Google says it fixed more Chrome bugs in June than over the past two years, thanks to AI

Good news! Is this how Google now covers up a buggy software? Just kidding!

With ML & AI software engineering will never be the same again! Rapid, unprecedented progress is to be expected!

I personally can confirm that about over the last two months or so my Google Chrome browser prompted me almost every day to perform an update.

"... The tech giant announced on Thursday [7/30/2026] that it has fixed a whopping 1,072 security bugs in the last two versions of Chrome, both released in June. That is more than the number of bugs patched in the previous 23 versions released over the last two years, which totaled 1,036 fixes. ..."

"We’re living through a massive shift in the software security industry. Large Language Models (LLMs) are unlocking unprecedented capabilities for automated vulnerability discovery, scaling far beyond the limits of human security expertise, and requiring new approaches for staying ahead of attackers. ..."

Google says it fixed more Chrome bugs in June than over the past two years, thanks to AI | TechCrunch

Stronger with every update: How we’re making Chrome and the web safer in the AI Era (original news release) "How Chrome is using AI to improve vulnerability discovery, triage, and patching."

Wow! Very impressive! Look at "internally found" (red line)

Chinese military researchers tap US AI models to train defense systems

Bad news! Why should AI not be dual use?

Maybe US AI models should train Chinese defense systems how to malfunction in case of a military conflict? 😊

What about the US spying on China how they use US AI models?

"Chinese military researchers have used outputs from leading U.S. artificial intelligence models developed by OpenAI and Anthropic to train domestic AI systems to advance China’s defense capabilities, according to a Reuters review of more than 80 Chinese academic papers and patents. ...

One paper published last year by researchers in PLA Unit 96941, a military intelligence and cyber-warfare unit in Beijing, described using OpenAI’s GPT-3.5 to process sensitive military source code. ..."

"Summary

• Chinese military researchers distil US models, documents show
• Distilling done to create smaller and more secure versions of a model
• Distillation has become a flashpoint in US-China ties as AI competition intensifies
• Distillation cannot replace massive compute needed to develop frontier AI from scratch

Chinese military researchers have used outputs from leading U.S. artificial intelligence models developed by OpenAI and Anthropic to ​train domestic AI systems to advance China's defence capabilities, according to a Reuters review of more than 80 Chinese academic papers and patents. ..."

"Executive Summary:

• Distillation has become essential to the rise of Chinese artificial intelligence (AI) models. Chinese academic and industry papers published in the period 2024–2026 detail how distillation takes place—including adversarial distillation—and provides insight into the kinds of entities involved in the work of using leading models to build capabilities that rival the world’s best models.
• The prize is frontier reasoning: the step-by-step “chain of thought” that is the most expensive capability to build and the one on which U.S. laboratories lead most clearly. Chinese research literature documents a deliberate, premeditated effort to distill that capability.
• Some published Chinese research also focuses on engineering distillation techniques to evade detection. Much of this work is conducted by researchers affiliated with the People’s Liberation Army, the defense industry, and the Chinese Academy of Sciences.
• Distilled models are either already being used, or are proposed to be used, for public security applications such as surveillance and monitoring systems, and for military and cyber operations.

..."

Chinese military researchers tap US AI models to train defense systems

Chinese military researchers tap US AI models to train defence systems (Reuter)

Chinese Research Details Distillation for Military Use

Spy vs. spy (remember the famous MAD cartoons, source) MAD = mutually assured destruction

Self-Supervised Learning of Structured Dynamics from Videos

This could be an interesting new, but narrowly focused paper by Andrew Zisserman and his team!

From the abstract:

"Understanding motion in video is a fundamental challenge for visual learning, as frame-to-frame change entangles two sources of dynamics:

camera motion and object motion.

This decomposition has remained underexplored in representation learning, partly because these factors are tightly coupled in natural videos and difficult to supervise separately.

Yet recovering it is important for learning robust motion representations that separate meaningful object dynamics from camera-induced variation. We study whether such structured motion representations can be recovered from frozen features of a pretrained image vision transformer.

