Lifeboat Foundation: Safeguarding Humanity
Toggle light / dark theme

Blog – Page 4

Get the latest international news and world events from around the world.

Log in for authorized contributors

Anthropic Compute Spend Outpaced Its Own 2025 Revenue

A leaked draft of Anthropic’s IPO prospectus shows the company spent $7.33 billion on compute and infrastructure in 2025 against $4.6 billion in revenue — roughly $1.60 of infrastructure for every dollar earned.

Operating losses exceeded $8 billion. Long-term cloud and infrastructure commitments total $518 billion over the next decade.

Revenue is accelerating: Q2 2026 alone hit $11.5 billion, more than the entire previous year, and the company reports operating profit in recent quarters.

The number that matters for buyers is the spend-to-revenue ratio. Every enterprise building long-term plans around a frontier AI vendor’s API pricing is implicitly betting that the vendor’s own infrastructure math holds.

Before locking multi-year contracts, ask vendors for their compute-spend-to-revenue ratio — not just today’s price per token.

Full analysis:

#AIEconomics #Anthropic #AIInfrastructure #Procurement

Continue reading “Anthropic Compute Spend Outpaced Its Own 2025 Revenue” | >

Human tau seeds trigger disease-specific shapes in mouse brains, new study finds

Tau is a protein that occurs naturally in the brain. It is essential for proper neuron development and for how neurons communicate with each other. The same protein, however, when misshapen or misfolded, can act as a driver of neurodegenerative diseases. Abnormal buildup of tau is a hallmark of Alzheimer’s disease, and scientists have also found an association between misfolded tau and more than 20 neurodegenerative diseases.

The prominent prion hypothesis proposes that misfolded proteins spread by forcing normally shaped proteins to copy their exact 3D structure. To test its validity, a recent study investigated at the atomic level how misfolded tau proteins spread and multiply through the brain.

Researchers injected tiny amounts of human tau seeds from Alzheimer’s disease (AD) or corticobasal degeneration (CBD) into the brains of healthy mice. High-resolution microscopy that captured tau at the atomic level revealed that the mice’s own tau had misfolded into exact 3D structural copies of the diseased original seeds. The human seeds vanished within a week, but not before setting off a chain reaction in which the mice’s tau continued to form fibers with matching disease-specific folds and accumulate for the next nine to 12 months.

New study investigates PFAS exposure and future inflammatory bowel disease

Exposure to per- and polyfluoroalkyl substances (PFAS), often called “forever chemicals,” may play a role in the development of inflammatory bowel disease (IBD) years before symptoms appear, according to a new five-year study supported by a $3.3 million National Institutes of Health (NIH) R01 grant to researchers at the Icahn School of Medicine at Mount Sinai.

Mount Sinai Researchers Identify How APOE4 Gene Damages Brain Blood Vessels in Alzheimer’s Disease

About the Mount Sinai Health System

Mount Sinai Health System is one of the nation’s leading integrated academic health systems and one of the largest in the New York metropolitan area. Its comprehensive system includes seven hospitals, more than 400 outpatient practices, over 600 research and clinical laboratories, the Icahn School of Medicine at Mount Sinai, the Graduate School of Biomedical Sciences, and the Mount Sinai Phillips School of Nursing. Together, the Health System comprises approximately 48,000 employees, more than 9,000 physicians, and 8,600 nurses.

As a leading learning health system, Mount Sinai combines clinical expertise with scientific discovery to improve patient care while training the next generation of health care and biomedical leaders. The Health System provides care across every stage of life, from prenatal care through geriatrics, while advancing personalized medicine through artificial intelligence, data science, and biomedical research.

Michelle Odemwingie on AI in Education | Goalkeepers 2026

A test score can show what a student understands, but it doesn’t always tell the full story.

As a former educator, Michelle Odemwingie learned this firsthand in her own classroom. Now, as CEO of Achievement Network, she’s using data and AI to give teachers clearer insights into student development and help them tailor learning to each student’s unique needs.

Hear her powerful story about helping students learn and feel authentically seen: https://youtu.be/Td6qbAPnhx4

#WorldTeachersDay

What if a test could tell a teacher more than what a student got wrong?

Continue reading “Michelle Odemwingie on AI in Education | Goalkeepers 2026” | >

Millisecond-timescale, genetically targeted optical control of neural activity

Karl Deisseroth just won the Nobel Prize.

He showed that brain cells can be turned on/off by shining blue light on them, after inserting a light-sensitive protein from algae.

The method was named optogenetics and has been used worldwide to link specific neural circuits to behavior.

But Ed Boyden is the main author of this paper? I was wondering.

It’s a question many people in neuroscience are asking today, and there are a few layers to it.

First, what the prize actually honors. The 2026 Nobel in Physiology or Medicine went to Karl Deisseroth, Peter Hegemann, and Georg Nagel “for their discoveries concerning light-gated ion channels and optogenetics” — essentially two scientific contributions: discovering the light-gated ion channel (Hegemann and Nagel’s channelrhodopsin work in the early 2000s) and turning it into a tool for controlling neurons (Deisseroth). Boyden’s name does appear in the Nobel committee’s own advanced documentation, which explicitly describes the 2005 paper as work by “Deisseroth, Nagel and collaborators” — the committee treats Deisseroth as the driving scientific force behind the neural-control half of the discovery.

