Topps Blog

Friday, 24 February 2023

NASA Is Working on a Battery That Won’t Melt on Venus

(Credit: NASA) Venus is considered Earth’s twin due to its size and composition, but the conditions on this world could not be more different. The choking, scorching environment on Venus has killed every probe that attempted a landing in mere minutes, but NASA is planning a new mission to Venus that will last a bit longer. The Long-Lived In situ Solar System Explorer (LLISSE) is a small but mighty Venus lander that will study the planet for several months. NASA’s Glenn Research Center is working with a company to develop a battery that can power that lander in one of the most extreme environments in the solar system.

Building an advanced robotic probe that can survive on Venus is no simple feat. Temperatures on the planet can reach 869 degrees Fahrenheit (465 degrees Celsius), which is hot enough to melt the lead in circuit boards and cook most batteries. Venus’s atmosphere is also highly corrosive, which causes common spacecraft materials like copper to fail quickly. The only landers to reach the planet’s surface were the USSR’s Venera missions, none of which lasted more than a few minutes. NASA plans to have LLISSE (above) operating on Venus for an incredible 60 Earth days.

NASA is working with Advanced Thermal Batteries, Inc (ATB) to develop the power systems for this mission — and it has to be a battery. You can’t use solar panels due to the planet’s thick atmosphere. A radioisotope thermoelectric generator (RTG) like the one used in the Perseverance rover would produce heat, and Venus is already hot enough to threaten the 22-pound lander.

Test battery hardware: High-temperature thermal batteries adapted for the Venus surface. (Credit: Dr. Michael Barclay, Advanced Thermal Batteries)

The heat-loving battery being developed for LLISSE is based on the short-lived thermal batteries used in smart missiles, and this technology could be ideal for use on Venus. The 17-cell battery developed by ATB uses a special high-temperature electrolyte that is solid and inert at normal Earth temperatures. When heated to high temperatures, the battery instantly provides high power output, and there’s plenty of heat to spare on Venus.

The main issue with thermal batteries is that they have high self-discharge rates, lasting just a few hours before running out of juice. ATB has created prototype batteries with modified materials and chemistry capable of running for 118 days in a laboratory. With the core capabilities reached, ATB is now working with Glenn Research Center to improve the robustness of the battery to ensure it can survive launch and landing. If successful on Venus, the new thermal battery could be useful in other environments where traditional batteries cannot operate, like in the atmosphere of a gas giant or inside high-temperature jet engines.

NASA plans to launch LLISSE aboard a Russian Venera-D mission, which is expected to leave Earth for Venus in 2029. It will include a Russian orbiter and a larger lander that won’t last long on Venus. LLISSE will have a compact suite of sensors to measure winds, radiance, temperature, pressure, and the concentration of important chemical compounds.

Now read:

Thursday, 23 February 2023

Nvidia CEO Says AI Will Allow Anyone to Be a Programmer

Nvidia’s H100 GPU (Credit: Nvidia) ChatGPT and generative AI has become all the rage in recent weeks. Nvidia has done its share of talking up the technology recently. The company’s CEO has hailed it as the next big thing in computing. This week, Jensen Huang expanded on that idea a bit. He said he believes AI will allow any human to become a computer programmer. A human could tell an AI what it wants to create, and the AI will do the coding.

Huang’s remarks were delivered in this week’s earnings call for fiscal Q4 of 2023. The company ended up beating expectations and saw strong earnings in its data center products. As expected, there was plenty of discussion about Nvidia’s AI products. When it came to the Q/A portion, though, Jensen went off on an extended tear about ChatGPT’s future. In the transcript saved by PC Gamer, Huang said ChatGPT caught everyone by surprise with one simple trick: its ability to “perform tasks and skills that it was never trained to do.”

By that, Huang means it can essentially translate human language into computer code. “It can be prompted in human language, but output Python, output Cobalt, a language that very few people even remember, output Python for Blender, a 3D program,” he said. “So it’s a program that writes a program for another program.”

Nvidia is already seen as a front-runner in the new “AI arms race.” (Credit: Nvidia)

Programs writing programs is certainly an interesting idea and makes us consider when humans can be removed from the equation altogether. Huang explained that in this scenario, human language is the new programming language. “We now realize — the world now realizes that maybe human language is a perfectly good computer programming language,” he said. With large language model AIs, he thinks it will allow anyone to write code. “We’ve democratized computer programming for everyone, almost anyone who could explain in human language a particular task to be performed.”

