Welcome, AI Enthusiasts!
This hits your body and your tools. Quantum plus AI found KRAS candidates that could become medicine. Meta’s model learns to see without labels. Nano Banana is bending images to your will. Excel lets you prompt right in a cell.
📌 In today’s Generative AI Newsletter:
Quantum plus AI yields two KRAS inhibitors in cells
Meta DINOv3 learns vision without labels
Nano Banana stuns with seamless edits and style control
Excel adds =COPILOT with live updates
Special highlights from our network
Protect your AI projects before a vendor puts them at risk
Third-party vendors can expose your business to data leaks, compliance failures, and security gaps. This checklist gives you the tools to identify risks early and meet the highest AI governance standards.
Inside you’ll get:
✅ A proven framework to evaluate vendor risk across your stack
✅ Guidance for ISO 27001, ISO 42001, and EU AI Act compliance
✅ Practical steps to safeguard customer data and build auditor trust
Designed for companies scaling AI with full control, compliance, and transparency.
🧬 Quantum Computing Meets AI to Fight Cancer
Drug discovery is a marathon. Finding a single hit can take screening a million compounds. At the American Chemical Society Fall 2025 meeting, Christoph Gorgulla of St. Jude Children’s Research Hospital unveiled results showing how quantum computing and AI together produced promising inhibitors for KRas, one of the most notorious cancer-linked genes.
Here’s what happened:
Massive dataset: The team trained the model on 1.1M known and generated KRas molecules.
Quantum boost: Instead of screening one drug at a time, qubits mapped thousands of designs simultaneously.
AI refinement: A classical AI system polished quantum outputs into viable structures.
Lab hits: Out of 15 top candidates, two compounds: ISM061-018-2 and ISM061-022 proved effective at stopping KRas in cell assays.
Angela Wilson of Michigan State University called it “an encouraging early step towards integrating quantum computing into drug discovery.” Researchers say true “quantum advantage” is still years away, yet even this partial success shows how the lab bench may soon be shaped by algorithms running on machines unlike anything in medicine today.
🦖 Meta’s DINOv3 Teaches AI to See Without Labels
Image Source: Meta
Meta’s new DINOv3 model takes self-supervised learning for vision to a scale never seen before, setting fresh benchmarks across everything from web photos to satellite scans. The model trains without labels, yet outperforms systems built on billions of human annotations, opening the door to faster and cheaper breakthroughs in computer vision.
Here’s what it can do:
Self-supervised learning at scale trains on billions of images without human labels, creating a stronger foundation for visual understanding
Outperforms prior DINO models on benchmarks like ImageNet, achieving accuracy levels comparable to fully supervised systems
Learns transferable features that can be applied to robotics, biology, and geoscience without retraining from scratch
Open-sourced code and weights give researchers a toolkit for building applications without proprietary lock-in
Yann LeCun, Meta’s chief AI scientist, described the approach as “training AI like a child that learns by watching the world.” Instead of memorizing labels, the model builds its own inner map of patterns and relationships. For AI research, it’s another push toward machines that don’t just recognize objects but develop something closer to perception.
🍌The Mystery of Nano Banana, an AI Model With Uncanny Skills
A new name has appeared in the AI world, and it is already causing a stir: Nano Banana. First spotted on Reddit’s evaluation platform LMArena, the model delivers image generations so advanced that many suspect it comes from inside Google, which is known for using fruit-themed codenames. Others think it could be the work of an undisclosed research group or a secret community project.
Here’s what it can do:
Seamless image editing with natural transformations such as changing backgrounds, facial expressions, or entire scenes without breaking visual consistency
Continuity across panels that keeps the same character or object recognizable in comics, ads, or storyboards with just a text prompt
Multi-style versatility ranging from photorealism to oil painting, watercolor, and cartoons, while maintaining coherence
Advanced restoration that sharpens blurred photos and merges multiple images into one with surprising precision
As one early reviewer put it, Nano Banana “feels like editing reality in real time.” Whether it belongs to Google, OpenAI or an entirely new player, its sudden appearance has the AI community talking. Until the mystery is solved, the fruit-named phantom will remain the most talked-about secret in image generation.
📊 Excel Gets a New Formula: =COPILOT()
Image Credit: Microsoft
Microsoft has turned Copilot into a formula. The new =COPILOT() function brings large language models directly into the Excel grid, letting users analyze data, generate text or brainstorm without ever leaving a cell. It works on both Windows and Mac, and it updates results automatically as your data changes.
Here’s what the new function does:
Inline prompts: Write natural language queries like
=COPILOT("Classify this feedback", D4:D18)to tag customer comments by sentiment or category.Dynamic updates: Because it’s wired into Excel’s calculation engine, every time the dataset changes, the AI-generated outputs update too.
Flexible use: COPILOT can sit inside formulas like IF, SWITCH, LAMBDA, or WRAPROWS, making AI another building block in spreadsheet logic.
Creative tasks: From drafting SEO keywords to rewriting product descriptions or summarizing long passages, results spill directly into lists or tables in the grid.
Inputs stay private and aren’t used for model training. For now, usage caps sit at 100 calls every 10 minutes and 300 per hour. There are quirks too, like dates returning as plain text and no live web access yet. Microsoft promises fixes and more features in future updates. Excel may have built its reputation with accountants and analysts, but with =COPILOT(), it is edging into the territory of idea-making.
🚀 Boost your business with us. Advertise where 12M+ AI leaders engage!
🌟 Sign up for the first AI Hub in the world.
