Robotics & AI

Katakem’s ‘robot chef’ speeds up drug development with reliable chemistry • ZebethMedia

By zebethcontrolPosted on October 20, 2022Posted in Battlefield 200, Biotech & Health, Disrupt, Katakem, Robotics & AI, Startups, TechCrunch Disrupt 2022No Comments

Organic chemist Manuela Oliverio was working on a new drug when he noticed that test results on mice weren’t consistent, because the molecule being administered was always different depending on the chemist who produced it. It occurred to him that automation and robotics could make the drug development process more predictable, and so he founded Katakem, one of the startups in the ZebethMedia Disrupt Battlefield 200. With Katakem, Oliverio aims to develop what he calls a “robot chef” for chemists — a device that makes chemical reactions more consistently reproducible while accelerating the experimental process. He claims that the current prototype, dubbed OnePot, can collect data about chemical processes 150 times every second and automate repetitive, mundane tasks like heating, cooling and mixing different molecules. “The production of a chemical product is strictly regulated and standardized. [But] the development phase between discovery and production is still carried out manually and no significant data is extracted,” Oliverio told ZebethMedia in an interview. “Through data, we can help companies develop new life-saving drugs faster and, of course, this means higher revenues and better margins for them … Data [from OnePot] is reliable, clean and immediately usable.” To Oliverio’s point, drug development today is a lengthy and expensive endeavor. Only about 12% of drugs entering clinical trials are ultimately approved for introduction by the U.S. Food and Drug Administration. And estimates of the average R&D cost per drug range from less than $1 billion to over $2 billion, with errors and mistakes adding to the price tag. Katakem developed OnePot over the course of three years, designing both the mechanical and electrical components in-house. The company is seeking chemists to beta test the device, particularly those in corporate and academic settings, to collect data that it plans to use to train an algorithm that can recommend “faster and more sustainable” ways to develop molecules. Image Credits: Katakem Image Credits: Katakem Given the size of the problem — and addressable market — it’s not surprising that Katakem has competition. Automata is also creating a robot to handle basic lab tasks, and it recently raised $50 million to do so. There’s Kebotix, a startup developing AI and robotics tools to expedite the discovery of chemicals, and Artificial, which sells a lab automation platform aimed at life sciences R&D. But while Katakem has the dual challenges of proving its technology works and overcoming rivals, Oliverio isn’t concerned. Based on existing commitments, he expects Katakem’s annual recurring revenue to hit $350,000 by the end of the year and $3 million by the end of 2023. Presumably, those projections assume Katakem finds success with its early customers and demonstrates that OnePot does all the company says it can do. “As our clients — chemical companies — are key to economies, we are not subject to high variability in demand,” Oliverio said. “The robot is ready to be commercialized.” To date, Calabria, Italy-based Katakem has raised €1.3 million ($1.27 million) in capital from undisclosed seed investors, according to Crunchbase data.

France fines Clearview AI maximum possible for GDPR breaches • ZebethMedia

By zebethcontrolPosted on October 20, 2022Posted in AI regulation, clearview ai, GDPR, Privacy, Robotics & AINo Comments

