Connect with us

Tech

Musk said that the self-driving Tesla has never crashed. Organizers counted 8

Published

on

Elon Musk has long used his powerful Twitter megaphone to amplify the idea that Tesla’s automated driving program isn’t just safe — it’s safer than anything a driver can achieve.

That drive began picking up pace last fall when the electric car maker expanded its “pilot” program for fully autonomous driving from a few thousand people to a fleet that is now growing. more than 100,000. The $12,000 feature allegedly allows Tesla to drive itself on highways and neighborhood streets, changing lanes, making turns, and complying with traffic lights and signals.

While critics berated Musk for testing the experimental technology on public roads without safety-trained drivers as a backup, Santa Monica chief investment officer and vocal Tesla supporter Ross Gerber was among the allies who defended him.

Advertisement

“There hasn’t been a single accident or injury since the launch of the FSD beta,” he said chirp in january. “Not one. Not even one.”

Musk replied with one word: “True.”

In fact, by that time, dozens of drivers had already submitted safety complaints to the National Highway Traffic Safety Administration about accidents involving fully autonomous driving—and at least eight of them were involved in crashes. Complaints in the public domain, in a Database on the NHTSA website.

One driver reported that the FSD “turned right onto a semi-truck” before accelerating to a mid-center, causing a wreck.

Another said, “The car went into the wrong lane” with FSD engaged “and another driver hit me in the lane next to my car.”

Advertisement

YouTube and Twitter are full of videos exposing FSD misconduct, including a recent Mail It appears to show a Tesla car steering itself on the track of an oncoming train. The driver pulls the steering wheel to avoid a head-on collision.

It is nearly impossible for anyone except Tesla to say the number of accidents, injuries, or deaths associated with FSD; The NHTSA is investigating several recent fatal incidents in which it may have been involved. The agency recently ordered automakers to report serious accidents involving automated and semi-automated technology to the agency, but it has not yet released crash-by-crash details to the public.

Automated car companies such as Cruise, Waymo, Argo and Zoox are equipped with over-the-air software that immediately reports faults to the company. Tesla pioneered such a program in passenger cars, but the company, which does not maintain a media relations office, did not respond to questions about whether it receives automated malfunction reports from FSD-powered cars. Automakers without over-the-air software must rely on public reporting and communications with drivers and service centers to judge whether an NHTSA report is necessary.

Attempts to reach Musk also failed.

Gerber said he wasn’t aware of the crash reports in the NHTSA database when he posted his tweet, but he believes the company would have known of any crashes. “Due to the fact that Tesla records everything that happens, Tesla is aware of every incident,” he said. He said it was possible the drivers were at fault in the accidents but he did not review the reports himself.

Advertisement

There are no accurate public statistics on motor vehicle accidents because the police officers who write crash reports only have the drivers’ data. “We’re not experts at how to pull that kind of data,” said Amber Davis, a California Highway Patrol spokesperson. “At the end of the day, we ask for the best memories of how you did it [a crash] Event. “

Mahmoud Hekmat, head of research and development at Umeo Autonomous Shuttle, said the data that Tesla’s vehicle’s automated driving system collects and sends back to headquarters is known only to Tesla. He said that Musk’s definition of an accident or accident may differ from that of an insurance company or an ordinary person. NHTSA requires crash reports for fully or partially automated vehicles only if someone is injured, an air bag is deployed, or the vehicle must be towed away.

Reports of the FSD disruption were first discovered by FSD critic Taylor Ogan, who runs Snow Bull Capital, a China-oriented hedge fund. The Times downloaded and evaluated the data separately to verify Ogan’s findings.

The data – covering the period from January 1, 2021 to January 16, 2022 – shows dozens of safety complaints about the FSD, including several reports of phantom braking, in which the car’s automatic emergency braking system applies the brakes without apparent appearance. the reason.

