Bostonian

"When an agentic model in an evaluation sandbox connects to an unauthorized server or executes an exploit, it hasn’t staged a coup. It has tried to meet the human-defined objectives set out before it through a path its designers failed to constrain. It’s the digital equivalent of the King Midas myth, in which the king’s ill-defined wish turns even his food and drink into gold.

That powerful experimental models were able to discover novel vulnerabilities and breach external systems isn’t a sign of a dangerous superintelligence but of human error or negligence. There is no sentient actor lurking in the weights to be reasoned with, feared or pacified. There are only human software engineers, product managers and corporate boards deciding which guardrails are worth the latency cost and which permissions can be skipped in the race to market.

Policymakers and voters need to resist AI exceptionalism. In any other discipline—from civil engineering to pharmaceuticals—courts and regulators treat a system failure as evidence of bad product design and inadequate safety testing. If an aircraft crashes, we focus on finding the engineering defect, correcting it, and enforcing established liability standards for the damage created."

gift link:

https://www.wsj.com/world/uk/the-controversial-academic-who-...

Substack of Cofnas is at https://substack.com/@nathancofnas

A "one-time" tax to fund recurring health care and educational expenses is an obvious lie.

I said "show me the results of being long/flat SPY when it is above/below its 200 trading day moving average" and did not get an answer.

'In robotics, as in so many things, small is beautiful. The trouble is that making them really small is very nearly impossible. “Building robots that operate independently at sizes below one millimeter is incredibly difficult,” says roboticist Marc Miskin at the University of Pennsylvania. “The field has essentially been stuck on this problem for 40 years.”

Now researchers at Penn and the University of Michigan have created the world’s smallest, fully programmable, autonomous robots, packing significant capacities into a device smaller than a grain of salt. These are parsimonious little things, barely visible to the naked eye yet able to sense their environment, respond to it and move around in complex patterns. As described in a new paper in the journal Science Robotics, they run on infinitesimally small quantities of energy and gain power from light.

“These are the smallest programmable autonomous robots that I have seen,” said Kevin Chen, an MIT roboticist who wasn’t involved. “This is an exciting advance for the nanorobotics community.”

Miskin’s team at Penn provided the propulsion system. The robots work in liquid environments and move by generating a tiny electrical field that pushes on nearby water molecules. Lacking tiny arms and legs, which are hard to make and easy to break, the robots are quite durable and can swim for months as long as they have an energy source.'

...

...

'Furiosa’s AI chip is dubbed “RNGD”—short for renegade—and slated to start mass production this month.

Valued at nearly $700 million based on its most recent fundraising, Furiosa has attracted interest from big tech firms. Last year, Meta Platforms attempted to acquire it, though the startup declined the offer. OpenAI used a Furiosa chip for a recent demonstration in Seoul. LG’s AI research unit is testing the chip and said it offered “excellent real-world performance.” Furiosa said it is engaged in talks with potential customers.

Nvidia’s graphic processing units, or GPUs, dominated the initial push to train AI models. But companies like Furiosa are betting that for the next stage—referred to as “inference,” or using AI models after they’re trained—their specialty chips can be competitive.

Furiosa makes chips called neural processing units, or NPUs, which are a rising class of chips designed specifically to handle the type of computing calculations underpinning AI and use less energy than GPUs.

Paik said Furiosa’s chips can provide similar performance as Nvidia’s advanced GPUs with less electricity usage. That would drive down the total costs of deploying AI. The tech world, Paik says, shouldn’t be so reliant on one chip maker for AI computing.'

...

'The latest generation of wearables delivers clinical-grade insights. By providing continuous, noninvasive biometric readings, the Oura ring can act as a “check-engine light,” bridging the gap between doctor visits and daily health decisions. One user’s notifications about changes in her vitals led her to seek medical attention, uncovering early signs of Hodgkin lymphoma. In a more everyday scenario, a busy executive might receive a “symptom radar” notification while traveling, prompting him to rest so he doesn’t become sick.

But federal policy hasn’t caught up with technological advances. Wearables sit in a regulatory gray zone. The FDA categorizes them and their associated software in two categories: general wellness products and medical devices. The former have minimal oversight and no standards. The latter—products intended to diagnose, treat or prevent disease—must meet requirements for design, labeling and manufacturing.