Second, why first authorship doesn’t settle it. In the 2005 paper (Boyden, Zhang, Bamberg, Nagel, Deisseroth), Boyden was the first author, but Deisseroth was the senior/last author — meaning it was his lab, his conception of the project, and his scientific program. Nobel committees don’t reward authorship order; they try to identify who originated the decisive insight. Boyden was a graduate student at the time, doing the experiments within Deisseroth’s research program (Boyden himself has described working first in Richard Tsien’s lab, then moving into Deisseroth’s new Stanford lab in spring 2005).

Continue reading “Millisecond-timescale, genetically targeted optical control of neural activity” | >

FieldAI Robot Brain Hits A $10 Billion Valuation

FieldAI has raised $700 million at a $10 billion valuation — more than quadrupling its worth in just over a year.

The company builds no robots. It sells a “universal general-purpose brain” designed to run humanoids, drones, robot dogs and industrial rovers across different hardware platforms.

The thesis is clear: robot hardware will keep commoditizing. The durable value sits in the intelligence layer that can control many bodies.

FieldAI now has more than $135 million in revenue and customer contracts, with $35 million added since June. Backers include Nvidia, Intel, and Hyundai Motor Group.

It sits in a three-way race with Physical Intelligence (~$11B) and Skild AI (~$14B) to own that layer.

For buyers the question is shifting: do you buy a robot whose intelligence is locked to one manufacturer, or one running a third-party brain that can move across platforms later? The second option buys flexibility — and adds a new vendor dependency.

Full analysis:

Continue reading “FieldAI Robot Brain Hits A Billion Valuation” | >

AI Generated Youth Drugs Exist Now

The biggest thing in this is Sinclair mentioning A.I. turned 160 years worth of drug cocktail discovery into 2 months.

An AI designed a drug for a deadly lung disease, and six different aging clocks say it made people biologically younger. Rentosertib, from Insilico Medicine, was built to treat idiopathic pulmonary fibrosis. But when researchers ran patients’ blood through six proteomic aging clocks from six separate labs, all six read the treated patients as younger, by roughly 3 to 4 years at the peak.

We break down how AI found both the target and the molecule, what the new Nature Biotechnology paper actually shows (and what the hype got wrong), and why Insilico’s \.

Tail-Likelihood Reinforcement Learning: Teaching AI to Keep Its Best Possibilities Alive

Most reinforcement learning methods train an AI to maximize its average reward. That sounds sensible—but averages can hide something important. https://arxiv.org/abs/2609.

Think of two AI systems that both score an average of 70. One almost always produces results around 70. The other usually scores 60, but occasionally discovers an exceptional solution scoring 100. If we are allowed to sample the AI many times and choose the best result, the second system could be far more valuable.

This is the problem addressed by Tail-Likelihood Reinforcement Learning (TailRL).

Instead of asking only, “What reward does the AI achieve on average?”, TailRL asks a different question across the entire range of outcomes:

“How likely is the AI to produce something better than a particular reward threshold?”

The method effectively turns continuous rewards into a collection of success-or-failure questions. Importantly, it gives greater learning weight to rare, high-reward outcomes rather than allowing them to disappear into the average.

There is an interesting connection to Best-of-k sampling: when we generate multiple attempts and keep the best one, performance depends on whether the model has retained enough probability mass around those unusually good solutions. TailRL explicitly trains the model to preserve that potential.

The practical advantage is significant. The researchers report that TailRL helps models avoid getting trapped in mediocre solutions across tasks including object localization, maze navigation, GUI grounding, and code optimization. More importantly, the resulting models become better at exploiting additional sampling at inference time—the more attempts you give them, the more useful those attempts become.

Continue reading “Tail-Likelihood Reinforcement Learning: Teaching AI to Keep Its Best Possibilities Alive” | >

Coolest lava world yet with signs of an atmosphere offers clues to early Earth

In the search for extraterrestrial life, it makes sense to first look for rocky planets with an atmosphere, like Earth. Without an atmosphere, a planet can’t have surface water. But of the more than 6,300 exoplanets cataloged thus far, the vast majority are not rocky, and only a handful of the rocky worlds appear to have an atmosphere.

In a study published in The Astrophysical Journal Letters, a group led by University of Chicago scientist Brandon Park Coy reports another: a rocky super-Earth 154 light-years away in the constellation Pisces. Named HD 3,167 b, this very hot “lava world” zips around its host star in just one Earth day.

“What’s so surprising is that the closer a rocky planet orbits its star, the harder it should be to have an atmosphere, because it’s bombarded by stellar wind and gets more high-energy photons from the star. But it seems that many of these lava worlds do,” explained Edwin Kite, UChicago associate professor of geophysical sciences and co-author of the study. “These planets are too hot for life, but by studying them, we can say something about the processes that matter for other rocky worlds.”

/* */