robotics – Techdirt (original) (raw)
Robot Vacuums That Collect Photos, Videos And Audio Of Users To Train AI Models Start Yelling Obscenities And Chasing Dogs
from the not-so-clever dept
As far back as 2017, Techdirt was warning that robot vacuum cleaners represented a threat to privacy. In that instance, it concerned the possibility that iRobot, makers of the robot vacuum Roomba, might sell the data that its device collected about the size and layout of a home. Five years later, it was becoming clear that a new danger was emerging because robot vacuums were starting to incorporate cameras. These are the kind of images a system could gather, as reported by MIT Technology Review in 2022:
The photos vary in type and in sensitivity. The most intimate image we saw was the series of video stills featuring the young woman on the toilet, her face blocked in the lead image but unobscured in the grainy scroll of shots below. In another image, a boy who appears to be eight or nine years old, and whose face is clearly visible, is sprawled on his stomach across a hallway floor. A triangular flop of hair spills across his forehead as he stares, with apparent amusement, at the object recording him from just below eye level.
At that point, the images were taken by a development version, not a consumer product. They were then sent to Scale AI, a startup that used contract workers to label data for companies like iRobot. But the MIT Technology Review article presciently noted:
Ultimately, though, this set of images represents something bigger than any one individual company’s actions. They speak to the widespread, and growing, practice of sharing potentially sensitive data to train algorithms
Today, this kind of data is needed not to tweak the odd algorithm or two, as was the case a few years back, but to train powerful, large-scale AI systems. The current AI frenzy has led to a desperate race to gather and use as much training data as possible. A story on the Australian ABC News site shows that this includes the latest generation of robot vacuums:
Ecovacs robot vacuums, which have been found to suffer from critical cybersecurity flaws, are collecting photos, videos and voice recordings – taken inside customers’ houses – to train the company’s AI models.
The Chinese home robotics company, which sells a range of popular Deebot models in Australia, said its users are “willingly participating” in a product improvement program.
When users opt into this program through the Ecovacs smartphone app, they are not told what data will be collected, only that it will “help us strengthen the improvement of product functions and attached quality”.
According to the ABC News article, users are instructed to click “above” in order to find out the details of what data is collected and how it is used, but that — unhelpfully — there is no link available on the page. Ecovacs’s general privacy policy allows for the “blanket collection of user data for research purposes”, including:
- The 2D or 3D map of the user’s house generated by the device
- Voice recordings from the device’s microphone
- Photos or videos recorded by the device’s camera
Ecovacs confirmed to ABC News that data from users who had opted into its product improvement program was being used for training its AI model:
“During this data collection, we anonymise user information at the machine level, ensuring that only the anonymised data is uploaded to our servers,” the spokesperson said in a statement.
“We have implemented strict access management protocols for viewing and utilising this anonymised user data.”
Well, that’s what they all say. But supposedly anonymized data can often be de-anonymized, and “strict access management protocols” can be abused or circumvented. In fact, another article on the ABC News site shows how it was possible to gain remote access to an Ecovacs robot vacuum and watch live images from its camera. It was also possible to access all the unit’s logs, Wi-Fi credentials and sensors, the report claimed. More recently, it seems that hackers have been taking advantage of these security weaknesses. According to ABC News again:
Robot vacuums in multiple US cities were hacked in the space of a few days, with the attacker physically controlling them and yelling obscenities through their onboard speakers.
As well as spewing “racial slurs”, another Ecovacs Deebot under remote control chased its owner’s dog around a Los Angeles home. In a statement to ABC News, Ecovacs said:
ECOVACS has always prioritised product and data security, as well as the protection of consumer privacy. We assure customers that our existing products offer a high level of security in daily life, and that consumers can confidently use ECOVACS products.
ABC News noted that researchers at the Australian Centre for Robotics have developed a “privacy-preserving” camera that scrambles input before it is used, while retaining enough information for the robot vacuum to operate. But the problem isn’t lack of technical solutions here — those often exist. It’s the reluctance of companies to take extra steps to protect personal data that is present among the rest of the training set. Given today’s pressure to bring out new products with built-in AI as quickly and as cheaply as possible, it seems unlikely that many companies will be willing to do that. As a result, we can probably expect more horror stories of personal or even highly intimate data being leaked by your “AI-enhanced” robot vacuum and other so-called “smart” devices.
