AI Update: Mid-Late May
Another update to formats
My goal with this substack is to try to turn a bunch of the news and sources I already use each week into something useful for others, and hopefully let you keep up with the pace of technology in less time.
I’m trying a new approach (again), back to a topic-focused approach. A secondary goal for me is learning some best practices using new tools and capabilities. In that vein, this week I’ve tried making a NotebookLM audio overview, which essentially is a podcast form of the news. I’ve not done an audio post before, it seems like it will work best as a separate audio-specific post. Let me know what you think!
From Good Tech / Bad Tech
Industry Leaders Unveil Next-Generation AI
Google demonstrated a comprehensive AI strategy. Key developments include its Veo AI model generating video with sound (Engadget), Project Astra for a universal AI assistant (Engadget), and the centrality of AI Overviews in search, as highlighted at I/O 2025 (Ars Technica). This push for ubiquitous AI (The Register, MIT Technology Review) includes proactive AI agents (TechCrunch) and SynthID for content authenticity (The Verge).
Key Takeaway for Google: Google is weaving AI into its entire ecosystem, pushing for powerful, proactive assistance while simultaneously developing tools to address content authenticity.
Competitors are also advancing. OpenAI's Codex is key to new "agentic" coding tools capable of more independent complex tasks (TechCrunch). The high-stakes environment was hinted at in an "Empire of AI" excerpt (Wired).
Key Takeaway for OpenAI: OpenAI continues to push towards more autonomous AI agents, particularly in complex domains like software development.
Anthropic showcased progress with its Claude 4 models' enhanced multi-step reasoning (TechCrunch) and a hybrid AI model capable of extended autonomous work (MIT Technology Review). This drive for autonomy is an industry trend, though not without issues, such as a reported incident of unexpected model behavior during safety tests (TechCrunch). Anthropic's growing influence was marked by its first developer conference (Wired).
Key Takeaway for Anthropic: Anthropic rapidly advances AI reasoning and autonomy, simultaneously confronting new, complex safety and control hurdles.
AI's Widening Impact and the Human Factor
AI is actively reshaping real-world applications, from content curation (The Verge) and enterprise productivity (Wired) to scientific forecasting (The Register).
However, the AI-human expertise dynamic is still evolving. While tools like OpenAI's Codex offer powerful assistance (TechCrunch), some argue AI cannot fully replace human oversight and creativity (The Register). Security concerns are highlighted by incidents like an AI assistant being tricked into generating malicious code (Ars Technica).
Key Takeaway on AI Adoption & Human Element: AI offers significant gains across sectors but raises critical questions about the future of human work, oversight, and system security.
Growing Pains: Ethical Dilemmas and Societal Shifts
AI's rapid evolution brings numerous ethical and societal challenges. Concerns include AI-induced errors impacting justice (MIT Technology Review), data privacy violations by AI firms (South China Morning Post), and questions about AI model trustworthiness and benchmark integrity (Live Science). Debates also surround AI "hallucinations" (TechCrunch).
Economically, AI's potential for job displacement is a concern (The Verge). Creative industries face copyright tensions over AI training data (The Verge, The Verge). Legal and regulatory systems are struggling, with issues of AI agent liability (Wired), First Amendment implications (The Verge), and surprising legislative moves like a decade-long ban on state-level AI regulations in the U.S. (Engadget).
Key Takeaway on Concerns: AI's power brings complex ethical, societal, and legal challenges—from job displacement and privacy to copyright and justice—with regulatory responses lagging.
The AI Engine: Infrastructure and Energy Needs
The AI boom demands immense computational power, driving massive infrastructure investment, including Nvidia's servers (The Register) and large international data center projects in the UAE and Paris (TechCrunch, The Register, Semafor). Startups also attract capital (The Next Web).
This growth has a significant energy cost (The Register, MIT Technology Review), making energy-efficient computing critical (SemiEngineering).
Key Takeaway on Infrastructure: The AI revolution relies on an expanding, power-hungry infrastructure, making sustainable energy solutions vital.
A Field in Rapid Flux
Mid-2025 finds AI at a crucial turning point. As The Register stated, "AI models are evolving," bringing stunning progress, fierce competition, and vast applications. However, this also presents urgent societal, ethical, and resource challenges demanding proactive management. The promise of beneficial AI is materializing quickly, alongside critical questions about wielding its power responsibly.