Here are excerpts from the eight reports of incidents involving the FSD:

Advertisement
  • Southampton, New York: A Model 3 traveling at 60 mph hit an SUV parked on the shoulder of the highway. The Tesla drove itself “straight across the side of the SUV, tore off the car’s mirror.” The driver called Tesla to say, “Our car has gone crazy.”
  • Houston: The Model 3 was traveling at 35 mph “when the car suddenly jumped over the curb, damaging the bumper, wheel and flat tire.” “The accident appeared to be caused by a distorted patch of road that gave the FSD the wrong perception of the obstacle it was trying to avoid.” The Tesla Service Center denied the warranty claim, charged him $2,332.37 and said he wouldn’t return the car until the bill was paid.
  • Priya: “While turning left, the car went into the wrong lane and another driver hit me in the lane next to my car.” The car “controlled itself and forced itself into the wrong lane…putting everyone involved in danger. The car was badly damaged by the driver.”
  • Collettsville, North Carolina: “The road was sloping to the left, and as the car took a very wide turn and veered off the road… He climbed up the right side of the car and over the start of the rocky slope. The front right tire had exploded and only the side airbags (both sides) were deployed. The car traveled 500 yards along The road then stopped itself.” Estimated damages were $28,000 to $30,000.
  • Troy, Mo: The Tesla was turning through a curve when “suddenly, about 40% of the way through the turn, the Model Y straightened the wheel and crossed the center line into the direct lane of the oncoming car. When I tried to pull the car back into my lane, I lost control and slid into a ditch and through the woods, causing Causing serious damage to the vehicle.
  • Jackson, Mo: Model 3 “Go right toward a semi-truck, then go left toward the pillars in the center where it was accelerating and the FSD wouldn’t turn off. …we owned this car for 11 days when our wreck happened.”
  • Hercules, California: “Phantom braking” caused the Tesla to come to a sudden stop, and “the car behind me did not react.” A rear-end collision caused “significant damage to the vehicle”.
  • Dallas: “I was driving with full self-driving assistance… There was a car in the blind spot so I tried to grab the car by pulling the wheel. The car sounded an alarm indicating I was about to hit the left-hand average. I think I was fighting with the car to get back Taking control of the car and I ended up hitting the left middle that bounces[ed] The car is all the way to the right, hitting the middle.”

Critics say full self-driving is a misnomer, and no vehicle available for sale to an individual in the United States can drive itself. “It’s just a fantasy,” said Meredith Broussard, a professor at New York University, author of Lack of Artificial Intelligence published by MIT NFSD Press. “It’s a safety nightmare.”

California regulations prevent the company from advertising a car as fully self-driving when it is not. The state Department of Automobiles is conducting a review of Tesla’s marketing, which is in its second year.

DMV President Steve Gordon has refused to speak publicly about the matter since May 2021. “A review is ongoing. We’ll let you know when we have something to share,” the department said Wednesday.



Source link

Advertisement
Continue Reading
Advertisement
Click to comment

Leave a Reply

Your email address will not be published.

Tech

This stunning crypto character home is for sale

Published

on

By

Her first foray into the rental market was to advertise her landlord’s properties on Airbnb. She decided to make the cannabis rental themed, decorating it with ganja leaves and providing visitors with free joints upon arrival.

But she said, “He was kind of a brawler.” Visitors from out of town were being robbed in the neighborhood, and there were parties with strippers that annoyed the locals. In the end, I decided to close it.

Levi got into crypto in 2017, after a teenage acquaintance advised her to buy bitcoin. “I’m like, ‘You’re 17.’ Like what the hell, you know? And he’s like, ‘Download Coinbase! Buy Bitcoin! “

“I wanted it to be a bit extravagant and tacky, but in a good way,” she said. “You know, like, crypto overload.” She added that the neighbors weren’t happy with the influx of hard-line cryptocurrency, which was another reason she eventually decided to sell the place. (She said the city council contacted her about the complaints.)

Advertisement

As for her next project? It’s a pancake delivery company called boffins, which launched smoothly. The selling point is that the pastries are delivered by huge men in branded tank tops.