Follow me @glynmoody on Bluesky and on Mastodon.
Filed Under: algorithms, anonymization, china, photos, privacy, robot vacuum, robotics, surveillance, training set, videos
Companies: ecovacs, irobot, scale ai
$2,350 Amazon ‘Astro’ Business Robots To Become Pointless Paperweights 11 Months After Release
from the I'm-sorry-I-can-no-longer-do-that,-dave dept
Thu, Jul 11th 2024 01:14pm - Karl Bode
One of the common themes here at Techdirt over the last 20 years is how in the digital and internet-connected era, the very meaning of “ownership” and “property” has changed — often for the worse. It simply takes a merger or an ill-timed firmware update to render something you thought you owned — completely obsolete.
Case in point: last fall Amazon announced that it would be selling its Astro robot for use as a security guard for businesses. The robots, which can patrol up to a 5,000 square foot area, start at a price tag of 2,350,witha2,350, with a 2,350,witha180 per month subscription charge. But eleven months after the announcement, Amazon has announced that they’re shutting the effort down and bricking the robots.
The robots will simply stop working on September 15, and unfortunately can’t be repurposed in any way (Amazon is providing free recycling for your expensive and now completely pointless 2,350robot).Businesssubscriberswillatleastgetarefundfortheirunitsand2,350 robot). Business subscribers will at least get a refund for their units and 2,350robot).Businesssubscriberswillatleastgetarefundfortheirunitsand300 in Amazon credit. All personal data will be wiped from the device by Amazon.
In a statement to The Verge, Amazon indicates that they’re shifting their attention to Astro robots for the home. Employees that worked on the business version will be migrated to that version:
“We are fully committed to our vision of bringing world-class consumer robotics solutions to the home. To accelerate our progress and ongoing research to make Astro the best in-home robot, we’ve made the decision to wind down support for Astro for Business. We’re excited about the in-home experiences we’re inventing for Astro, and look forward to sharing more in the future.”
The home version of Astro is now only available as a $1,600, invite-only preview. Hopefully those users don’t have the same experience another few months from now.
Filed Under: amazon, bricked, consumers, hardware, ownership, robotics, robots, shut down
Companies: amazon
DailyDirt: How Quickly Can You Solve A Rubik's Cube?
from the urls-we-dig-up dept
The original Rubik’s cube puzzle was invented in 1974, but there were similar puzzles made before — such as a 2x2x2 cube puzzle and a spherical 3x3x3 puzzle. The patents for these toys have expired, but people playing with these puzzles are still going strong. Speedcubing or speedsolving is a competitive sport, and there are variations on the activity to solve it blindfolded or with feet only or with just one hand. If you have a scrambled Rubik’s cube sitting in drawer somewhere, maybe you’ll be able to buy a robot to solve it for you soon.
- The fastest machine for solving a Rubik’s cube can currently do it in less than 2 seconds. This bot beats the previous non-human Rubik’s cube-solve bot (aka The Cubestormer 3) which holds a Guinness World Record of just 3.253 seconds, and it has solved a scrambled Rubik’s cube in less than 1 second. [url]
- Speedcubers (that’s apparently what you call people who solve Rubik’s cubes in seconds) are getting slightly faster at solving this puzzle and have broken the 5 second barrier. Lucas Etter, a 14yo, solved a cube in 4.904 seconds. Clearly, there must be a physical limit to how fast a cube can be solved. How fast can anyone or anything rotate a face of a Rubik’s cube? Multiply that time by 20 (though maybe there’s a way to rotate multiple faces simultaneously) or so. [url]
- There are 43,252,003,274,489,856,000 possible Rubik’s cube positions. The minimum number of turns required to solve a cube from an arbitrary starting position is known as “God’s number” by some folks, and it was established to be 20 turns in 2010. [url]
After you’ve finished checking out those links, take a look at our Daily Deals for cool gadgets and other awesome stuff.
