1 2 3 4 5 |
|
app/build.gradle
中添加以下配置:1 2 3 4 5 6 7 |
|
excludes
排除重复文件即可解决。1 2 3 |
|
app/build.gradle
中添加依赖替换规则:1 2 3 4 5 |
|
com.google.ai.edge.litert
,若项目同时包含新旧包名,会导致类冲突。通过依赖替换强制使用新包解决。1 2 3 4 |
|
android/gradle.properties
文件中添加:1
|
|
I subscribe to Proud Mary’s monthly “deluxe” offering for $39. Every month they send me 100 grams of a special coffee. For June they sent Sitio da Pedra’s Cocarive Auction #1, which is a naturally processed Arara (a variety I’ve never had before).
The beans smell lovely. They were roasted on June 2nd, so they have been resting a little over two weeks, which may be the perfect time.
Here are the ground beans (Comandante C40 at 25 clix) in the sifter:
28 grams of ground coffee yields 20 grams of consistent grind once you filter out the fines using a sifter. It’s very important that you get this step of the coffee making right when making pour-over. Coffee that is too finely ground will extract quickly and make it taste too strong and muddy. Spend $100 on a sifter, you won’t regret it!
Here’s the perfectly consistent grind (no grind finer than 800 microns):
I use a Chemex (hand-blown of course, and pronounced “sha-may”) to make coffee. 200 degree Fahrenheit water with three pours: first pour to 50 grams (30 seconds), second pour to 150 grams (90 seconds), third and final pour to 300 grams (150 to 180 seconds).
A “three-cup” Chemex makes one cup of coffee. A “five cup” Chemex makes two cups of coffee. That’s a three-cup Chemex below:
This method makes a perfect 10 ounce cup of coffee.
This Arara coffee from Sitio da Pedra in Brazil is very nice. Clean and good coffee flavor, a pretty strong coffee aftertaste, very little acidity (sourness), maybe a touch at first but it’s subtle and a complex sour (not a bad sour). This is good coffee. I’d recommend trying it if you have never had Arara.
Earlier:
I paid $52 for 100 grams of Esmeralda Buenos Aires 2NC, which will make around four cups of coffee. This is a naturally-processed Geisha. It’s no longer in stock, but the web page remains. I’m glad to see that Blendin Coffee Club no longer scrubs the web pages of the coffees they sell once they are out of stock.
The beans were SUPER hard to grind, which is often the case with Geishas. It’s like grinding up little stones. Strong berry aroma.
Here’s the 28 grams of grind in the sifter after using the Comandante C40 at 25 Clix. This will yield 20 grams of perfect grind after sifting.
No grounds smaller than 800 microns. This is all-important when making pour-over. When I see people dump coffee ground straight from the grinder into the filter, it makes me shudder. Do you even sift, bro?
Usual recipe: 15:1, 200-degree Fahrenheit water, three pours, 50 ml to 30 seconds, 150 ml to 90 seconds, 300 ml to 150-180 seconds. Hand-blown sha-may. When I see people make coffee in anything but a Chemex, it makes me shudder. Yes, again.
Lovely color.
This Geisha is much more coffee-forward than other Geishas I’ve tried. The really superb Geisha don’t taste of coffee at all on the tongue. It’s like a sneak attack. The coffee flavor hits and hits hard only after a minute. This Buenos Aires 2NC has coffee right up front, which makes me think it’s a lesser Geisha, but I could be wrong about that. It’s still good. Maybe my expectations were too high. Anything from Hacienda La Esmeralda commands a super premium price, even if it’s one of their lesser crops. That’s my opinion anyway.
Earlier:
Reliable Robotics is one of a couple of companies that are outfitting Cessna Caravan airplanes with advanced software to provide a high level of autonomy, for applications that include cargo transportation. Reliable Robotics
The economic argument for autonomy is even more compelling in the emerging air-taxi industry, where hundreds of hopefuls—including a dozen or so serious contenders—are racing to develop electric vertical takeoff and landing aircraft to ferry passengers around crowded urban areas. Most of these eVTOLs are the size of helicopters, with space for just four or five passengers, and their proponents envision scores or even hundreds of them in the air over major cities, collectively moving millions of passengers annually. The concept is called urban air mobility, and in the speculative math that underpins it, eliminating the expense of a pilot and freeing up another seat for a paying passenger are seen as key to maximizing profits and scale.
