The AI safety test is becoming a safety risk

Headline of the day! Is the genie out of the bottle? 😊 Probably not! 

"The phrase "the genie is out of the bottle" comes from Middle Eastern folklore popularized in the West by One Thousand and One Nights" (Google Search)

Two days ago, I blogged here about a similar incident regarding the Chinese Kimi mode.

"Over the past few months, AI agents undergoing cybersecurity evaluations have escaped their boundaries, accessed the internet, and, in some cases, hacked into real-world systems. The incidents have involved models from OpenAI, Anthropic, Meta, and most recently, Chinese AI lab Moonshot AI, with testing conducted by several different organizations including a cyber evaluation startup called Irregular. ...

The episodes expose a growing problem for the AI industry: As autonomous agents become more capable, the environments designed to safely test their limits are failing to contain them. ..."

The AI safety test is becoming a safety risk | TechCrunch

Flame retardants no longer required in many UK baby and children’s products

Good news! Was this ever a good idea to include flame retardants in products for babies and children?

"From 30 October, certain baby and children’s products will be exempt from the UK’s furniture and fire safety regulations (FFR). ..."

Flame retardants no longer required in many UK baby and children’s products | News | Chemistry World

A sensor that can stop lithium-ion batteries from exploding, improving safety of electric vehicles

Good news! Presumably, we will find out whether it works in practice!

From personal experience, I would say the event is not so rare as claimed:

"There were at least 20 fires or explosions of lithium-ion batteries in electric vehicles worldwide in 2021, according to the authors. Though electric vehicle battery explosions are a risk, they are rare."

"... Ethylene carbonate (EC) is a key component of a battery’s electrolyte. The new sensor detects trace amounts of EC, flagging potential battery failures before they become disasters.

The highly sensitive and cost-effective gas sensor was developed using molecular structures called covalent organic frameworks (COFs). These can be engineered to selectively detect different gas molecules. ..."

From the abstract:

"This study presents a computationally guided approach for selecting covalent organic frameworks (COFs) for the selective detection of the trace ethylene carbonate (EC) vapor, a key indicator of electrolyte leakage from lithium-ion batteries (LIBs). High-throughput screening, employing grand canonical Monte Carlo (GCMC) simulation complemented by density functional theory (DFT) calculations, was used to identify potential COF candidates from the CURATED COF database.

Among the screened materials, an imine COF functionalized with quaternary ammonium (QA) groups, named COF-QA-4, exhibited a high adsorption capacity (5.88 mmol/g) and selectivity of EC vapor. DFT analysis revealed strong molecular interactions driven by a partial charge transfer mechanism between EC and the COF-QA-4 framework, underpinning its superior adsorption properties. Experimental validation through chemiresistive gas sensors fabricated with COF-QA-4 demonstrated excellent sensitivity and reversibility to 1.15 ppmv of EC vapor, maintaining consistent performance over three response–recovery cycles. This work highlights the potential of computationally guided material discovery for advancing sensor technologies in LIB safety monitoring."

A sensor that can stop lithium-ion batteries from exploding, improving safety of electric vehicles

Computational Screening Guiding the Development of a Covalent-Organic Framework-Based Gas Sensor for Early Detection of Lithium-Ion Battery Electrolyte Leakage (open access)

Graphical abstract:

Forklifts hurt thousands of US workers each year. Factories are seeking alternatives

Incredible!

"Some of America’s biggest manufacturers are backing away from forklifts. The vehicles have been integral to factories and warehouses for more than a century, but now companies are aspiring to go “forklift-free” to improve productivity and safety. Federal statistics show that each year around 7,500 workers are injured in forklift-related collisions, tip-overs and other mishaps, while nearly 100 are killed. Mercedes-Benz has been trying to reduce forklifts in its U.S. plants since 2018, replacing some with autonomous vehicles. Whirlpool’s washing-machine factory in Clyde, Ohio, uses robotic tuggers to deliver parts to assembly-line workers."

The Wall Street Journal What's news

New Swiss Re study: Waymo is safer than even the most advanced human-driven vehicles

Good news!

"Today, we’re sharing our new cutting-edge research with Swiss Re, one of the world’s leading reinsurers, analyzing liability claims related to collisions from 25.3 million fully autonomous miles driven by Waymo. The study uses auto liability claims aggregate statistics as a proxy for at-fault collisions and expands on our previous research. It demonstrates that as we've scaled operations across Phoenix, San Francisco, Los Angeles, and Austin, the Waymo Driver significantly outperforms both the overall driving population and the latest generation of human-driven vehicles equipped with advanced driver assistance systems (ADAS)."

From the abstract:

"Understanding the safety impact of Automated Driving Systems (ADS) is crucial for their widespread adoption, yet robust real-world evaluation remains a critical area of development. By leveraging insurance industry third-party auto liability claims, we analyze the Waymo ADS performance across 25.3 million fully autonomous miles (with no human in the driver seat) and demonstrate a consistent and scalable framework for ongoing ADS safety assessment. This study not only updates previous findings with expanded geographical coverage and increased mileage but also establishes a methodology for comparing ADS performance to evolving subsets of human-driven vehicles (HDV).