Filed Under: cubestormer 3, god's number, guinness world record, lucas etter, puzzles, robotics, rubik's cube, speedcuber, speedcubing, speedsolving, toys
DailyDirt: Welcoming Our Robot Overlords…
from the urls-we-dig-up dept
Despite our supposed intelligence, humans don’t actually know how our own brains work. But even in our ignorance, we’re still developing algorithms and machines that might catch on to how we learn and think. Google’s autonomous vehicle project has a pretty good driving record, except that the world is messy, and predicting how human drivers will react isn’t always certain — especially when they drive buses. Our relationship with robots is going to be more and more complex in the next few years. We’ll need to recognize when robots are faulty, and that might get harder and harder to do.
- We’ve seen people blindly follow GPS navigation directions that endangered the driver’s safety, so would you expect people to follow a robot unquestioningly in an emergency? Yup. A study at Georgia Tech created a “guide robot” that was purposely made to be unreliable (and human participants in the study were told the robot was broken), but during a faked emergency, humans still followed the robot’s bad directions to evacuate a building — even when they conflicted with clearly-marked exit signs. [url]
- A “Data Science Machine” might have better intuition than teams of humans when it comes to big-data analysis. MIT researchers entered their data machine in three data science competitions, and the machine placed ahead of more than half of the human contestants. [url]
- It’s difficult to really grasp exponential improvements in technology — other than to look back and be amazed that things like the iPhone didn’t exist just a few short years ago. And it’s even more challenging to predict where technology is headed when artificial intelligence projects look so primitive now (and have for some time), but by 2040, we could have significantly more autonomous devices performing complex tasks for us. (Assuming the robots haven’t enslaved us instead….) [url]
After you’ve finished checking out those links, take a look at our Daily Deals for cool gadgets and other awesome stuff.
Filed Under: ai, algorithms, analytics, artificial intelligence, autonomous vehicles, big data, data science, hmi, predictions, robotics, robots, trust
Awesome Stuff: Tech Toys
from the all-play-and-some-work dept
This week, we’ve got a lineup of crowdfunded fun with three high-tech toys, only one of which is designed primarily for kids (and it’s the most mature and productive of the three).
The foam dart arms race continues with the FDL-1, which may be the most fearsome contender yet. It’s a high-power, fully-automatic robotic dart launcher that can be configured as a standalone turret or a handheld blaster. But the truly cool part is how it’s made: apart from the electronic guts, the entire thing can be produced with most average hobbyist 3D printers with a 6″ cube build size (not just high-end professional numbers). All of the schematics, instructions and software is open source and/or Creative Commons ShareAlike, so upon release the FDL-1 will be free and easy for anyone to build and modify. In the mean time, its 3D-printed construction also enables several ways to order one on Kickstarter at different tiers (though the prices of all three are high): as a 3D printing kit that includes components and filament, as an assembly kit with components and pre-printed pieces, or as a fully assembled unit.
Though I’m sure there are plenty of kids who wouldn’t mind getting their hands on an FDL-1, it’s a pretty advanced project with a price tag of several hundred dollars to boot. In the mean time, there’s the Kamibot: a papercraft robot kit designed to teach kids to code. To keep things at a beginner’s level, the robot itself is a single pre-made unit based on open source Arduino, with IR and ultrasound sensors, multicolor LEDs, and a single servo in addition to its dual-motor drive. It’s wirelessly controllable and, more importantly, highly programmable via a robust drag-and-drop “learn to code” interface. To keep things fun and interesting for kids, it also has a bunch of papercraft templates for building cool-looking skins on top of the robot itself, from tanks to Frankenstein.
“Moving seats” that rise and fall and tilt and sway according to what’s on screen were a staple of Universal Studios when I went there as a kid, and if you’d asked me then (or yesterday, for that matter) whether that technology would be coming to the living room anytime soon, I’d probably have dismissed the possibility. Well, the Immersit has shown otherwise: it’s a home system that adds motion and vibration feedback for video games to just about any sofa. It works with PC, X-Box and Playstation and is preconfigured to respond to 120+ games, not to mention a whole bunch of movies (it works with plain old video, too). For games, the motion is based on various signals detected from the game, and can be configured at a granular level to change what motions go with what game actions. For movies, the team is using a combination of software and human adjustment to create motion codes for various movies; the Immersit detects the movie being played, and looks up the appropriate motion track. As with all such devices, it has to be tried to be properly evaluated, and I’d be pretty dubious about dropping $700+ on one without doing so — but the reviews from those who’ve had the chance are so far pretty positive.