She said, “You know, delivering flowers on Mother’s Day is great, but imagine a hot guy bringing you a cake. That would be fun!”

Source link

Advertisement
Continue Reading

Tech

Health tech products that I think are going to explode this year

Published

on

By

While I’m still skeptical of the accuracy of this kind of AI, I visited two other models who analyzed my skin with wildly variable results—one told me I had 25-year-old skin, the other 42-year-old. About my eye bags but it gave me higher ratings for purity and softness then fully recommended products Skin care routine It meets my skin type. I don’t know who to trust, if anyone is.

Similar technology can also monitor internal health. I tested a platform that scans your face to determine your risk level for various health issues (including mental health) from estimated blood pressure, heart rate, and other vital signs. I have nothing to compare it to, so it’s hard to say how accurate it is, but it’s another sign that we’ll see more options for assessing our health from home to share with doctors or receive AI-generated feedback.

Smart home tools for proactive healthcare

Speaking of assessing our health from home, I predict that smart health and fitness monitoring devices will skyrocket in 2023.

We’ve already seen an influx of wearables Fitness and sleep monitoringheart rate, blood pressure, and even blood oxygen levels — all of which can be used to alert us to potential problems before a visit to the doctor. I attended a session where he was the CEO Aura ring He talked about the usefulness of the smart ring, particularly for sleep and for “digital birth control,” or cycle tracking via wearable devices, which he believes will become more prominent in the future.

Advertisement

Taking wearables even further, a medical officer from Healthify explained how people can use a continuous glucose monitor, or continuous glucose monitor, along with a human coach or AI facilitated from an app to help them understand their metabolic panel. them and the best way to eat and work for them. Corpses.

Source link

Continue Reading

Tech

How computers learned to be predictors of the COVID-19 outbreak

Published

on

By

Imagine a time when your virus-blocking face covering was like a parachute. Most days, it stays in your locker or stowed somewhere in your car. But when the COVID-19 outbreak is in the forecast, you can use it.

Moreover, the intense viral forecast may prompt you to choose an outdoor table when meeting a friend for coffee. If contracting the coronavirus has the potential to make you seriously ill, you can choose to work from home or attend church services online until the threat has passed.

Such a future assumes that Americans will heed public health warnings about a pandemic virus — and that’s a big deal if. It also assumes a system that can reliably predict impending outbreaks with few false alarms, and with enough timing and geographic accuracy that the public can trust its predictions.

Advertisement

A group of would-be forecasters says it has the makings of such a system. they Suggestion To build a viral weather report published this week in the journal Science Advances.

Like the meteorological models driving weather forecasts, the COVID-19 outbreak prediction system emerges from a river of data fed by hundreds of local and global information streams. They include time-stamped online searches for symptoms such as chest tightness, loss of smell, or fatigue; geotagged tweets that include terms like “corona,” “pandemic,” or “panic buying”; location data aggregated from smartphones that reveal how many people are traveling; and a drop in online requests for directions, indicating fewer people are getting out.

The resulting volume of information is far greater than humans can manage, let alone interpret. But with the help of powerful computers and software trained to winnow, interpret and learn from the data, the map is beginning to emerge.

Advertisement

If you check this map against historical data—in this case, two years of epidemiological experience in 93 counties—and update accordingly, you might have the makings of an outbreak forecasting system.

This is exactly what the team from Northeastern University is leading Computer scientist it’s over. In their attempt to create an early warning system for the COVID-19 outbreak, the study authors built a “machine learning” system capable of chewing through millions of digital traces, integrating new local developments, improving its focus on subtle signs of disease, and issuing timely notifications of impending local surges of COVID. -19.

Of his many Internet searches, one proved to be an especially good warning sign of an impending outbreak: “How long will COVID last?”

Tested against real data, the researchers’ machine-learning method predicted an increase in local virus prevalence up to six weeks early. Alarm bells were going off almost at the point where every infected person was likely to spread the virus to at least one other person.

