Welcome to another edition of This Week in AI, your go-to round-up of the latest breakthroughs, industry movements, and regulatory updates in artificial intelligence. As the AI landscape continues to evolve at an astonishing pace, this week has brought news that spans healthcare, semiconductor innovation, regulatory pushback, and new horizons in productivity tools. Let’s dive in.
🧠 DeepMind’s AMIE Surpasses Human Doctors
In a landmark study published this week, Google DeepMind’s AMIE (Articulate Medical Intelligence Explorer) has shown it can outperform doctors in diagnostic accuracy. The AI system was tested across a range of medical scenarios and bested human physicians in both recall and precision metrics. This represents more than just a technical milestone; it could be a paradigm shift in how frontline healthcare is delivered, especially in under-resourced regions.
AMIE’s success builds on large language model (LLM) foundations but layers in domain-specific reasoning and multimodal understanding. As regulatory and ethical frameworks catch up, AMIE could redefine triage, remote consultations, and even the training of new medical professionals.
💜 Profluent Bio Unlocks Protein Design Scaling Laws
Biotech startup Profluent Bio announced a new family of models – ProGen3 – which uncovers that protein design follows similar scaling laws to natural language processing. This discovery might not grab mainstream headlines, but it’s groundbreaking. By treating protein sequences like a language and leveraging massive transformer models, Profluent is creating the tools to engineer enzymes, therapeutics, and diagnostics faster and more reliably than ever before.
This could represent a turning point in synthetic biology, with profound implications for medicine, agriculture, and climate tech.
📱 Google’s AI Email Assistant: Inbox Zero Gets Real
At SXSW London, Google DeepMind revealed an experimental AI email assistant that promises to autonomously manage inboxes and even write replies in your personal tone. The tool uses multimodal inputs and user-specific language modelling to mirror your communication style and prioritise based on historical behaviour.
While still in development, this assistant offers a tantalising vision of what AI-powered productivity could look like. For busy professionals and digital nomads alike, the dream of “Inbox Zero” might finally be within reach.
🛠️ Duolingo Goes “AI-First”
Duolingo, the world’s most popular language learning app, declared this week that it is shifting to an “AI-first” business model. What does that mean? It means AI will be baked into everything: content generation, language translation, user assessment, and even internal HR functions like recruitment and performance reviews.
The shift signifies how AI is moving from being a product feature to becoming the foundation of entire company operations. As other edtech and consumer platforms take note, expect more businesses to follow suit.
🚀 Broadcom and AMD: The AI Hardware Race Heats Up
Two major moves in the AI hardware space made headlines this week:
- Broadcom reported record demand for its networking and custom AI chips, driven by hyperscalers and data centres scaling large model inference. Their new Tomahawk 6 chip pushes the envelope on interconnect speeds, a critical factor for distributed training.
- AMD acquired the team from Untether AI, a startup known for its energy-efficient AI inference chips. This acqui-hire strengthens AMD’s positioning in the low-power AI hardware market and complements its existing GPU and CPU lines.
These developments underscore how essential custom silicon has become to competing in the AI arms race.
🌍 Saudi Arabia Bets Big with Humain
Saudi Arabia launched a new AI entity called Humain under its Public Investment Fund, with an ambition to develop a multimodal Arabic language model and foundational AI infrastructure. This move signals the country’s intent to become a regional AI powerhouse and reflects a broader trend of sovereign-backed AI initiatives.
Expect Humain to focus heavily on culturally and linguistically aligned AI systems, which have been historically underserved by Western-centric models.
🌎 China’s Q1 AI Surge
According to a report this week, China made substantial gains in AI development during Q1 2025. With massive investments in model training, chip design, and localised applications, China is consolidating its position as a key global player.
Notably, this includes an increased focus on AI in manufacturing, logistics, and government services. The pace and scale of these advancements suggest the global AI race is intensifying.
📄 US State Lawmakers Push Back on Federal AI Regulation
In the regulatory sphere, a bipartisan coalition of 260 US state lawmakers signed a letter opposing a proposed 10-year federal moratorium on state-level AI laws. They argue that innovation must be balanced with state sovereignty and local governance.
This highlights the growing tension between national standardisation and regional agility in AI policymaking. It also signals that AI legislation will remain fragmented across the US for the foreseeable future.
🏆 Databricks Recognises AI Trailblazers
The 2025 Data Intelligence Industry Awards, hosted by Databricks, honoured companies like Barclays, Novo Nordisk, and 7-Eleven for exceptional AI innovation. These recognitions validate real-world applications of data and AI to drive measurable business impact.
🎓 Looking Ahead: Apple’s WWDC
All eyes turn to Apple’s WWDC 2025, starting 9 June, where major AI feature updates are anticipated. iOS 26 is rumoured to include a major revamp of Siri, AI-driven image editing tools, and tighter integration of on-device intelligence. Apple’s approach to privacy-preserving AI will be one to watch closely.
Final Thoughts
From healthcare to semiconductors, productivity tools to geopolitical shifts, this week reinforced that AI is no longer an experimental add-on. It is core infrastructure. The pace of progress is breathtaking, and the stakes are only getting higher.
We’ll be back next week with more insights from the frontlines of intelligent systems. Until then, stay curious and keep building.
— Darren Redmond, https://darrenredmond.com’s take on This Week in AI
No responses yet