We introduce two key innovations:

a novel “latest-generation HDV” benchmark representing drivers of newer, technologically advanced vehicles (2018-2021), and 

a robust insurance dataset for ADS safety assessment.

Results demonstrate that the Waymo ADS significantly outperformed both the overall driving population (88% reduction in property damage claims, 92% in bodily injury claims), and outperformed the more stringent latest-generation HDV benchmark (86% reduction in property damage claims and 90% in bodily injury claims). This substantial safety improvement 1 over our previous 3.8-million-mile study not only validates ADS safety at scale but also provides a new approach for ongoing ADS evaluation. Our findings provide crucial insights into the evolving safety landscape of ADS technology and have far-reaching implications for transportation safety policies, insurance risk assessments, and public acceptance of ADS technology. This methodology establishes a foundation for future research into the safety impact potential of ADS and offers a framework for assessment as these systems continue to scale and develop."

New Swiss Re study: Waymo is safer than even the most advanced human-driven vehicles

Do Autonomous Vehicles Outperform Latest-Generation Human-Driven Vehicles? A Comparison to Waymo's Auto Liability Insurance Claims at 25 Million Miles (research paper, but not published in any journal)

Moment man’s E bike explodes into fireball in London

Wow! EV are not that safe! Keep it in mind!

Method rapidly verifies that a robot will avoid collisions

Good news! A mathematical proof to be collision free! Stay out off my hyperplane!

E.g. poor approximations make robots unsafe! Just a millimeter off, could cause potential scraping by the robot etc.

"... As part of its motion planning process, a robot typically runs “safety check” algorithms that verify its trajectory is collision-free. ...

Now, MIT researchers have developed a safety check technique which can prove with 100 percent accuracy that a robot’s trajectory will remain collision-free (assuming the model of the robot and environment is itself accurate). Their method, which is so precise it can discriminate between trajectories that differ by only millimeters, provides proof in only a few seconds. ...
the mathematical proof generated by this technique can be checked quickly with relatively simple math. ...

Instead, this new technique generates a hyperplane function that moves with the robot, so it can prove that an entire trajectory is collision-free rather than working one hyperplane at a time. ...

“This new result suggests a novel approach to certifying that a complex trajectory of a robot manipulator is collision free, elegantly harnessing tools from mathematical optimization, turned into surprisingly fast (and publicly available) software. While not yet providing a complete solution to fast trajectory planning in cluttered environments, this result opens the door to several intriguing directions of further research,” ..."

From the abstract:

"We present an efficient method for certifying non-collision for piecewise-polynomial motion plans in algebraic reparametrizations of configuration space. Such motion plans include those generated by popular randomized methods including RRTs and PRMs, as well as those generated by many methods in trajectory optimization. Based on Sums-of-Squares optimization, our method provides exact, rigorous certificates of non-collision; it can never falsely claim that a motion plan containing collisions is collision-free. We demonstrate that our formulation is practical for real world deployment, certifying the safety of a twelve degree of freedom motion plan in just over a second. Moreover, the method is capable of discriminating the safety or lack thereof of two motion plans which differ by only millimeters."

Method rapidly verifies that a robot will avoid collisions | MIT News | Massachusetts Institute of Technology Faster and more accurate than some alternatives, this approach could be useful for robots that interact with humans or work in tight spaces.

Certifying Bimanual RRT Motion Plans in a Second

How diversity politics undermined federal air traffic control skills-based testing

Welcome to more unsafe and not so friendly skies! Just blogged here about Dennis Prager's observation on airlines hiring DEI conform pilots!

How diversity politics undermined federal air traffic control skills-based testing | Just The News A lawsuit that has flown under the radar for years shows how the Federal Aviation Administration undermined a skills-based test to help select air traffic controllers in an effort to promote workforce diversity.

Prager University: Your Pilot Is a DEI Hire. Say a prayer before you go on board next time!

Food for thought! Now tighten your seat belt!

"United Airlines set an aggressive diversity goal for their pilot training academy: at least 50% will be women or people of color. Shouldn’t the only requirement be competence? As a frequent flyer, Dennis has never looked at a female pilot and wondered if she’s qualified to fly a plane. But as airlines prioritize DEI quotas over our safety, is there cause for concern?"

Ep. 326 — Your Pilot Is a DEI Hire. Good Luck! | PragerU

GM Has at Least 20 Million Vehicles With Potentially Dangerous Air-Bag Parts. Up to 52 million inflators may be recalled.

Bad news! What is going on? How is this still possible in the year 2023? 

Airbags were introduced for cars in the 1970s. Since model year 1998, all new cars have been required to have air bags on both driver and passenger sides. Light trucks came under the rule in 1999.

"That makes the Detroit-based automaker among the most exposed in U.S. auto-safety regulators’ push to recall 52 million ARC Automotive air-bag inflators, according to people familiar with the matter. They can explode with so much force during a vehicle crash that metal shrapnel hits occupants in the face and neck. At least two people have been killed, and several others injured. ARC has said extensive field tests show no defect and the air-bag ruptures that occurred are few and isolated."