We propose the Structured Dynamics Model (SDM), which explicitly separates the dominant source of temporal change from residual dynamics through future-feature prediction, rather than representing video change with a single entangled latent or with unstructured, spatially dense transition tokens.

Training combines self-supervised learning on real video with weak supervision of scene dynamics on synthetic Kubric data.

We evaluate SDM on ProbeMotion, a new evaluation suite spanning synthetic and real videos with camera motion, object motion, and combined dynamics.

SDM outperforms backbone baselines using global CLS or average-pooled features, and compares favorably to strongly supervised representations such as VGGT on several probes, despite using substantially weaker supervision. 

These results suggest that pretrained image models can be readily repurposed into structured video-dynamics representations, providing a useful inductive bias for learning and analyzing latent video dynamics."

[2607.21576] Self-Supervised Learning of Structured Dynamics from Videos

Robostral Navigate

This could be an interesting new paper in robotics by over 140 authors!

This work appears to rely heavily on generating synthetic data and it focuses on indoor navigation.

Here is the official news release (a blog) by Mistral AI: Introducing Robostral Navigate "Summary: Robostral Navigate is an 8B model that enables robots to autonomously navigate complex environments using only a single RGB camera, achieving 76.6% success on unseen R2R-CE benchmarks—outperforming multi-sensor approaches while being more efficient. Built entirely in-house with simulated data and token-efficient techniques, it generalizes across robot types and adapts to real-world obstacles unseen during training. The model combines pointing-based navigation with reinforcement learning for continuous improvement, paving the way for unified embodied AI in robotics."

From the abstract:

"Deploying navigation systems at scale requires a recipe that minimizes sensor assumptions, generalizes across robot embodiments, and trains efficiently.

Yet, today's best systems depend on depth sensors, multi-camera rigs, or pre-built maps, limiting the hardware they support and increasing deployment cost.

We introduce Robostral Navigate, an 8B vision-language model built around this scalability objective. The model consumes only a stream of monocular RGB images - the most ubiquitous sensor across robotic platforms and predicts waypoints by pointing to the next target location in the current camera view.

Operating purely in image space, rather than robot-specific coordinates, makes the policy naturally robust to changes in camera intrinsics and scene scale, enabling deployment across wheeled, legged, and aerial robots without recalibration.

We generate 2.4 million trajectories across 350k simulated scenes to reduce the reliance on real-world data collection and scale easily. We further introduce a prefix-caching training recipe that packs entire episodes into single training sequences, reducing training tokens by 22x and cutting training time from months to days.

A tree-based attention mask prevents conditioning on previous ground-truth actions, encouraging visually grounded action prediction, and reinforcement learning is used to further improve exploration and recovery capabilities.

On the Room-to-Room and Room-Across-Room in Continuous Environments (R2R-CE and RxR-CE) benchmarks, Robostral Navigate sets a new state of the art. On R2R-CE, it achieves a 77.4% success rate, surpassing the best monocular method by 10.5 points and the strongest depth- or multi-camera system by 5.3 points despite using only a single RGB camera. On RxR-CE, it reaches 75.1% success rate, outperforming all monocular baselines."

[2607.20785] Robostral Navigate

Attributes Should Come from Images, Not Class Names: Distribution-Conditioned Attribute Selection for Vision-Language Models

This could be an interesting research paper by Deepak Pathak and his team!

From the abstract:

"A popular route to interpretable zero-shot classification asks a large language model (LLM) to describe each class name and prompts CLIP with the resulting descriptors.

We show that these descriptors carry little visual evidence of their own: removing the class name from the prompt collapses ImageNet accuracy from 59.5% to 15.5%.

The diagnosis is that the descriptors are conditioned on the label rather than on the images, so they describe the concept in general and mislead exactly when the data shifts; an LLM insists that strawberries are red, but every strawberry in ImageNet-Sketch is a colorless line drawing.

We therefore select attributes from the target image collection instead: we score a large attribute pool against the images in CLIP's joint embedding space and keep the top-scoring attributes per class.

Selected this way, class-name-free attribute prompts reach 23.8% on ImageNet (against 15.5% for LLM descriptors), the gain holds on four shifted ImageNet variants, and reselecting from the LLM's own pool isolates the selection mechanism as the cause.