Filed Under: 3d printing, awesome stuff, coding, robotics, toys
DailyDirt: Speedy Deliveries Coming Via Robots
from the urls-we-dig-up dept
The FAA hasn’t exactly been quick to figure out how its going to regulate drones, and its current rules are a bit of an arbitrary mess of trying to determine what “commercial use” really means and how to register pilots and/or the UAVs they fly. Still, plenty of projects are moving forward with plans to use (semi-)autonomous robots to deliver packages more efficiently and quickly.
- Google’s (ok, Alphabet’s) Project Wing expects to be a commercial business by 2017 — delivering packages with low-flying drones. It’s still uncertain how these UAVs will get FAA approval by 2017, but if everything goes smoothly (cough!), there could be a low altitude “Class G” airspace specifically for drones in a year or so. [url]
- Starship Technologies, despite its name, is planning to use a 6-wheeled robot on the ground to deliver groceries — instead of any kind of flying contraption. This robot will be able to haul around 20 pounds of cargo at 4 mph with a range of about 4 miles. That’s not exactly impressive, but it doesn’t need to worry about flying into anything or getting shot out of the sky. [url]
- Tacocopters are feasible. It’s not a matter of technology. All sorts of things can be delivered by multicopters — mail, ice cream, drink orders — it’s just a matter of how much it’s worth it to you to do it (FAA approval, the cost of equipment, logistics, etc). [url]
After you’ve finished checking out those links, take a look at our Daily Deals for cool gadgets and other awesome stuff.
Filed Under: class g airspace, commercial use, drones, faa, multicopter, robotics, robots, tacocopter, uav
Companies: alphabet, google, starship technologies
DailyDirt: Can't We Just Play Games For Fun?
from the urls-we-dig-up dept
We’ve seen plenty of advances in game algorithms that make us humans look pretty weak compared to the best chess (and checkers and poker and RPS and air hockey and Flappy bird and…) playing computers. Computers aren’t having any fun beating us at all these games, but they do it nonetheless. As always, let’s just hope they figure out quickly that no one wins at thermonuclear war.
- It seems a bit irrational for humans to keep playing a game that a computer can play better than 99.999999% of all humans, but that doesn’t mean we shouldn’t try to create better and better chess playing algorithms. A deep learning program called Giraffe has taught itself how to play chess at an FIDE International Master level in just three days (on a modern mainstream PC, not a supercomputer). It’s not playing at a (super-)Grandmaster level yet, but it’s also not evaluating millions of moves per second like Deep Blue and other chess supercomputers can. [url]
- Google’s DeepMind AI is beating humans at more classic video games — now up to 31 titles, such as Q*Bert and Zaxxon. However, it hasn’t yet mastered games like Ms. Pac-Man or Asteroids. Phew! We’re not obsolete yet…. [url]
- If you think humans are safe by sticking to sports like soccer, basketball or baseball, you might want to see a few robots in development for playing some of these sports. It might take some time for robots to catch up, but I doubt anyone really wants to play any kind of full-contact sport against a robot, anyway. [url]
After you’ve finished checking out those links, take a look at our Daily Deals for cool gadgets and other awesome stuff.
Filed Under: ai, artificial intelligence, chess, deep blue, deepmind, game algorithms, robotics, robots, sports, video games
Companies: google, ibm
DailyDirt: Lethal Machines
from the urls-we-dig-up dept
Artificial intelligence is obviously pretty far from gaining sentience or even any kind of disturbingly smart general intelligence, but some of its advances are nonetheless pretty impressive (eg. beating human chess grandmasters, playing poker, driving cars, etc). Software controls more and more stuff that come in contact with people, so more people are starting to wonder when all of this smart technology might turn on us humans. It’s not a completely idle line of thinking. Self-driving cars/trucks are legitimate safety hazards. Autonomous drones might prevent firefighters from doing their job. There are plenty of situations that are not entirely theoretical in which robots could potentially harm large numbers of people unintentionally (and possibly in a preventable fashion). Where should we draw the line? Asimov’s 3 laws of robotics may be insufficient, so what kind of ethical coding should we adopt instead?