After testing the prediction of 367 actual outbreaks countywide, the program provided accurate early warnings for 337 – or 92% – of them. Of the 30 remaining outbreaks, 23 have been identified just as they would have become apparent to human health officials.

Advertisement

Once the Omicron variant began spreading widely in the United States, the early warning system was able to detect early evidence of 87% of outbreaks county.

A predictive system with these capabilities could be useful to local, state, and national public health officials who need to plan for COVID-19 outbreaks and warn vulnerable citizens that the coronavirus threatens an imminent local resurgence.

But “we’re looking beyond” COVID, he said Mauricio Santayanawho runs Northeastern’s Machine intelligence group to improve health and the environment.

“Our work aims to document technologies and approaches that may be useful not only for this, but for the next pandemic,” he said. “We’re gaining the trust of public health officials, so they won’t need any more convincing” when yet another disease begins to spread across the country.

This may not be an easy sell for the state’s public health agencies and the Centers for Disease Control and Prevention, which have all struggled to keep up with pandemic data and incorporate new ways to track the spread of the virus. The CDC’s inability to adapt and communicate effectively during the pandemic has led to some “dramatic and very public missteps,” said Dr. Rochelle Walensky, the agency’s administrator, I acknowledge. And she warned that only a “changing culture” would prepare the federal agency for the next pandemic.

Advertisement

CDC’s lackluster efforts to develop prediction tools haven’t paved the way for easy acceptance either. 2022 Assess Of the forecasting efforts used by the CDC it concluded that most “failed to reliably predict rapid changes” in COVID-19 cases and hospitalizations. The authors of this assessment cautioned that the systems developed to date “should not be relied upon to make decisions about the possibility or timing of rapid changes in trends.”

Anas Barryan expert in machine learning at New York University, called the new early warning system “very promising,” though it was “still experimental.”

“The machine learning methods presented in the paper are good, mature and well-researched,” said Barry, who was not involved in the research. But he warned that in a once-in-a-lifetime emergency such as a pandemic, it would be dangerous to rely too heavily on a new model to predict events.

For starters, Barry noted, the coronavirus’ first encounter with humanity didn’t yield the long historical record needed to fully test the model’s accuracy. And the three-year period of the pandemic gave researchers little time to recognize the “noise” that comes when too much data is thrown into a jar.

The centers for disease control and state health departments have begun to use epidemiological techniques such as phylodynamic gene sequencing And Wastewater monitoring To monitor the spread of the Corona virus. Using machine learning to predict the location of upcoming viral spikes, Santillana said, could take another leap of imagination for these agencies.

Advertisement

In fact, accepting early warning tools like the one developed by Santillana’s group may require some leaps of faith, too. As computer programs digest large amounts of data and begin to discern patterns they can reveal, they often generate surprising “features” — variables or search terms that help predict an important event, such as a viral mutation.

Even if these visible signs prove to accurately predict such an event, their relevance to a public health emergency may not be immediately apparent. A sudden signal may be the first sign of a new trend – for example, a previously unseen symptom caused by a new variant. But they may also seem so random to public health officials that they cast doubt on the software’s ability to predict an imminent outbreak.

My review, said Santillana, who also teaches at the Harvard School of Public Health early work of his group She responded suspiciously to some of the signals that appeared as warning signs of an upcoming outbreak. Santayana said one of them — the tweets referring to “panic buying” — seemed like a false signal from machines that ran into a random event and imparted meaning to it.

He defended the inclusion of the “panic buying” signal as a signal of an imminent outbreak domestically. (After all, the early days of the pandemic were marked by lack of basic elements Including rice and toilet paper.) But he acknowledged that the “black box” early warning system could face resistance from public health officials who need to trust its forecasts.

“I think the concerns of decision makers are a legitimate concern,” Santayana said. “When we find a signal, it has to be reliable.”

Advertisement

Source link

Continue Reading

Trending