With one image per class, the selected attributes outperform the prompt-tuning method CoOp by 3 points while fitting in under a minute instead of 14 hours, with no learned soft prompt to obscure the decision. Because the attribute set is chosen by the data, it doubles as a readable summary of a dataset, which we use to describe distribution shift in words."

[2607.18695] Attributes Should Come from Images, Not Class Names: Distribution-Conditioned Attribute Selection for Vision-Language Models

Patch Policy: Efficient Embodied Control via Dense Visual Representations

This could be an interesting new paper by Yann LeCun and his team!

From the abstract:

"Pretrained dense visual features from Vision Transformers (ViTs) are powerful yet have been underutilized in robot learning. Modern robot policies either compress each observation into a single global token, or rely on visual backbones trained from scratch, sacrificing both fine-grained spatial detail and the benefits of large-scale visual pre-training.

While there exist policies that do operate on dense patch features like large vision-language-action models (VLAs), they tend to be heavy and slow, inheriting the full cost of a billion-parameter vision-language model (VLM) backbone.

We close this gap with Patch Policy, a minimal architectural extension that enables transformer-based policies to consume dense pre-trained patch tokens directly without the computational overhead of a full VLM.

At its core is a block-causal attention mask that preserves the temporal causality of standard policies while letting the model attend over many patch tokens per observation, alongside other state information. Patch Policy is lightweight, fast, and highly effective.

Across four simulated and three real-world environment suites, our method achieves a 40% relative improvement over policies using state-of-the-art global-pooled representations.

Furthermore, it surpasses fine-tuned OpenVLA-OFT by 18% while using roughly 0.7% of the parameters.

We believe Patch Policy provides a pipeline for the robotics community to readily leverage continuing progress in visual representation learning, without sacrificing the training efficiency or inference speed required for high-frequency, reactive control. Videos can be viewed at this https URL"

[2607.18236] Patch Policy: Efficient Embodied Control via Dense Visual Representations

Artificial intelligence (AI) forces a fundamental reform of primary and secondary pedagogy

Does anyone seriously doubt it?

The centuries lasting traditional setup of classrooms with teacher in front and fixed school hours is doomed!

Best of all it may allow personalized/individualized learning! Thus, each student can e.g. learn at their own pace and curiosity.

How about brain computer interfaces?

How will AI extend the human brain?

AI Companies Are Purchasing and Destroying Millions of Used Books for Training Data. Really!

Sounds more like a conspiracy theory! The company accused of it denies it!

This piece of sensationalism is junk journalism! It was repeated all over in different media outlets!

What about books stored in public and private libraries? They will still be around. Most of these books are most likely not rare!

Just look at the headline of the supposedly originating article (see second link from the bottom) suggests a almost laughable motive and it is behind paywall!

"Tom’s Hardware reports that an investigation has revealed that AI companies are reportedly engaging in large-scale purchases of secondhand books to feed their language models, with the physical books subsequently being destroyed after scanning. The practice involves using middlemen to secretly acquire the books, potentially to avoid the controversy that has surrounded AI companies’ use of copyrighted material. ..."

AI Companies Are Purchasing and Destroying Millions of Used Books for Training Data "Major AI companies are buying millions of used books through intermediaries and destroying them after digitization to obtain high-quality training data while avoiding public scrutiny, according to a recent investigative report."

AI companies are reportedly shredding millions of books after using them to train AI models — tech giants outsource to middlemen to secretly buy up books for training material

AI Companies Are Buying Tons of Old Books Because They're Free of AI Slop (behind paywall) "ISBNdb, a company that sources printed books for AI companies to turn into training data, tells clients “the optics problem is real.”"

ISBNdb "The facts: ISBNdb has never purchased, scanned, or destroyed a book - for AI training or anything else. We don't train AI models, and we never have. The page was up to explore demand for a service we never brought to life. We've taken it down."

Mark Zuckerberg predicts that billions of people will have personal AI agents in five years

Good news! Plus many publicly available AI agents on the Internet etc.! 