He concluded his remarks by saying this “new era of computing” will be “utterly revolutionary” for the entire world. “Everybody who develops software is either alerted or shocked into alert or actively working on something that is like ChatGPT to be integrated into their application or integrated into their service,” he said, noting this demand is global. Huang says the level of activity surrounding its own AI products has “gone through the roof” in the past 60 days.

Nvidia’s CEO could be onto something here. Just recently, an AI already helped with 100 semiconductor designs. Google is also building an AI to code, and it’s doing a decent job, too. Jensen has also likened ChatGPT to the iPhone in its ability to show non-tech-savvy folks what AI can accomplish. However, it’s yet to be demonstrated that ChatGPT or a competitor can spit out code when prompted by a human without an engineering background. If that does come to pass, it could begin to make a lot of software engineers pretty nervous. Still, a human will be needed to deliver the prompts, so that should give them some job security.

Now read:

Qualcomm Shrinks AI Art Generator to Run on an Android Phone

Creating art normally takes skill and dedication, but now it just takes a few keystrokes to create original AI art good enough to win contests. Most of the AI tools you’ve seen have been running on powerful servers in the cloud. Qualcomm just showed what’s possible on a smartphone with its latest ARM chip. The company has adapted the Stable Diffusion art generator to run on a smartphone, and it’s reportedly as fast as it is on a powerful desktop machine.

Stable Diffusion is one of a multitude of AI tools that can take a text prompt and turn it into original artwork. In Qualcomm’s new demo of Stable Diffusion on Android, it went with “super cute fluffy cat warrior in armor.” Running on the Snapdragon 8 Gen 2, it took the compressed Stable Diffusion just 14.4 seconds to generate a 512 x 512 cat image. In late 2022, Apple released optimizations to get Stable Diffusion running on its Core ML framework. However, the current-gen iPhone takes about a minute to generate an image.

Qualcomm and others are focusing more on machine learning workloads in the latest chips — similar projects on older phones might take up to an hour to create one image. While the Snapdragon 8 Gen 2 has a ton of advanced GPU and CPU cores, that’s not what matters here. Credit for the record-setting mobile AI art goes to the Hexagon digital signal processor, which has been overhauled in recent generations with AI-specific enhancements.

The company’s engineers used a technique called quantization to shrink the Stable Diffusion model and make it suitable to run on a smartphone. It started with the v1.5 release of Stable Diffusion in FP32 format, using its latest AI research to compress the model to INT8 (8-bit data space). This makes the model more efficient and faster on mobile hardware, while technologies like Qualcomm’s Adaptive Rounding maintain the model’s accuracy without the need for retraining.

Qualcomm says this is the beginning of “edge AI,” an era in which our personal devices will have enough power to run AI models locally instead of relying on the cloud. Running a large, advanced AI model like Stable Diffusion on a phone means it could also run on devices with similar hardware, such as VR/AR headsets, tablets, and laptops.

The Snapdragon 8 Gen 2 was announced late last year and has just started shipping in devices like the OnePlus 11 and Samsung Galaxy S23. It will be common in new Android phones throughout 2023, but you might not be able to enjoy the full might of its AI processing. Qualcomm created its version of Stable Diffusion for internal testing and makes no mention of releasing it publicly.

Now read:

Apple Might Be Closer to Adding a Non-Invasive Glucose Monitor to the Watch

(Image: Luke Chesser/Unsplash) Apple has been toying with the idea of a no-prick glucose monitor Watch integration for a while. According to sources close to the project, it’s finally made some significant headway, and an Apple Watch with a built-in, noninvasive glucose checker could come to market in the next few years.

Anonymous sources told Bloomberg this week that Apple has taken a significant step toward its goal of introducing a continuous, noninvasive blood glucose monitor. Apple’s highly secretive Exploratory Design Group (known internally as XDG) has been working on the tech for years, but progress has been slower than the company would like.

Previously, Apple’s version of a noninvasive blood glucose monitor was large enough to need to sit atop a table. The company has since managed to shrink the system to the size of an iPhone. A device of this size can be strapped to a person’s bicep to monitor their blood sugar continuously, but Apple isn’t gunning for a bicep wearable—it specifically wants to integrate the technology into the Apple Watch.