- An open letter from the Future of Life Institute (FLI) is warning against the possibility of an artificial intelligence (AI) arms race that could threaten humanity. Autonomous weapons are a reality that could hinder beneficial AI research — as well as systematically kill people without “meaningful human control” behind the algorithms. [url]
- Autonomous cars with an ethical code in addition to just software code… are getting increasing attention as the odds of self-driving vehicles on public roads grows ever more likely. If a child runs in front of an autonomous car, should the car swerve to avoid the kid? There is an ethical dilemma inherent in making vehicles that are smart enough to know the difference between a kid and some other moving object, but these questions might be avoided entirely by making smart systems only so smart and no smarter — minimizing liability for the companies making the machines. [url]
- A precursor to an artificial intelligence race might be a supercomputer hardware arms race, and we’re already ordering up a National Strategic Computing Initiative (NSCI) to build an exaflop computer to rival China’s Tianhe-2. Sure, artificial intelligence doesn’t need to be developed on super fast computers, but if fast computers are considered potential weapons, it’s not a huge leap of logic to see a supercomputer arms race as a military threat. [url]
After you’ve finished checking out those links, take a look at our Daily Deals for cool gadgets and other awesome stuff.
Filed Under: ai, algorithms, artificial intelligence, asimov, autonomous vehicles, drones, ethical code, fli, military, national strategic computing initiative, nsci, robotics, supercomputers, tianhe-2, war, weapons
Companies: future of life institute
DailyDirt: Computers Are Learning How To Play More Video Games, But They'll Never Appreciate A Good Game?
from the urls-we-dig-up dept
Researchers can program computers to play all kinds of games and even beat the best humans at them. So far, we’re not worried about AI that can beat us at chess or Jeopardy, but maybe we’ll be more worried when a computer can program another computer to play chess at a grandmaster level. Luckily, there’s at least one billionaire willing to chip in a few million bucks to try to keep Terminators from destroying humanity.
- Google DeepMind has created software that can play old Atari games without humans teaching it how to play — and the AI plays 22 out of 49 games better than expert human players. Those Atari games are more difficult than you might think, but it’s not hard to imagine that humans will be no match for AI playing Pac-man in the near future. [url]
- A Civilization V match pitted 42 computer-controlled players against each other. It wasn’t an endless match (sorry, no Wargames conclusion), and a winner of this “Battle Royale” emerged after 179 turns. [url]
- Flappy Bird is a pretty hard game for humans to play, but a robot can play it without getting tired or frustrated. It’s not exactly a breakthrough in robotics, but this bot demonstrates a big difference in how computers play video games. [url]
If you’d like to read more awesome and interesting stuff, check out this unrelated (but not entirely random!) Techdirt post via StumbleUpon.
Filed Under: ai, artificial intelligence, atari, civilization v, elon musk, flappy bird, game algorithms, machine learning, robotics, video games, wargames
Companies: deepmind, google
DailyDirt: I, For One, Welcome Ping Pong Robots
from the urls-we-dig-up dept
Not too long ago, Kuka filmed an ad hinting that their industrial robot arm was fast enough and had software that enabled it to play ping pong with professional table tennis player, Timo Boll. However, that match was pretty disappointing because it never really showed the robot arm returning a tournament-level serve from Boll (or even returning any kind of shot that wasn’t highly edited to make it look more dramatic). Here are a few ping pong playing robots in reality — and they are not yet ready to compete with humans.
- Japanese electronics maker Omron demonstrated a ping pong robot at the 2014 Ceatec tech expo, and its 600+kg bot can play nicely with a human for over 100 volleys. This robot isn’t exactly going to beat anyone at a game, but it has reflexes in the sub-millisecond range, and presumably, software/hardware upgrades could make it more intimidating. [url]
- German researchers trained a robot to play ping pong, and it can return some gentle shots and keep its returns on the table (for the most part). Katharina Muelling and her colleagues were learning about how to teach robots physical skills by imitation, so maybe if they’d used a professional table tennis player to train their robot…. [url]
- Chinese humanoid robots have played ping pong against each other in a rally lasting 176 strokes. It’s not the most exciting game, but these bots can do both a forehand and backhand stroke — and play against humans, too. [url]
If you’d like to read more awesome and interesting stuff, check out this unrelated (but not entirely random!) Techdirt post via StumbleUpon.
Filed Under: ai, artificial intelligence, game algorithms, katharina muelling, machine learning, ping pong, robotics, robots, table tennis, timo boll
Companies: kuka, omron