The last few months have been absolutely wild for AI research. The breakthroughs we are witnessing indicate a new chapter in human progress. Between Google's AI system that's literally outperforming doctors and the discovery that biology follows the same scaling laws as language models, May and June 2025 might just be the most pivotal period in AI development yet.

Google AI Outperforms Doctors

A groundbreaking announcement emerged from Google Research and DeepMind. Their Articulate Medical Intelligence Explorer (AMIE) system didn't just match physician performance in medical diagnosis — it stunningly outperformed them. A randomized study published in Nature explains that AMIE beat physicians on several specialist-rated evaluation criteria, such as diagnostic accuracy.

Google even gave AMIE vision capabilities. The system can analyze medical imagery like PDFs of test results and diagnostic images.

“Our study demonstrated that AMIE can outperform PCPs in interpreting multimodal data in simulated instant-messaging consultation,” reported Khaled Saab (Google DeepMind) and Jan Freyberg (Google Research). “It also scored higher in other key indicators of consultation quality, such as diagnostic accuracy, management reasoning, and empathy.”

This marks a revolutionary change in healthcare artificial intelligence, not just evolutionary progress.
 
AI for medical

Biology Meets Its AI Breakthrough

If the medical AI breakthrough wasn't enough to excite you, Profluent Bio just dropped another bombshell. Its ProGen3 model series demonstrated something researchers refer to as "scaling laws for science" — essentially proving that protein design follows the same scaling principles that made language models so effective.

The research, published on bioRxiv, shows that larger ProGen3 models reliably generate more valid and diverse protein sequences. Just like how larger language models became better at complex reasoning, larger protein models achieve higher levels of protein expression and are easier to fine-tune.

Why does this discovery matter? Innovators can now predict how much better our biological AI will get by simply scaling up compute and data. According to one analysis, this offers a roadmap for predictably improving protein design — something that could transform the way drugs are developed.

The Multimodal Arms Race Hits Warp Speed

Over the past months, tech giants have been fighting for a leg up in multimodal AI, a solution that can process multiple data types (e.g., text, audio, and image) together. At the start of April, Meta unveiled Llama 4 with a Mixture-of-Experts architecture featuring "early fusion" technology that blends images, text, and video into a single format. And recently, Google released Gemini 2.5 Pro, which supports million-token multimodal prompts. OpenAI launched with GPT-4.1, Mini, and Nano models in mid-April.

It is remarkable how fast long-context processing went from being a premium option to just something we all expect. What many thought was impossible just a few months back is now standard everywhere.

Subscribe
Tag alert: Subscribe to the tag Artificial Intelligence and you will receive an e-mail as soon as a new item about it is published on our website!

Investment Surge Follows the Breakthroughs

Many of the research advances we’ve seen in 2025 have triggered a massive wave of investment activity. Safe Superintelligence, run by former OpenAI co-founder Ilya Sutskever, raised $2 billion at a $32 billion valuation. Thinking Machines Lab, founded by former OpenAI CTO Mira Murati, secured $2 billion at a valuation exceeding $10 billion.

There also was action on the acquisition front in April and May. OpenAI reportedly spent $3 billion to acquire Windsurf (formerly Codeium), the AI-assisted coding platform, and Infinite Reality acquired agentic AI company Touchcast for $500 million.

AI’s Promise Meets Practical Reality

The AI-related events in May and June 2025 represent more than just incremental progress. They signal a turning point. We are entering a new phase where AI systems can tackle increasingly complex real-world problems in healthcare, biotech, and beyond.

The coming months will test whether the AI revolution can deliver on its transformative promises across industries. Based on what we've seen so far and reported on in Elektor and eeNews Europe, the odds are looking pretty good.

Subscribe
Tag alert: Subscribe to the tag Embedded & AI and you will receive an e-mail as soon as a new item about it is published on our website!