From public libraries to public agents in the 21st century!

"... “As we move toward a future where we’re all interacting with multiple agents, I think that WhatsApp and our other messaging surfaces are going to become increasingly important,” he said, noting that WhatsApp is already the leading platform where users interact with Meta AI. ..."

Mark Zuckerberg predicts that billions of people will have personal AI agents in five years | TechCrunch

Mark Zuckerberg

74% of workers ask AI questions instead of colleagues - with potentially serious consequences

My wife frequently asks me questions instead of me first after she consulted with AI! 😊 Caution: Satire!

74% of workers ask AI questions instead of colleagues - with potentially serious consequences | ZDNET "A series of studies revealed that employees are spending less time asking their coworkers for help. Here's how that tactic could backfire and how organizations need to adapt."

National Science Foundation funds nationwide network self-driving research laboratories at US universities

Good news! From self driving cars to self driving science labs!

The US National Science Foundation (NSF) has launched a new $380 million (£285 million) effort to establish a nationwide network of automated laboratories enabled by artificial intelligence (AI), and as part of this initiative North Carolina State University (NCSU) will lead a project that aims to build ‘self-driving’ chemistry labs.

Dubbed the self-driving platforms for experimental co-design in chemistry and materials science

NSF funds $380m AI lab network including self-driving chemistry platform at NC State University | Chemistry World

NSF announces $400M investment in new national network of AI-programmable cloud laboratories in alignment with the U.S. government's Genesis Mission (original news release) "Awardees will test, scale and demonstrate new methods and tools that advance automated science and engineering for discovery and translation"

Major Corporations Resume Hiring as AI Limitations Become Apparent

Oh, I can believe that instantly, given how often inaccurate or outdated or unreliable information is provided by AI.

AI output often seems very convincing, but requires e.g. serious double checking.

AI needs a lot more training not only more curated data!

Major Corporations Resume Hiring as AI Limitations Become Apparent "America’s largest companies are reversing course on hiring freezes, announcing plans to expand their workforces after months of restraint driven by AI adoption expectations."

The new art of herding robots or how to manage a swarm of robots (e.g. against drone incursions)

Food for thought!

"The urgency of this work comes from rapidly evolving real-world conditions. Drone incursions are increasing in both frequency and affordability ... “It still requires a centralized and expensive response from the Air Force just to take down a $200 drone.” 

Swarm robotics offers an alternative. By deploying many low-cost systems, the team aims to shift the economics of defense. “To take out a $200 drone, we want to do it in less than $2,000,” Nowzari explained. ...

The team’s goal is to enable military planners to assemble, adapt, and deploy swarm-control and coordination systems using modular,building-block components, reducing complexity while improving mission effectiveness. 

They are building a Smart Swarm system with three components:
a Swarm Decision Guidance Assistant,
Swarm Operations Assistant, and
Adversarial Risk Assistant.

Together, these tools help operators interpret data, assign tasks, and respond to changing conditions without being overwhelmed.  ...

To explore these questions, the project incorporates an unexpected element: a video game. ... the simulator translates swarm behavior into an interactive, accessible format. In the game, called Astroswarm, players design simple algorithms to control fleets of spaceships defending a target, mirroring real-world swarm coordination. ..."

The new art of herding robots | George Mason University "As autonomous systems grow in number and complexity, researchers at George Mason University are tackling a critical question: How can a single human operator manage an entire swarm of robots? Professors Alexander Brodsky, Cameron Nowzari, and Daigo Shishika are designing smart-swarm technologies to help military planners coordinate unmanned aerial and ground vehicles within Human-Machine Integrated Formations. "

A short biography of one of the best known engineers of machine learning & AI: Ilya Sutskever

Recommendable!

(831) The Greatest AI Engineer of All Time - YouTube

OpenAI Launches Health Platform to the public at large in the US

Good news! Self medication and self treatment most likely preceded the emergence of the medical profession in human history!

Physicians/medical doctors are sometimes referred to in German language as "Halbgötter in Weiß" (demigods in white clothing)! Iconoclasm by AI?