In 2010, founder Steve Jobs directed Apple to buy noninvasive blood glucose monitoring startup RareLight to bring its technology to mobile devices. RareLight’s developments were kept separate from Apple’s through project codenames and distinct employee badges, which is why the public hasn’t known to look for Apple-integrated glucose monitoring tech until recently. In the meantime, Apple has gradually added health tech to the Watch, including a pulse oximeter, an electrocardiogram (ECG), and heart rate monitoring.

(Image: Raagesh C/Unsplash)

Most continuous glucose monitors (CGMs) rely on a tiny sensor inserted into the skin of a person’s arm or abdomen. This pinprick-style sensor is coated in glucose-sensing enzymes, which detect “blood sugar” in the fluid between blood cells. The sensor sends its findings to a transmitter that sits above the skin, which sends data to a receiver or the person’s smartphone. Although some people find CGMs more convenient than test strips (which require regular finger pricks), they’re expensive, thanks to the frequent sensor and transmitter replacements. CGMs cost anywhere from $100 to $1,000.

If Apple integrates a CGM into the Watch, it could revolutionize diabetes care. Apple Watches start at $249; even if Apple were to make the CGM exclusive to its most expensive Ultra model ($799), its monitor wouldn’t require pinpricks or regular replacements. There are also the apparent bonuses of working alongside other elements of a person’s Apple ecosystem and facilitating tasks that have nothing to do with blood glucose checks. Sources for Bloomberg also say Apple’s tech would alert prediabetic Watch users, allowing them to make changes that could ultimately prevent adult-onset (Type 2) diabetes.

Now Read:

Russian Propagandists Are Using Paid Twitter Blue Checks to Spread Disinformation

(Image: Brett Jordan/Unsplash) Ever since Elon Musk made it possible for Twitter accounts to purchase the little blue “verification” check marks next to their names, it’s been challenging to determine in passing what information is reliable and what isn’t. Anyone can pose as a celebrity or major brand with a blue check and the right profile picture, and if virtual passersby don’t take the time to double-check whether the blue check was paid for, they can easily be duped.

This was already bad enough when Twitter users began impersonating brands in November. Now, things are arguably far worse. According to a new report from The Washington Post, Russian propagandists have started using paid Twitter Blue checkmarks to spread disinformation. While some of their tweets revolve around economic and social falsities (and outright bigotry), most involve disinformation about Russia’s invasion of Ukraine.

One of the dozen Russian propagandists uncovered by Reset.

The trend was first uncovered by Reset, a research initiative focused on how technology intersects with democracy. Reset located a dozen Twitter accounts using paid blue checks to disseminate Kremlin-aligned Russian propaganda under the guise of defying the vague and ubiquitous buzzword “wokeness.” While a couple of these accounts have just a few hundred followers, others have tens of thousands, who help to spread propaganda via likes and retweets. One such account was even given a boost by Musk himself, who replied to tweets saying thousands of NATO troops had died in Ukraine. (They hadn’t, as NATO never deployed troops to Ukraine to begin with.)

Russian propagandists’ abuse of Twitter Blue further proves how dangerous a paid verification model can be for major social platforms. Social media has become a go-to for people seeking quick information without complicated jargon, and while that certainly has its disadvantages, none are as artificially-produced as a site that invites users to impersonate other users. As Reset has helped to show, such an environment is ripe for manipulation.

Given Twitter’s lack of a safety advisory board and tendency to allow hate speech to proliferate, it’s unlikely the site will do anything about Reset’s findings anytime soon. A separate Washington Post report from this month showed that Twitter makes millions off of controversial accounts due to the levels of engagement they solicit, and Russian propagandists’ accounts appear to be no different. According to public engagement data, one propagandist’s account has reached the 10 million tweet impressions mark—a statistic that likely has Twitter execs seeing dollar signs.

Now Read:

Intel Xeon W CPU Sneak Peek Shows It Nearing 1,000 Watts of Power Consumption

Intel recently announced its lineup of Sapphire Rapids Xeon CPUs for workstations. They’re still a month or two away, but in the lead-up to the on-sale date, Intel invited German overclocker de8auer to its labs for a quick overclocking test. That some Sapphire Rapids workstation CPUs are overclockable is both a first for Intel and a big selling point. To demonstrate the possibilities, de8auer and an Intel employee showed some impressive preliminary results, despite it being on pre-release hardware. However, the performance comes at the cost of serious power consumption.