China has already certified a pilotless air taxi: the EH216-S, a two-seat multicopter developed by Guangzhou-based EHang that in March obtained initial approval from the Civil Aviation Administration of China for limited commercial sightseeing operations. However, many Western observers doubt that EHang’s design would pass muster by the U.S. Federal Aviation Administration (FAA) or the European Union Aviation Safety Agency (EASA), both of which have an especially conservative approach to safety. For that reason, most Western eVTOL makers have opted to develop piloted aircraft first and plan to introduce autonomous versions at some later date. They figure that seeking certification of novel electric aircraft designs, even without autonomy, is already a big ask of these regulators.
A notable exception to this strategy is Wisk Aero, which began as a project funded by Google cofounder Larry Page and is now a wholly owned subsidiary of Boeing. In January 2022, the company declared that it would obtain FAA certification for its self-flying air taxi by the end of the decade and be operating close to 14 million flights annually within five years after that—a staggering ambition, given that the entire U.S. air traffic system currently manages around 16 million flights per year. While overheated expectations around urban air mobility have cooled considerably in the three years since that announcement, Wisk continues to forge ahead with its autonomous Generation 6 eVTOL, the company’s sixth aircraft design and the first it plans to certify for passenger-carrying operations.
A mockup of Wisk’s sixth generation of electric vertical takeoff and landing aircraft was unveiled in October 2022. Wisk
Importantly, Wisk, Reliable Robotics, and Merlin Labs aren’t just developing autonomous aircraft—they have already launched formal certification programs with the FAA. That means they’re working closely with the agency to define the rules and standards by which autonomous aircraft will be approved for commercial operations, blazing a trail for others to follow. The task is a daunting one, but the regulators and industry are not starting from scratch. Rather, they’re building on decades of certification experience and best practices that have helped to dramatically improve the safety of the aviation industry over its history.
Merlin Labs launched the flight test campaign for its certification-ready autonomy system in June 2024. The Merlin Pilot system is integrated directly onto the aircraft and is intended in the near term to reduce crew workload rather than fully replace pilots.Merlin Labs
“Once you have the regulatory basis in place, then you need to come to agreement on how you’re going to demonstrate compliance to all of those regulations,” says Rose. “You can pull from existing standards, you can modify existing standards, or you can, in some cases, even just propose your own standards.” After agreeing upon the means of compliance, the applicant and regulator develop a detailed project plan that outlines the tests that will be performed and the reports—known as artifacts—that will be submitted to the regulator to support certification.
For conventional piloted aircraft with a history of real-world operations, much of how those aircraft will function in the national airspace system is assumed. “Large commercial airplanes operate from airports around the world with relatively known and static equipment that helps them navigate and approach and land,” says Brian Yutko, until recently Wisk’s CEO (he now heads commercial airplane product development at Boeing). This infrastructure, he adds, has been established over decades and is reflected in the design of aircraft in ways that are often taken for granted.
The existing system relies heavily on human pilots communicating with air traffic controllers over radio. Autonomous aircraft will require new concepts of operations, or “ConOps,” for how they will function, which could include using ground supervisors to handle radio calls, for example. In turn, the specifics of each ConOps will influence aircraft design requirements. According to Comer, crystallizing the ConOps at the beginning of the certification process “helps drive a common understanding of what you’re actually doing, and that may be different for every applicant with the FAA.”
Basically, Wisk intends for its autonomous air taxi, which Yutko has likened to “a tram in the sky,” to fly along very specific and limited routes with predetermined emergency landing locations. Such a narrow set of tasks is an easier thing to automate than the varied and flexible operations performed by most small piloted aircraft today (or, for that matter, most self-driving cars). Meanwhile, human supervisors on the ground will monitor flights and communicate with air traffic control as required.
Reliable Robotics’ automated Cessna Caravans will also have remote operators to handle communications with air traffic control, but they will fly over a much larger and more variable operating area. Because of this added complexity, Reliable has opted to split up the work of certifying its autonomous aircraft into chunks, beginning with certification of an advanced, always-on autopilot. This will assist but not replace the onboard pilot during all phases of flight, including landing as well as taxi and takeoff—which traditional autopilots are not capable of. Taking the pilot out of the cockpit will come as a follow-on certification project.
In the speculative math that underpins urban air mobility, eliminating the expense of a pilot and freeing up another seat for a paying passenger are seen as key to maximizing profits and scale.
Neither Wisk nor Reliable Robotics is using machine learning algorithms in its technical solutions, in large part because there’s no consensus on how to assure, to aviation’s exacting standards, the safety of such algorithms. These algorithms are frequently characterized as “nondeterministic,” meaning that their outputs can’t be reliably predicted from their inputs.