Democratize healthcare! Pronto! 😊

Unfortunately, it appears that the first release is too much focused on Apple Health only. Hopefully, this will quickly be corrected by OpenAI.

"OpenAI rolled out ChatGPT Health, a service that allows users to securely connect their medical records and wellness apps to the artificial intelligence chatbot. 

The new feature comes in response to nearly 300 million people who already turn to ChatGPT with health-related questions each week. But users are cautioned to consider the likelihood of inaccuracy as well as the issue of sharing sensitive, potentially deeply personal information with a chatbot. ..."

The Flyover - Daily Newsletter for You

Launching Health in ChatGPT (original news release) "Rolling out to U.S. users, you can securely connect your health information to help you better understand and navigate your health."

'Unprecedented' Self-Directed CyberAttack on Hugging Face by OpenAI's Rogue Model Was Repelled by Chinese Defensive AI

I had intended to publish this blog post earlier! I procrastinated! Mea culpa!

I have just blogged here about this complex and very sophisticated cyber attack.

'Unprecedented' Self-Directed CyberAttack by OpenAI's Rogue Model Was Repelled by Chinese Defensive AI "AI startup Hugging Face deployed a Chinese-developed AI model to counter an unprecedented cyberattack launched by a rogue OpenAI system, highlighting both the emerging threat of autonomous AI attacks and the growing capabilities of Chinese AI technology."

OpenAI used the ExploitGym to cyber attack Hugging Face

What an irony! 😊 There is even a research paper published in May as a preprint on arXiv about ExploitGym with the title containing "Turn Security Vulnerabilities into Real Attacks"! 

As they say reality is stranger than fiction! 😊

I blogged here and here about this incidence before.

"... OpenAI pitted its models against a benchmark called ExploitGym, released in May, which challenges LLMs to find ways to exploit real-world vulnerabilities found in commonly used software. ...

On July 9, according to reporting by Reuters, OpenAI’s models started trying to break through the proxy. They found an unknown bug in the proxy’s software and used it to access the internet. From there, they broke into Hugging Face’s computer systems on July 11, apparently looking for data sets and solutions that would help them complete their task. Hugging Face announced the hack on July 16. ..."

The Hugging Face AI hack is scary—but unsurprising

ExploitGym: Can AI Agents Turn Security Vulnerabilities into Real Attacks?

Language diversity: A historical context for language loss for the past 6,000 years or more

Over human history, many languages have disappeared and continue to disappear!

Why should the evolution of human languages be so different than the biological evolution? Do we need about 7,000 existing and actively spoken languages on our planet?

Hopefully, the Tower of Babel will go away in the 21st century. The world would be a much better place if we all speak the same language (with or without machine translation)!

Is this article another way of trying to indoctrinate with the DEI (Diversity, Equity, and Inclusion) ideology?

The approach taken in this study, i.e. a kind of extrapolation from present-day hunter-gatherers, is certainly questionable if not junk! This assumption is possibly a very strong inductive bias! What if e.g. the first humans spoke only one or a few languages and that for a long time before language diversity emerged maybe with the geographic spread of humans?

Notice that this study does not blame colonialism, but the preceding multinational empire building and expansion for the loss of languages.

"Humanity’s linguistic diversity is in trouble [???]. Roughly half of the 7500 languages signed or spoken today are endangered [???], and at least four go extinct every year. But while preserving the languages that do persist is a top priority, a new study in Science suggests that we are already more than one millennium beyond humanity’s linguistic “golden age.”

Linguists have long debated when the most languages were in use: Some have argued that the total number of languages remained roughly steady from the last ice age until modern-day European colonialism, others suggested that diversity has steadily declined since the dawn of agriculture about 12,000 years ago, and a third camp theorized a high point sometime before large states became established 4000 years ago.

Since writing was not developed until around 6000 years ago, however, understanding dynamics of early language remained challenging.

So, a team of anthropologists applied data on the languages, distributions, and behaviors of present-day hunter-gatherers to estimates of the total human population size and the size and number of individual cultures over the last 12,000 years. This allowed them to create a model of the total number of languages in use at any point since the last ice age ended.

