AI Surgeon 🧠 AGI in 2026 🤖 New Industry Reports 📈
Plus, AlphaFold 3 goes open-source, the progress of AI is slowing down and new medical breakthroughs
👋 Welcome to this week in AI.
🎵 Don’t feel like reading? Listen to two synthetic podcast hosts talk about it instead.
📰 Latest news
AlphaFold 3 Goes Open-Source: A New Era for Academic Research and Scientific Discovery
Google DeepMind has open-sourced AlphaFold 3, allowing academic researchers full access to its code and training weights, a first for the technology.
Capable of predicting interactions between proteins and molecules, AlphaFold has already mapped over 200 million proteins.
This open-source release follows the Nobel Prize in Chemistry awarded to its creators, Demis Hassabis and John Jumper, recognising AlphaFold’s impact on solving complex protein structures.
Why It Matters
AlphaFold 3’s open-source release is set to transform academic research by giving scientists unrestricted access to advanced protein prediction.
This democratisation of technology could drive breakthroughs in areas like drug development and antibiotic resistance, allowing researchers worldwide to explore new applications that benefit public health and biology.
AGI in 2026-2027: Insights from Anthropic’s Dario Amodei on Responsible AI Scaling
In his Lex Fridman Podcast interview, Anthropic CEO Dario Amodei predicts AGI could emerge as early as 2026-2027.
He explains that scaling laws continue to drive AI capabilities toward generalised reasoning, though challenges remain in predicting the exact nature and timeline of AGI development.
Amodei anticipates a gradual, stage-wise manifestation of AGI abilities rather than a single breakthrough, with improvements in language models like Claude bringing incremental steps toward general intelligence.
Anthropic’s Responsible Scaling Plan reflects this cautious approach, aiming to manage AGI’s risks by developing models in controlled phases to monitor and mitigate potential issues.
Why It Matters
Amodei’s prediction for AGI by 2026-2027 signals an accelerated timeline for transformative AI, heightening the importance of responsible scaling and safety protocols.
Anthropic’s phased approach serves as a model for balancing rapid AI advancements with safety, encouraging the industry to align innovation with careful oversight to address AGI’s societal and ethical implications.
Adapting to the Plateau: OpenAI’s Next Move as Model Progress Slows
OpenAI’s anticipated model, code-named Orion, shows signs of “slowing down” in improvement compared to the leap from GPT-3 to GPT-4, especially in areas like coding.
According to The Information, while Orion does exceed current models, “there was less improvement” than prior upgrades, signalling a plateau in frontier model advancements.
To counter this, OpenAI has formed a “foundations team” to explore alternative approaches due to “a dwindling supply of new training data” and the high costs associated with training complex models.
Strategies include training on synthetic data produced by AI models to bridge data gaps, alongside post-training refinements to enhance model quality without full-scale retraining.
Why It Matters
With the traditional methods becoming increasingly costly and limited by data scarcity, OpenAI’s move to use synthetic data and post-training optimisations could set a new standard in AI development.
This shift indicates how AI firms are adapting to sustain progress when conventional resources reach their limits, potentially reshaping how the industry tackles future advancements in model capability and sustainability.
📰 Article by The Information (paywall)
AI-Trained Robot Operates by Watching Videos
Johns Hopkins researchers have advanced robotic surgery by training the da Vinci Surgical System to perform complex tasks, such as suturing and needle manipulation, solely by watching surgical videos.
Using an AI model that translates video into robotic movements, the system achieved human-level skill and adaptability, even retrieving dropped needles without explicit programming.
Why It Matters
This approach could fast-track the development of autonomous surgical robots, bypassing years of hand-coding to train robots in days.
With the potential to improve surgical precision and accessibility, this innovation may lead to safer, more efficient procedures and broaden access to high-quality surgical care.
📝 Blog post on the Johns Hopkins University hub
AI Detects Blood Pressure with 86% Accuracy in 30 Seconds
Japanese researchers have developed an AI that screens for high blood pressure and diabetes through a short video of a person’s face and hands, reaching “94% accuracy” for high blood pressure and “75% accuracy for diabetes.”
Using high-speed video at 150 frames per second, the system analyses blood flow patterns across 30 facial and hand regions, detecting blood pressure with up to 86% accuracy in a 30-second clip and 81% in just 5 seconds.
Why It Matters
This AI-driven, contactless method bypasses traditional tools, making health screenings simpler and more accessible.
As researcher Ryoko Uchida notes:
“if it were to require only a non-invasive photo or video, that could be a game-changer”
This will potentially allow people to monitor their health from home with just a smartphone or mirror.
📝 Blog post by the American Heart Association
📝 Reports
Beyond the AI Hype: Why 26% of Companies Are Seeing Real AI Results
CG’s Where’s the Value in AI? report shows that only 26% of companies are realising measurable AI results, with a select 4% achieving "substantial value" as AI leaders.
These leaders generate “50% higher revenue growth” and “60% greater shareholder returns” by focusing AI investments on core business functions like operations, which drive 62% of their AI value.
Rather than spreading resources thin, leaders double their digital investment and prioritise fewer, high-impact initiatives, achieving twice the RoI of their peers.
Why It Matters
For companies, the takeaway is clear: “focus on core areas, invest strategically, and build AI skills in-house.”
By adopting these methods, AI leaders establish a strong competitive advantage, moving beyond proof-of-concept to deliver consistent financial and operational benefits.
This disciplined approach to AI can help other businesses unlock sustainable value and innovation across their operations.
Semrush Study: How AI Overviews Are Shaping the Future of SEO
In a study by SEO leader Semrush, researchers analysed 200,000 AI-generated search overviews to understand their impact on search visibility and ranking strategies.
AI-generated search overviews (AIOs) are summaries generated by artificial intelligence that appear at the top of search engine result pages, offering users quick, concise answers to common informational queries.
The Semrush study found that AIOs are especially prevalent for informational, low-volume keywords—with 82% of desktop and 76% of mobile overviews for terms searched fewer than 1,000 times monthly.
AIO length varies widely, from 5 to 488 words, with an average of 11 links per overview. However, achieving high organic SEO rankings doesn’t guarantee presence in these summaries, as only 20-26% of top 10 organic results appear in AIOs.
Why It Matters
This Semrush study reveals how AIOs are shifting SEO dynamics by changing which content is surfaced to users.
Traditional SEO efforts focused on securing top organic rankings may no longer be sufficient for visibility in AIOs, leading marketers to explore new ways to capture attention in AI-driven results.
The study also highlights an opportunity for niche content to gain visibility in low-volume, informational queries, providing a fresh path to connect with audiences through AI-enhanced search results.
📝 Read the report from Semrush
From Testing to Production: 11x Growth in AI Model Deployment
The State of Data + AI report from Databricks reveals that companies are now 3x more efficient in deploying AI, moving models from testing to real-world use at a faster rate, with 11x more AI models going into production this year.
Natural language processing (NLP) has become the top AI application, with usage up 75% year-over-year, as businesses increasingly leverage it for insights from text-heavy data.
Another standout is the rapid adoption of vector databases, growing 377% YoY, which help companies tailor LLMs to their specific needs using their own data.
Why It Matters
These improvements in AI deployment mean companies can innovate faster and more effectively, reducing costs and driving value.
NLP’s growth highlights how AI is helping businesses unlock useful insights from everyday data, while the rise in vector databases reflects a shift toward custom AI solutions that are more aligned with each company’s unique needs, giving them an edge in today’s competitive market.