Some autonomous-aircraft developers are incorporating artificial intelligence into their designs. Merlin Labs, for example, is developing natural-language-processing algorithms to communicate with air traffic control. For the most part, however, autonomous-aircraft developers aren’t counting on technology alone to solve the innumerable contingencies that can arise in flight—that’s where the ground operators come in.
“We basically have taken everything that can be [automated] deterministically, and we’re making it deterministic,” Rose explains. “And all of the things that are…very hard to automate, that a human can do easily, then let the human do it.”
Which raises the question: If humans are required to supervise autonomous aircraft, does the business case for them still hold up? Their developers say it does, but in ways that aren’t as simple as just striking “pilots” from the balance sheet. For example, those remote supervisors will need training, but that’s likely to be far less extensive and costly than the training required to competently fly an aircraft. For Reliable Robotics and other companies targeting cargo delivery, autonomy also promises to improve the efficiency of the existing cargo feeder network.
“The reality is, in cargo aircraft, especially small cargo aircraft, pilots are super underutilized,” says Rose. Pilots at the feeder airlines may spend most of their day hanging out in a hotel room between their morning and evening flights. If people were instead managing autonomous cargo aircraft remotely, they could conceivably oversee additional flights across multiple time zones. “Our analysis has shown you can easily double the productivity of a pilot by putting them into our control center, potentially triple or quadruple the productivity [depending] on the mission set,” Rose says.
aspect_ratio
Despite not appearing together right away—the JPEG first appeared in Netscape in 1995, three years after the image standard was officially published—the JPEG and Web browser fit together naturally. JPEG files degraded more gracefully than GIFs, retaining more of the picture’s initial form—and that allowed the format to scale to greater levels of success. While it wasn’t capable of animation, it progressively expanded from something a modem could pokily render to a format that was good enough for high-end professional photography.
For the Internet’s purposes, the degradation was the important part. But it wasn’t the only thing that made the JPEG immensely valuable to the digital world. An essential part was that it was a documented standard built by numerous stakeholders.
We’re going to degrade the quality of this image throughout this article. At its full image size, it’s 13.7 megabytes.Irina Iriser
At 50 percent of its initial quality, the image is down to about 2.6 MB. By dropping half of the image’s quality, we brought it down to one-fifth of the original file size. Original image: Irina Iriser
That same forest, saved at 5 percent quality, now down to about 419 kilobytes.Original image: Irina Iriser
When an image is compressed with DCT, the change tends to be less noticeable in busier, more textured areas of the picture, like hair or foliage. Those areas are harder to compress, which means they keep their integrity longer. It tends to be more noticeable, however, with solid colors or in areas where the image sharply changes from one color to another—like text on a page. Ever screenshot a social media post, only for it to look noisy? Congratulations, you just made a JPEG file.
Other formats, like PNG, do better with text, because their compression format is intended to be non-lossy. (Side note: PNG’s compression format, DEFLATE, was designed by Phil Katz, who also created the ZIP format. The PNG format uses it in part because it was a license-free compression format. So it turns out the brilliant coder with the sad life story improved the Internet in multiple ways before his untimely passing.)
In many ways, the JPEG is one tool in our image-making toolkit. Despite its age and maturity, it remains one of our best options for sharing photos on the Internet. But it is not a tool for every setting—despite the fact that, like a wrench sometimes used as a hammer, we often leverage it that way.
Our forest, saved at 1 percent quality. This image is only about 239 KB in size, yet it’s still easily recognizable as the same photo. That’s the power of the JPEG.Original image: Irina Iriser
Then in 1997, a company named Forgent Networks acquired Compression Labs. The company eventually spotted the patent and began filing lawsuits over it, a series of events it saw as a stroke of good luck.
“The patent, in some respects, is a lottery ticket,” Forgent CEO Jay Peterson told CNET in 2005. “If you told me five years ago that ‘You have the patent for JPEG,’ I wouldn’t have believed it.”
While Forgent’s claim of ownership of the JPEG compression algorithm was tenuous, it ultimately saw more success with its legal battles than Unisys did. The company earned more than US $100 million from digital-camera makers before the patent finally ran out of steam around 2007. The company also attempted to extract licensing fees from the PC industry. Eventually, Forgent agreed to a modest $8 million settlement.