The researchers showed that linguistic diversity rose steadily for around 10,000 years, reaching a “golden age” of tens of thousands of languages around 1000 to 3000 years ago before entering a language loss free-fall that continues to this day. The results reveal that the language crisis is not the result of modern colonialism but rather the culmination of millennia of cultural exchange and homogenization."

"... The golden age was followed by a period of rapid decline in linguistic diversity that coincided with the rise of large states and multinational empires, such as the Roman Empire, the researchers found. The finding challenges a commonly held view that widespread language extinction began later, about 500 years ago, with the onset of European colonial expansion. ..."

"To the Point

• Fewer languages before agriculture: The researchers estimate that between approximately 4,500 and 6,200 languages were spoken at the beginning of the Holocene, around 12,000 years ago - probably fewer than the roughly 7,500 languages spoken and signed today.
• A linguistic “golden age”: As the global human population grew, language diversity also increased. The models suggest that tens of thousands of languages may have existed between 1,000 and 3,000 years ago.
• A deep history of language loss: The decline in linguistic diversity appears to have begun with the expansion of large states and empires, long before modern European colonialism.
• Today’s languages are survivors [???]: The languages spoken today represent a small and historically biased sample of past linguistic diversity [???], with important implications for how researchers explain global patterns in language and culture.

...

They began with ethnographic data from 171 hunter-gatherer and fisher societies whose traditional subsistence and mobility had not been profoundly transformed by contact with food-producing populations. These data allowed the researchers to estimate the likely distribution of ethnolinguistic group sizes near the beginning of the Holocene.

The team then combined these estimates with reconstructions suggesting that the global human population 12,000 years ago was between approximately 4.4 and seven million. Assuming that ethnolinguistic groups at that time generally corresponded to distinct languages, the models produced an early-Holocene estimate centred on roughly 4,500 to 6,200 languages, although broader plausible estimates ranged from around 3,300 to 7,800. ..."

From the editor's summary and abstract:

"Editor’s summary

More than four languages are lost every year, and this rate is predicted to increase. Blasi et al. investigated how language diversity has changed over the Holocene using a Bayesian modeling approach, ethnographic data, and estimates of the total human population size over time.

They assumed a maximum size of ethnolinguistic groups and that the number of different ethnolinguistic groups is related to the number of languages. Their findings suggested that language loss is not a new phenomenon, and that language diversity has been decreasing rapidly over the past one to three millennia following a peak of an order of magnitude higher diversity than we see today. ...

Abstract

Characterizing the factors that have shaped linguistic diversity is fundamental for understanding human history, culture, and cognition.

In this study, we combined statistical and social computational modeling, ethnographic data, and paleodemographic inference to model trajectories of global linguistic diversity.

Before the onset of plant and animal domestication, the number of languages was smaller than it is today (4500 to 6000 compared with 7500).

Subsequent increases in global population precipitated increased linguistic diversity.

We uncovered a linguistic “golden age” with tens of thousands of languages 3000 to 1000 years ago. Great loss of linguistic diversity did not begin with recent colonial expansion but as multinational empires first spread along with their languages, pathogens, and cultures. Thus, extinction has likely played a much greater role in shaping linguistic and cultural diversity than previously thought."

ScienceAdviser

In Science Journals

Study uncovers lost ‘golden age’ of languages (original news release) "A new study coauthored by Yale linguist Claire Bowern suggests that tens of thousands of languages were spoken between 1,000 and 3,000 years ago."

The lost golden age of language diversity (original news release)

The rise and fall of language diversity through the Holocene (no public access)

Map of currently spoken and signed languages of the world, colored by endangerment status.

Wearables: Monitoring movement in infants

Amazing stuff! However, it appears this research is not quite new, see e.g. A smart jumpsuit tracks infants’ motor development (published in 2022 by the same institution as the senior author of this research article)

From the editor's summary and abstract:

"Editor’s summary

Evaluation of infant spontaneous motor activity currently requires clinic visits and trained observers.

Airaksinen et al. used a wearable motion sensing jumpsuit to monitor infant movement and postures at home during free play.