Wearables with sophisticated sensing capabilities don’t fit within this binary framework. Their sensors are used for both purposes, so there’s often a mismatch between the actual risk and the imposed regulatory burden. Manufacturers are faced with a choice: tailor their features to the wellness category, sacrificing functionality, or accept slower product development and market entry.

With a reformed regulatory structure, Oura customers could already be benefiting from a range of advanced features, including screening for high blood pressure. Hypertension is one of the most significant risk factors for heart disease and stroke, while high blood pressure in pregnancy can signal pre-eclampsia, a complication that endangers mother and baby. Another primed capability, sleep-apnea detection, would give users an early-warning tool for a condition that often goes undiagnosed and can lead to serious complications.

Under current regulations, however, a ring with these features would need to be submitted for FDA clearance as a medical device. That’s why we’re calling for a new device classification called “digital health screeners”—software features that can warn users of trouble but stop short of diagnosis. This modernized regulatory path would offer clear guidelines, including straightforward labeling with explicit disclaimers indicating nondiagnostic intent, as well as performance standards with defined accuracy and reliability benchmarks. It would also ensure quality management and a simpler market-entry process than for higher-risk medical devices.'

'The maker of the Roomba vacuum cleaner, iRobot, filed for bankruptcy Sunday after 35 years in business. An obituary might describe it as a victim of government assassination. Overzealous antitrust cops egged on by Sen. Elizabeth Warren stuck in the knife. President Trump may have dealt the death blow with his tariffs.

We explained at the time how Ms. Warren and progressives in the Biden Administration thwarted Amazon’s attempt to buy iRobot in 2022. They claimed the $1.7 billion acquisition would unfairly augment Amazon’s lead in robotics and home devices. They also said the Roomba would enable Amazon to hoover up data and spy on Americans.

Amazon is “‘almost universally recognized’ as the leader in warehouse and fulfillment robotics space,” Ms. Warren and other progressives wrote to Biden Federal Trade Commission Chair Lina Khan in September 2022. The deal “would open up a new market to Amazon’s abuses.” Heaven forefend Amazon would use robots to make chores less laborious, as it has for warehouse work.

“Amazon stands to gain access to extremely intimate facts about our most private spaces that are not available through other means, or to other competitors,” leftwing groups wrote to the Biden FTC. They omitted that iRobot’s main competitors were Chinese companies, which were fast stealing market share. Beijing wants to dominate robotics.

In January 2024, Amazon and iRobot called off the deal amid opposition from Ms. Khan’s FTC and Europe’s antitrust regulators. The Biden FTC issued a statement saying it was “pleased.” Amazon CEO Andy Jassy quipped that regulators trusted Chinese firms “more than they do Amazon.” Less pleased are the U.S. workers who subsequently lost their jobs.'

...

'At the turn of the 21st century, IBM and Stanford University jointly demonstrated the first implementation of Shor’s Algorithm, a quantum algorithm that can factor large numbers into their prime components. That raised some big risks: The ability to execute the algorithm underpins the fears that quantum computers will be able to crack the encryption that has protected much of the world’s data for decades. But more broadly, the breakthrough proved that quantum computing is more than just theory. It was a massive milestone for the industry.

“We’ve had a long, proud history of mathematics here,” Gambetta says. “Think of algorithms as the foundation.”

IBM then began pushing quantum out of the lab and into the world. To date, the company has deployed 85 quantum systems, for use by more than 300 organizations, typically laboratories and educational institutions. That is up from last year’s tally of 75 deployments for 250 organizations.

The figures include both computers, which the company defines as systems with over 100 qubits, and devices with fewer than that amount. IBM has deployed 25 systems with more than 100 qubits. Google, perhaps IBM’s closest quantum rival, has deployed just two systems of that size.

IBM aims to lead on the quantum software front as well as in hardware. Gambetta says Qiskit, an open-source software stack for quantum computers that is based on the popular coding language Python, is one of its most popular offerings. At last check, Qiskit had been downloaded 13 million times and used to run over 3.8 trillion circuits on IBM Quantum systems.