As the company took an increasingly aggressive approach to its acquired patent, it began to lose battles both in the court of public opinion and in actual courtrooms. Critics pounced on examples of prior art, while courts limited the patent’s use to motion-based uses like video.
By 2007, Forgent’s compression patent expired—and its litigation-heavy approach to business went away. That year, the company became Asure Software, which now specializes in payroll and HR solutions. Talk about a reboot.
See the forest JPEG degrade from its full resolution to 1 percent quality in this GIF. Original image: Irina Iriser
Other image technologies have had somewhat more luck getting past the JPEG format. The Google-supported WebP is popular with website developers (and controversial with end users). Meanwhile, the formats AVIF and HEIC, each developed by standards bodies, have largely outpaced both JPEG and JPEG 2000.
Still, the JPEG will be difficult to kill at this juncture. These days, the format is similar to MP3 or ZIP files—two legacy formats too popular and widely used to kill. Other formats that compress the files better and do the same things more efficiently are out there, but it’s difficult to topple a format with a 30-year head start.
Shaking off the JPEG is easier said than done. I think most people will be fine to keep it around.
Ernie Smith is the editor of Tedium, a long-running newsletter that hunts for the end of the long tail.
Neuroprosthesis Reproduces a Man’s Speech UC Davis
In experiments, the scientists found that the BCI could detect key aspects of intended vocal intonation. They had the patient attempt to speak sets of sentences as either statements, which had no changes in pitch, or as questions, which involved rising pitches at the ends of the sentences. They also had the patient emphasize one of the seven words in the sentence “I never said she stole my money” by changing its pitch. (The sentence has seven different meanings, depending on which word is emphasized.) These tests revealed increased neural activity toward the ends of the questions and before emphasized words. In turn, this let the patient control his BCI voice enough to ask a question, emphasize specific words in a sentence, or sing three-pitch melodies.
“Not only what we say but also how we say it is equally important,” Wairagkar says. “Intonation of our speech helps us to communicate effectively.”
All in all, the new BCI could acquire neural signals and produce sounds with a delay of 25 milliseconds, enabling near-instantaneous speech synthesis, Wairagkar says. The BCI also proved flexible enough to speak made-up pseudo-words, as well as interjections such as “ahh,” “eww,” “ohh,” and “hmm.”
The resulting voice was often intelligible, but not consistently so. In tests where human listeners had to transcribe the BCI’s words, they understood what the patient said about 56 percent of the time, up from about 3 percent from when he did not use the BCI.
Neural recordings of the BCI participant shown on screen.UC Davis
“We do not claim that this system is ready to be used to speak and have conversations by someone who has lost the ability to speak,” Wairagkar says. “Rather, we have shown a proof of concept of what is possible with the current BCI technology.”
In the future, the scientists plan to improve the accuracy of the device—for instance, with more electrodes and better AI models. They also hope that BCI companies might start clinical trials incorporating this technology. “It is yet unknown whether this BCI will work with people who are fully locked in”—that is, nearly completely paralyzed, save for eye motions and blinking, Wairagkar adds.
Another interesting research direction is to study whether such speech BCIs could be useful for people with language disorders, such as aphasia. “Our current target patient population cannot speak due to muscle paralysis,” Wairagkar says. “However, their ability to produce language and cognition remains intact.” In contrast, she notes, future work might investigate restoring speech to people with damage to brain areas that produce speech, or with disabilities that have prevented them from learning to speak since childhood.
“The new website is more visual, with video and other media to engage all visitors. It also showcases our global community’s commitment as a public charity advancing technology for the benefit of humanity.”
The next section, the IEEE Is the Global Community of Technology Professionals module, has options to Find Your Path to learn about resources available for industry professionals, authors and researchers, students and young professionals, volunteers, new members, and retirees.
The following section, Latest Innovations, features videos and articles from publications including IEEE Spectrum and The Institute on cutting-edge technology engineers are working on, such as electronic tattoos.
Keep scrolling down and you’ll get to know IEEE members and their thoughts on what’s next for technologies such as artificial intelligence and quantum computing.
“This redesign marks a key milestone in IEEE’s digital transformation,” Muirhead says. “The use of rich media, video content, and dynamic storytelling features allows for deeper engagement with IEEE and understanding its various offerings.
“However, it is just the beginning. In the months ahead, we will continue to enhance the site with new features, updated content, and richer tools.”
This is the first successful vertical takeoff of a jet-powered flying humanoid robot, developed by Artificial and Mechanical Intelligence (AMI) at Istituto Italiano di Tecnologia (IIT). The robot lifted ~50 cm off the ground while maintaining dynamic stability, thanks to advanced AI-based control systems and aerodynamic modeling.