Deep-learning classifiers objectively characterized 220 motor metrics in 92 typically developing infants aged 4 to 19 months, generating motor growth charts of emerging postures and their dynamics.

Approximately 91% of these motor metrics generalized to an independent clinical cohort of infants including both typical and atypical neurodevelopment, and 55% of the metrics statistically distinguished the two groups.

These findings suggest that wearable sensors could provide objective, home based quantitative measures of motor development dynamics to support routine care or to better understand developmental patterns. ...

Abstract

Early gross motor performance is a key component of neurodevelopmental assessments, and longitudinal follow-ups in personalized health care need better objective methods for measuring it. We studied how the detailed characteristics of infants’ gross motor development could be modeled with age-normative growth charts. We performed serial at-home measurements (total of 580 measurements, 1227 hours of free playtime) from a cohort of 92 typically developing infants at 4 to 19 months of age, using wearable suits with four movement sensors. A previously developed and validated, fully automated analysis pipeline detected postures and movements at a second-by-second resolution, to be used for deriving 220 detailed motor metrics of postures, movements, transition dynamics, locomotion types, and activity counts. The results showed that early gross motor development is sufficiently stable for constructing robust growth charts, most prominently for modeling evolving postures and their dynamics, and a wide range of motor metrics have sufficient temporal stability to provide reliable individual-level tracking. External validation in a clinical cohort with 39 infants, including participants with both typical and abnormal neurodevelopment, showed that the growth charts were highly generalizable (91% of the studied motor metrics were statistically comparable to those of the normative cohort), and 55% of the motor metrics showed statistical differentiation of abnormal neurodevelopment. These findings indicate that measurements of at-home wearables can provide assessments to meet a wide range of needs in health care and developmental science."

In Science Journals | Science

At-home wearable measurements provide detailed growth charts of infant early gross motor development (no public access)

How Open Models Are Driving AI Research: Nvidia as An example

Impressive! This is only the beginning of an exciting revolution in science and medicine!

"Every year, the International Conference on Machine Learning (ICML) reveals where thousands of AI researchers have decided to put their work. 

This year’s accepted papers reveal a clear direction: open frontier models and open AI infrastructure have become foundational to how modern AI science gets done.

NVIDIA had 74 papers accepted at ICML 2026. Approximately 2,000 accepted papers cite NVIDIA GPUs, and 145 cite NVIDIA Nemotron — a family of open models, including open datasets — as the foundation for new research.

Hundreds more draw on NVIDIA Cosmos, NVIDIA Isaac GR00T, BioNeMo and other NVIDIA open model families, spanning physical AI, robotics, autonomous vehicles and biomedical research.

The Themes Defining This Year’s Research

Areas including vision and video generation, reinforcement learning for large language models (LLMs) and agent training as well as AI inference remained prominent themes across this year’s papers, reflecting sustained investment these fields command — while several new areas also broke through.

Robot world models drew significant attention, with papers like DreamDojo pushing the boundary of how AI systems learn to reason about and act in physical environments. DreamDojo, for example, learns how the physical world behaves from human video and builds on NVIDIA Cosmos open frontier models to predict how a robot would handle objects and operate in environments it was never trained on. It lets researchers evaluate policies, plan actions and teleoperate a virtual robot, accelerating development without the costs and risks of physical deployment.

AI for life sciences was fueled by NVIDIA BioNeMo open models and research contributions that help researchers understand protein function, molecular behavior and genetic code. Papers like FLIP2 introduce public benchmarks for testing how well AI predicts the effects of protein mutations. KERMT is a new BioNeMo open model for predicting molecular properties important to drug discovery. 

Synthetic data generation (SDG) drew particular interest at ICML this year with several Nemotron and physical AI open datasets, reflecting a broader shift in how researchers are thinking about training at scale without relying solely on human-labeled data. ..."

How Open Models Are Driving AI Research | NVIDIA Blog "At the International Conference on Machine Learning, nearly 145 papers cite NVIDIA Nemotron open models and datasets as the foundation for new research."

International Conference on Machine Learning (ICML) 2026 (South Korea 7/6-7/11)