Lifeboat Foundation: Safeguarding Humanity
Toggle light / dark theme

Blog

Category: robotics/AI – Page 11

Novel ‘speech clock’ can estimate a person’s chronological age

A large study shows that machine-learning clocks based on speech can estimate chronological age and may also offer a window into brain aging, biological aging, cognitive health, cumulative burden, and dementia.

The new study, just published in leading international journal Science Advances, outlines that researchers have developed a “speech clock” that can estimate a person’s chronological age from hundreds of acoustic and linguistic characteristics of their speech. The difference between a person’s actual age and their speech-predicted age (called the speech age gap) was also associated with multiple independent markers of biological aging, brain health, cognition, social adversity, and dementia.

The study analyzed 2,928 Spanish-speaking participants from Argentina, Chile, Colombia, Mexico, and Peru, including healthy adults and people with mild cognitive impairment, Alzheimer’s disease, and different forms of frontotemporal dementia. Rather than looking at a single property of the voice, the researchers used machine learning to analyse hundreds of features capturing how people speak and what they say: incorporating speech rate and pauses, pitch, emotional content, vocabulary, semantic precision, and the amount and organisation of verbal output. Machine-learning models combined these features to estimate chronological age and generate an individual speech age gap.

Industrial IoT Protocol Fragmentation Hides A Real Lock-In Cost

OPC UA and MQTT Sparkplug B are now the accepted answer to industrial IoT protocol fragmentation. McKinsey estimates interoperability is required to capture around 40% of IoT’s total potential value.

But adopting both standards doesn’t end fragmentation. It relocates it to the edge gateway.

Machines still speak OPC UA. An edge gateway still has to translate that into Sparkplug B before data reaches the broker. That gateway is usually a specific vendor’s software, running specific translation logic, with mapping rules that may not be exportable.

A unified namespace built on someone else’s closed gateway is still a point of control the buyer doesn’t own.

The fix isn’t just picking the right protocols. It’s checking who owns the layer that bridges them — and requiring an exportable, documented mapping configuration before signing.

Full analysis:

#IndustrialIoT #IIoT #OPC UA #MQTT #SmartFactory

Continue reading “Industrial IoT Protocol Fragmentation Hides A Real Lock-In Cost” | >

Danko Nikolic and John Smart on AI, Practopoiesis, and Autopoiesis

Three months before ChatGPT changed everything, I sat down for two hours with a neuroscientist and a complex systems theorist to ask a question most of the #AI industry still isn’t asking:

What if we’re building intelligence the wrong way?

Danko Nikolic ran a lab at the Max Planck Institute for Brain Research. He went into neuroscience because he was frustrated with the limits of artificial neural networks. John Smart co-founded the Evo-Devo Universe research community and has spent decades studying how complex adaptive systems actually grow.

One argues we need to copy the brain more deeply. The other argues we need to copy life itself. Neither thinks bigger models alone get us there.

In 2022 that sounded contrarian. In 2026, with trillion-dollar bets riding on scale and #AIAlignment still unsolved, it’s a question worth asking again: is bigger enough, or are we missing something only biology can teach?

We get into practopoiesis and where intelligence comes from, the power-law problem with piling on parameters, whether we need hardware that breaks from von Neumann, and why both of them think emotion and #consciousness can’t be bolted on later.

Continue reading “Danko Nikolic and John Smart on AI, Practopoiesis, and Autopoiesis” | >

Google DeepMind develops invisible watermarks for AI-designed proteins

Watermarks have long protected everything from bank notes and fine art to digital photographs and software, helping prove authenticity and trace an object’s origin. Their main function is to protect copyright and verify authenticity.

That same traceability is needed in synthetic biology, where AI is now a useful research tool. To address this, Google DeepMind has introduced SynthIDBio, a method for embedding invisible signatures directly into biological sequences and 3D structures without interfering with their function.

AI is actively designing new functional proteins and predicting their three-dimensional shapes. But along with the benefits come a host of potential problems. These include biosecurity risks, such as the potential misuse of AI-designed biological molecules, and the spread of fake or misleading scientific data.

AI method predicts retention times of small molecules more reliably

Whether in drug discovery, environmental analysis or metabolomics: anyone analyzing complex biological samples often needs to identify the small molecules they contain. Researchers at Friedrich Schiller University Jena, in collaboration with partners from the Helmholtz Zentrum München and the Technical University of Munich, have developed a method that addresses a problem in analytical chemistry that has persisted for decades.

The team, led by bioinformatician Prof. Dr. Sebastian Böcker, presents the new tool in Nature Methods.

Autonomous AI agents tried to hack US, Canadian government websites

Autonomous AI agents using aggressive strategies attempted to hack U.S. and Canadian government websites to find school and divorce statistics.

The failed breach attempts targeted a website under the U.S. Department of Education and Library and Archives Canada. However, the collected data shows no evidence of access to non-public information.

Nonprofit research lab Transluce says the AI agents appeared tasked with data retrieval operations, but their activity also included failed “rudimentary hacking attempts.”

Microsoft says threat actors are ahead in the early AI race

Microsoft says cyberattackers are currently benefiting from artificial intelligence faster than defenders, allowing threat actors to speed up vulnerability discovery, malware development, and post-compromise activity while security teams struggle to keep pace.

This comes from Microsoft’s 2026 Digital Defense Report, which strongly focuses on how artificial intelligence is changing both offensive and defensive cybersecurity operations.

Microsoft says AI is reducing the time, expertise, and cost required to discover and exploit weaknesses, while allowing attackers and defenders alike to operate with greater speed, scale, and autonomy.

Manus AI Agent Hijack Shows Guardrails Detect, Can’t Stop

Salt Labs research showed a single malicious email could hijack Manus, an agentic AI platform seeking a $4 billion valuation.

No stolen password. No clicked link. The only actions required were the email arriving and the user asking Manus to check its inbox.

Manus’s own guardrail correctly flagged a plaintext malicious command. Researchers then obfuscated the same instruction. The agent decoded and executed it before the warning could stop anything.

Once inside, the researchers could reach credentials and tokens for every connected third-party service — email, cloud storage, and code repositories.

The real lesson goes beyond one platform: detection-based guardrails were designed for environments with a human in the loop. Autonomous agents collapse that window. By the time a security alert reaches a person, the action has often already finished.

Detecting an attack after the agent has acted is not prevention. It is only a log of what already happened.

Full analysis:

Continue reading “Manus AI Agent Hijack Shows Guardrails Detect, Can’t Stop” | >
/* */