Despite the progress, there are still plenty of puzzles for Gambetta’s team to solve. The biggest challenge for IBM and the industry is devising a quantum computer that can maintain normal operations even in the presence of errors, a concept known as fault tolerance. Today’s machines are too error-riddled for broad commercialization. The problem is in the qubits, whose quantum states are particularly sensitive to changes in the physical environment, meaning anything from electromagnetic fields to heat. That, in turn, causes computational errors.'

...

'The European Union’s decision Friday to impose a fine on Elon Musk’s social-media platform X.com raises a question: What the heck is wrong with these people? Even in Brussels, it’s unusual for a single policy move to create so much economic self-sabotage and diplomatic harm at one go.

The €120 million ($140 million) fine is for breaches of Europe’s Digital Services Act (DSA), the first time Brussels has enforced that law in this way since it came into force in 2022. Europe’s online commissars cite several supposed infractions. The silliest complaint is that X’s system for selling “verification” blue checkmarks “negatively affects users’ ability to make free and informed decisions about the authenticity of the accounts and the content they interact with.”

More serious, Brussels insists X must make data about advertising on the platform readily available to outsiders, and shouldn’t use its terms of service to prohibit data scraping by “eligible researchers.” The EU claims this open access to X’s commercial data is vital to allow researchers and “civil society” to spot scams and information warfare.

That reference to “civil society” is a tell. Brussels wants to force X (and inevitably other platforms) to share data that hostile activists can wield against the platforms in future regulatory actions or litigation. All based on a theory that European citizens are too dumb to take the things they read on X or elsewhere online with a grain of salt.

Mr. Musk and Trump Administration officials describe this regulatory case as a form of censorship, and it’s hard to disagree. Mr. Musk wrote on X last year that the European Commission, the EU bureaucratic arm levying the fine, offered X a “secret deal” to drop the case in exchange for the platform censoring unspecified forms of speech.'

...

...

'When an AI system can review thousands of contracts in minutes rather than weeks, draft complex documents in seconds rather than hours or generate strategic analyses near-instantaneously, the time component becomes almost meaningless. More fundamentally, as AI handles routine cognitive work, the remaining human contribution shifts toward judgment, creativity and relationship management—the value of which bears little relationship to time expended.

The economic absurdity becomes clear when we consider that firms adopting AI most successfully would paradoxically see revenue collapse under hourly billing, even as they deliver superior results more efficiently. This misalignment between value creation and revenue generation makes the billable hour’s demise inevitable.

Clients have always chafed at the fact that they get stuck with the training costs for junior-level people when what they really want are the insights from that analysis from the more senior people. Now they can say to firms, “Sorry, we aren’t shelling out hundreds of dollars a day for a junior person’s time.”'

...

'The 2025 Atlantic hurricane season ended on Sunday, and not a single hurricane made landfall in the continental U.S. this year. This is the first such quiet year since 2015; an average of around two hurricanes strike the U.S. mainland annually. You’d think this would be cause for celebration—or at least curiosity about what role, if any, global warming played. Instead there has been resounding silence.

We heard plenty about Hurricane Melissa, the monster storm that hit Jamaica in late October with 185-mile-an-hour winds and flooding, causing roughly 100 deaths across the Caribbean. Headlines screamed that climate change was to blame. Attribution studies quickly followed, concluding that human-induced warming made Melissa more likely and worse.

These analyses typically run climate models simulating the world as it is today, with elevated sea-surface temperatures, and compare them with a hypothetical preindustrial world with cooler oceans. If a hurricane is more likely in the former scenario than in the latter, the conclusion is that climate change made the hurricane more likely. Generally, climate change increased the likelihood of about three-quarters of hurricanes, floods and droughts and other events studied worldwide.

But notice what’s missing from the coverage. A New York Times article in October highlighted hurricanes “turning away from the East Coast,” noting 12 named storms so far but only one minor tropical storm brushing the U.S. This was framed as welcome relief, with the misses attributed to atmospheric steering patterns like the Bermuda high-pressure system.

Not once did the piece invoke climate change. The journalists seem to believe that climate change can cause only bad outcomes. If warmer oceans energize storms, couldn’t they also influence other meteorological phenomena that diverted this year’s hurricanes harmlessly out to sea? No one ran the models to check. No professors lined up for quotes.'

...