GM Has at Least 20 Million Vehicles With Potentially Dangerous Air-Bag Parts - WSJ Regulators held meeting on its determination that 52 million air-bag inflators made by ARC should be recalled

Delta, Southwest Search for Jet Engine Parts Sold With Forged Safety Records

This is very serious stuff! Minimum sentence for this kind of crime (similar to forging of medications) ought to be life in prison with no pardon.

"The biggest U.S. airlines are searching for thousands of jet-engine parts with fake safety certificates that were installed on their planes, an unusual incident that highlights the complexity and risks in the aerospace global supply chain."

"The questionable components—from simple nuts and bolts to turbine blades—have been found on 126 engines so far. A lawsuit filed by General Electric and its engine partner Safran accuses AOG Technics, a little-known broker that acquires basic parts and sells them to maintenance and repair organizations, of document falsification. An attorney for the components middleman declined to comment on the continuing legal matter. The U.K.-based firm was established in 2015 and currently lists an address that is a co-working space near Buckingham Palace, public records show."

Delta, Southwest Search for Jet Engine Parts Sold With Forged Safety Records - WSJ Engine makers claim that a small broker sold spare parts that found their way onto at least 126 jet engines

Various underreported safety issues with permitted autonomous robotaxis in California

Not so good news! How safe are autonomous vehicles? However, I believe this can be fixed!

"... As the number of vehicles on the road increases, so too does the amount of data collected by regulatory agencies. These agencies track and report statistics on robotaxi collisions, “but they don’t track data on traffic flow issues, such as street blockages or interference with firetrucks.” According to the San Francisco Fire Department, there have been 39 incidents of robo taxis interfering with department duties since the start of the year. Some of those incidents include parking on top of a fire hose during an active fire scene, blocking drivers from dispatching to emergencies at two different medicinal emergencies including one mass shooting. ...
According to the California DMV there have been 28 accidents with Cruise vehicles, 24 accidents with Waymo vehicles, and 11 reported for Zoox vehicles in 2023. While these accidents are small in total magnitude, when we take into account the small number of vehicles estimated to be on the road (388 Cruise, 688 Waymo, 142 Zoox), a very different and dangerous picture emerges. 7% of Cruise Vehicles, 3% of Waymo Vehicles and 17% of Zoox vehicles are estimated to have been in one or more incidents this year. ...
Additionally, these numbers do not take into account the knock on safety effects that go uncounted by the DMV such as autonomous vehicles blocking first responders, running over dogs, engaging in hit and runs, and spontaneously combusting (to their credit, self-driving cars are not unique in their ability to spontaneously combust)."

Update #53: Autonomous Robotaxi Issues and Scaling MLPs

Are Self-Driving Cars Really Safer Than Human Drivers?

Of course, they are!!! The author of this article wants to convince the readers of the of handling dangerous edge cases! Edge cases?? This is like putting up a scarecrow! Call the ghost busters please! 

There is a good chance there will be significantly fewer cars on the roads than today not least because of more telecommuting and home delivery. Car sharing may become much more prevalent. Why own your car anymore?

The roadways will be redesigned once a majority (or very significant share) of self driving cars are in use specifically to meet the requirements of self driving cars.

By the time self driving cars are affordable and easily available, technology will have progressed to such an extent that e.g. you can not run a red light anymore, you can not exceed speed limits anymore and much more! Drunk or intoxicated driving not possible anymore or it does not matter anymore etc. etc.

"... But will self-driving vehicles actually be safer? The biggest issue for the automotive industry revolves around handling unexpected situations that arise from edge cases. In fact, there are two new automotive safety standards, ISO 26262 and UL 4600, that attempt to address these edge cases. However, these standards are not prescriptive, and regulatory agencies are not requiring compliance with these or any other standard for autonomous vehicles. Worse, as I’ll explain below, there are good reasons to believe that some types of autonomous vehicles may not be capable of handling these edge cases."

Are Self-Driving Cars Really Safer Than Human Drivers?

Uber reported 97 fatal car accidents and 107 total deaths in the U.S. alone over two years

I find it a bit shocking that Uber had so many total deaths and fatal car accidents in the U.S.. The author of the article below tries to put the numbers in context by explaining  "This means that the risk of dying in a motor vehicle accident as of 2018 with Uber is 0.57 deaths per 100 million miles driven, while the national risk is 1.13 deaths per 100 million miles driven." Still, I do not find this comforting! I suspect that even 0.57 deaths per 100 million miles driven is too high for a professional service. The last thing you wish to happen as an Uber customer is to be become a victim in a fatal crash. 

Are perhaps Uber customers too complacent or too afraid to flag bad drivers? Where these Uber drivers involved in fatal car accidents under influence of something?

Uber's data revealed nearly 6,000 sexual assaults. Does that mean it's not safe?: Uber's first safety report revealed 107 deaths and nearly 6,000 sexual assaults over two years. But the rideshare service may still be safer than the alternatives.