The chip being tested here is the flagship W9-3495X, which has 56 cores and 112 threads. The motherboard is an Intel validation board, so it’s fairly pedestrian as far as VRM cooling goes. The CPU is cooled by a 360mm AIO made by Cooler Master. There’s 192GB of DDR5 running at 5,600MHz across eight channels. As the host notes, a system like this would probably start at around $10,000.

In the video, the Intel employee runs Geekbench 5 Pro but CPU-Z is open to monitor clock speeds. For the test, all 56 cores were running at 4.2GHz. For context, this CPU has a maximum single-core turbo frequency of 4.6GHz and a Turbo Boost 3.0 frequency of 4.8GHz. In other words, this is a modest all-core overclock.

Despite the above reality, it was still able to smash the world record in Geekbench. The title was previously held by a 64-core AMD Threadripper Pro CPU that scored 48,025 points. The Sapphire Rapids CPU crushed that to the tune of a 53,817 multi-core score. That’s a 12% delta, impressive as the Intel CPU has fewer cores and was also running at lower clocks: 4.2GHz versus 4.3GHz on the Threadripper.

While the chip was busy setting a new world record in Geekbench, it seems to have also set a new record for power consumption. Power draw was measured at the wall with a Watts Up? meter, and it showed it hitting 1,097W briefly. When running the single-threaded portion of the benchmark, it hovered around 375W before spiking up to around 700W, then 800W, and so forth. After going over 1,000W, it settled back down to around 650W. de8auer noted that with better cooling, it could have probably overclocked to around 4.6GHz.

If you’re not into overclocking, Cinebench was also run in “stock” trim on an Asus workstation board. One interesting highlight of the visual walkaround is the Asus board offers dual 24-pin power connectors in case you want to run two PSUs. Overall the system hit around 70,000 in Cinebench with only XMP enabled on the memory. The W9-3495X CPU was running all cores at 2.9GHz. For comparison, our colleagues at PCMag got 39,022 out of a Core i9-13900K in its review.

One final note; right as we went to press, we noticed someone has indeed already applied exotic cooling to this CPU. According to a new submission to the HWBOT leader boards for Cinebench, a W9-3495X now holds the top spot. A user named Safedisk applied some LN2 cooling to the chip, allowing it to hit 5.2GHz on all cores. That enabled a mind-bending score of over 128,000 on Cinebench.

Now Read:

Wilson 3D-Printed a Futuristic Basketball That Never Goes Flat

You’ve probably heard of “reinventing the wheel,” but what about reinventing the ball? That’s what engineers at Wilson are up to with a new airless basketball prototype. The team leveraged additive manufacturing technology to make a new kind of basketball from a loose hexagonal lattice. You can’t buy one, but Wilson plans to work toward making it more than a prototype.

The airless prototype has the same size and shape as a traditional basketball, with eight lobe-like panels covering the surface. Instead of a solid leather skin, the prototype ball is hollow with large hexagonal holes. So, calling it “airless” is a bit of a misnomer — there’s plenty of air in the ball, but it doesn’t stay there. The ball bounces because of the structure’s elastic support, not a pressurized bladder inside.

Wilson started by designing the ball in a virtual environment. Then, it partnered with a 3D printing firm called EOS to create the airless prototype. Most 3D printing involves putting down layers of melted plastic with an extruder nozzle, but EOS has a unique system that has already found use in medical devices and aerospace. It uses a resin powder that solidifies when hit with a laser. Like a 3D printer, the EOS technique builds the ball layer by layer by etching the structure into the powder.

The finished structure can be dyed any color, but Wilson opted to go with an understated black for the prototype. The company says the prototype has been designed to replicate the properties of a standard basketball without the air. That means it should weigh the same with an identical rebound when bounced. However, Wilson doesn’t mention how putting holes all over the surface of the ball might affect aerodynamics. There is surprisingly little bouncing (zero) in the video above. The commercial-style video below does show a little more action.

The ball debuted at the NBA’s Slam Dunk Contest this past weekend. As this is just a prototype, Wilson won’t sell you an airless basketball yet. You can, however, sign up for Wilson’s mailing list to be made aware of any airless ball innovations. The NBA, which is helping to test the ball, has no plans to switch to an airless design, but you never know what the future will bring.

Now read: