to refine its approach, users may encounter more such quirks, prompting some to explore alternative search engines.

Source: TechCrunch AI

At this week’s Google I/O conference, attendees got a hands-on look at the company’s upcoming AI-powered glasses, which combine audio and visual features through an in-lens display. Unlike the audio-only model set to ship this fall, these Android XR glasses overlay helpful information—such as weather updates, walking directions, Uber pickup details, and live translations—directly onto the real world. Users can even create custom widgets using AI. The glasses will pair with both iOS and Android devices, marking a step beyond the first-generation audio version.

The prototype, developed with Warby Parker, Gentle Monster, and Samsung, prioritizes display technology and battery life over cosmetic details. The test units lacked head-detection sensors and featured a basic frame, but the final version will include this capability. To activate Google’s Gemini assistant, users press and hold the right side of the frame for two seconds, which also starts the camera in the demo—though the shipping version will allow users to toggle this setting.

Testing revealed mixed results. Playing music via Gemini was difficult to evaluate due to noisy surroundings, but the glasses seemed better suited for casual listening than replacing high-quality earbuds. A photo capture test worked seamlessly, transferring images to a paired phone or watch. Video capture was not available in the prototype. While the glasses show potential, they remain a work in progress, with Google focusing on refining the experience before a broader release.

Google's recent overhaul of its search engine, announced at Google I/O 2026, has shifted the platform toward a conversational, AI-driven experience. The company now offers an AI mode by default, complete with chat boxes and AI Overviews that summarize results. While Google touts this as the biggest upgrade to its search box in 25 years, many users are frustrated by the forced integration of AI agents and chatbots, especially after earlier missteps like the infamous advice to stare at the sun. Critics argue this change, combined with Google's monopoly ruling in 2024, makes now the perfect time to explore alternatives.

One standout option is Kagi, a paid, ad-free search engine that costs $5 to $10 per month. It offers customizable features like "lenses" to filter results by category, such as academic articles, and an optional AI-powered "Quick Answer" tool. Unlike Google, Kagi lets users disable AI summaries entirely, putting control back in their hands. For those who prefer a free service, DuckDuckGo provides privacy-focused searching without tracking user data, instead serving ads based solely on the current search topic.

Other alternatives include Brave Search, which blends its own index with anonymous contributions, and Ecosia, which uses ad revenue to plant trees. For a more experimental approach, Open Web Engine offers a chaotic, unfiltered web crawl. As Google becomes increasingly AI-centric, these six search engines provide diverse options for users seeking simplicity, privacy, or customization without the chatbot overload.