We will have much more on this in the coming weeks!
[Nature] via [IIT]As a first step toward our mission of deploying general-purpose robots, we are pushing the frontiers of what end-to-end AI models can achieve in the real world. We’ve been training models and evaluating their capabilities for dexterous sensorimotor policies across different embodiments, environments, and physical interactions. We’re sharing capability demonstrations on tasks stressing different aspects of manipulation: fine motor control, spatial and temporal precision, generalization across robots and settings, and robustness to external disturbances.
[Generalist AI]
Thanks, Noah!Ground Control Robotics is introducing SCUTTLE, our newest elongate multilegged platform for mobility anywhere!Teleoperation has been around for a while, but what hasn’t been is precise, real-time force feedback.That’s where Flexiv steps in to shake things up. Now, whether you’re across the room or across the globe, you can experience seamless, high-fidelity remote manipulation with a sense of touch.
This sort of thing usually takes some human training, for which you’d be best served by robot arms with precise, real-time force feedback. Hmm, I wonder where you’d find those...?
[Flexiv]The 1X World Model is a data-driven simulator for humanoid robots built with a grounded understanding of physics. It allows us to predict—or “hallucinate”—the outcomes of NEO’s actions before they’re taken in the real world. Using the 1X World Model, we can instantly assess the performance of AI models—compressing development time and providing a clear benchmark for continuous improvement.
[1X]SLAPBOT is an interactive robotic artwork by Hooman Samani and Chandler Cheng, exploring the dynamics of physical interaction, artificial agency, and power. The installation features a robotic arm fitted with a soft, inflatable hand that delivers slaps through pneumatic actuation, transforming a visceral human gesture into a programmed robotic response.
I asked, of course, whether SLAPBOT slaps people, and it does not: “Despite its provocative concept and evocative design, SLAPBOT does not make physical contact with human participants. It simulates the gesture of slapping without delivering an actual strike. The robotic arm’s movements are precisely choreographed to suggest the act, yet it maintains a safe distance.”
[SLAPBOT]
Thanks, Hooman!Inspecting the bowels of ships is something we’d really like robots to be doing for us, please and thank you.
[Norwegian University of Science and Technology] via [GitHub]
Thanks, Kostas!H2L Corporation (hereinafter referred to as H2L) has unveiled a new product called “Capsule Interface,” which transmits whole-body movements and strength, enabling new shared experiences with robots and avatars. A product introduction video depicting a synchronization never before experienced by humans was also released.
[H2L Corp.] via [RobotStart]How do you keep a robot safe without requiring it to look at you? Radar!
[Paper] via [IEEE Sensors Journal]
Thanks, Bram!We propose Aerial Elephant Trunk, an aerial continuum manipulator inspired by the elephant trunk, featuring a small-scale quadrotor and a dexterous, compliant tendon-driven continuum arm for versatile operation in both indoor and outdoor settings.
[Adaptive Robotics Controls Lab]This video demonstrates a heavy weight lifting test using the ARMstrong Dex robot, focusing on a 40 kg bicep curl motion. ARMstrong Dex is a human-sized, dual-arm hydraulic robot currently under development at the Korea Atomic Energy Research Institute (KAERI) for disaster response applications. Designed to perform tasks flexibly like a human while delivering high power output, ARMstrong Dex is capable of handling complex operations in hazardous environments.
[Korea Atomic Energy Research Institute]Micro-robots that can inspect water pipes, diagnose cracks, and fix them autonomously—reducing leaks and avoiding expensive excavation work—have been developed by a team of engineers led by the University of Sheffield.We’re growing in size, scale, and impact! We’re excited to announce the opening of our serial production facility in the San Francisco Bay Area, the very first purpose-built robotaxi assembly facility in the United States. More space means more innovation, production, and opportunities to scale our fleet.
[Zoox]Watch multipick in action as our pickle robot rapidly identifies, picks, and places multiple boxes in a single swing of an arm.
[Pickle]And now, this.
[Aibo]Cargill’s Amsterdam Multiseed facility enlists Spot and Orbit to inspect machinery and perform visual checks, enhanced by all-new AI features, as part of their “Plant of the Future” program.This ICRA 2025 plenary talk is from Raffaello D’Andrea, entitled “Models are Dead, Long Live Models!”
Will data solve robotics and automation? Absolutely! Never! Who knows?! Let’s argue about it!