Clearview AI, the controversial facial recognition firm that scrapes selfies and other personal data off the Internet without consent to feed an AI-powered identity-matching service it sells to law enforcement and others, has been hit with another fine in Europe. This one comes after it failed to respond to an order last year from the CNIL, France’s privacy watchdog, to stop its unlawful processing of French citizens’ information and delete their data. Clearview responded to that order by, well, ghosting the regulator — thereby adding a third GDPR breach (non-cooperation with the regulator) to its earlier tally. Here’s the CNIL’s summary of Clearview’s breaches: Unlawful processing of personal data (breach of Article 6 of the GDPR) Individuals’ rights not respected (Articles 12, 15 and 17 of the GDPR) Lack of cooperation with the CNIL (Article 31 of the RGPD) “Clearview AI had two months to comply with the injunctions formulated in the formal notice and to justify them to the CNIL. However, it did not provide any response to this formal notice,” the CNIL wrote in a press release today announcing the sanction [emphasis its]. “The chair of the CNIL therefore decided to refer the matter to the restricted committee, which is in charge for issuing sanctions. On the basis of the information brought to its attention, the restricted committee decided to impose a maximum financial penalty of 20 million euros, according to article 83 of the GDPR [General Data Protection Regulation].” The EU’s GDPR allows for penalties of up to 4% of a firm’s worldwide annual revenue for the most serious infringements — or €20M, whichever is higher. But the CNIL’s press release makes clear it’s imposing the maximum amount it possibly can here. Whether France will see a penny of this money from Clearview remains an open question, however. The US-based privacy-stripper has been issued with a slew of penalties by other data protection agencies across Europe in recent months, including €20M fines from Italy and Greece; and a smaller UK penalty. But it’s not clear it’s handed over any money to any of these authorities — and they have limited resources (and legal means) to try to pursue Clearview for payment outside their own borders. So the GDPR penalties look mostly like a warning to stay away from Europe. Clearview’s PR agency, LakPR Group, sent us this statement following the CNIL’s sanction — which it attributed to CEO Hoan Ton-That: “There is no way to determine if a person has French citizenship, purely from a public photo from the internet, and therefore it is impossible to delete data from French residents. Clearview AI only collects publicly available information from the internet, just like any other search engine like Google, Bing or DuckDuckGo.” The statement goes on to reiterate earlier claims by Clearview that it does not have a place of business in France or in the EU, nor undertake any activities that would “otherwise mean it is subject to the GDPR”, as it puts it — adding: “Clearview AI’s database of publicly available images is lawfully collected, just like any other search engine like Google.” (NB: On paper the GDPR has extraterritorial reach so its former arguments are meaningless, while its claim it’s not doing anything that would make it subject to the GDPR looks absurd given its amassed a database of over 20 billion images worldwide and Europe is, er, part of Planet Earth… ) Ton-That’s statement also repeats a much-trotted out claim in Clearview’s public statements responding to the flow of regulatory sanctions its business attracts that it created its facial recognition tech with “the purpose of helping to make communities safer and assisting law enforcement in solving heinous crimes against children, seniors and other victims of unscrupulous acts” — not to cash in by unlawfully exploiting people’s privacy — not that, in any case, having a ‘pure’ motive would make any difference to its requirement, under European law, to have a valid legal basis to process people’s data in the first place. “We only collect public data from the open internet and comply with all standards of privacy and law. I am heartbroken by the misinterpretation by some in France, where we do no business, of Clearview AI’s technology to society. My intentions and those of my company have always been to help communities and their people to live better, safer lives,” concludes Clearview’s PR. Each time it has received a sanction from an international regulator it’s done the same thing: Denying it has committed any breach and refuted the foreign body has any jurisdiction over its business — so its strategy for dealing with its own data processing lawlessness appears to be simple non-cooperation with regulators outside the US. Obviously this only works if you plan for your execs/senior personnel to never set foot in the territories where your business is under sanction and abandon any notion of selling the sanctioned service to overseas customers. (Last year Sweden’s data protection watchdog also fined a local police authority for unlawful use of Clearview — so European regulators can act to clamp down on any local demand too, if required.) On home turf, Clearview has finally had to face up to some legal red lines recently. Earlier this year it agreed to settle a lawsuit that had accused it of running afoul of an Illinois law banning the use of individuals’ biometric data without consent. The settlement included Clearview agreeing to some limits on its ability to sell its software to most US companies but it still trumpeted the outcome as a “huge win” — claiming it would be able to circumvent the ruling by selling its algorithm (rather than access to its database) — to private companies in the U.S. The need to empower regulators so they can order the deletion (or market withdrawal) of algorithms trained on unlawfully processed data does look like an important upgrade to their toolboxes if we’re to avoid an AI-fuelled dystopia. And it just so happens that the EU’s

Battlefield bots • ZebethMedia

By zebethcontrolPosted on October 20, 2022Posted in Actuator, Hardware, MIT, Robotics & AI, TechCrunch Disrupt 2022No Comments

Greetings from the bowels of Moscone Center West. As I type this, Kevin Hart just exited the stage and Serena Williams is presiding over a packed house. No exaggeration: I attempted to make my way to grab a seat in the few rows up front allotted to the ZebethMedia staff, but I physically couldn’t get through the crowd. A solid one-two punch to kick off this Wednesday morning. I’ve had a little time to walk the halls here, mostly scouring for hardware and robotics firms, as is my wont. It’s always fun to see the sorts of microcosms that develop at events like this, identifying groupings that are indicative of broader current and future trends in the startup world. I’m happy to say for my own edification that robotics firms, in particular, were well represented. Not sure that’s something I would have felt comfortable asserting five or so years back. Coupled with all of the various ongoing market indicators, it truly feels like we’ve comfortably entered a new era for robotics and robotic investing. Yesterday I hosted what amounted to a two-hour marathon pitch-off, which involved 30 startups offering two-minute pitches. It was a bit exhausting, frankly, but I’m looking forward to unpacking some of those offerings in the coming weeks. One definitely warrants mention in this week’s Actuator, because I ended up speaking with the CEO and profiling the firm late last week—Touchlab. Image Credits: Touchlab Touchlab was the winner of our TC Sessions: Robotics event back in July, so this thing is long overdue. One bit that’s especially interesting to me is how the company’s outward focus has shifted in that short time. The Edinburg-based firm originally pitched us on its robotic skin. The applications are pretty clear there — effectively adding another layer of sensing to supplement existing vision systems and the like. That’s still the core of the startup’s play, but Touchlab has also begun to implement its own technology into a robotic system. It showcased an eldercare robot that is essentially an off-the-shelf TIAGo++ robot, outfitted with its sensor technology. Eldercare makes sense, as a highly pressure-sensitive sensor is required to interact with human patients — the elderly in particular. “We have a layer of software that translates the pressure of the skin to the suit. We’re also using haptic gloves,” co-founder and CEO Zaki Hussein told me. “Currently, our skin gathers a lot more data than we can currently transmit to the user over haptic interfaces. So there’s a little bit of a bottleneck. We can use the full potential of the best haptic interface of the day, but there is a point where the robot is feeling more than the user is able to.” The haptic sensations are translated into a wearable suit donned by a VR-wearing operator. I’m interested in exploring the state of teleoperation a bit more. There’s a weird sort of stigma around this technology in a category where everyone seems to be constantly chasing full autonomy. Image Credits: RIF Robotics RIF Robotics (pronounced “riff”), another one of the entries in the Battlefield 200, operates in a similar space. Specifically, it’s building systems designed to streamline the disinfecting of medical equipment in-hospital. Co-founder Kevin DeMarco tells ZebethMedia: The major challenges that the sterile processing industry is facing are a lack of experienced surgical technicians, instrument-level tracking, infection traceability and cost traceability. Medical device manufacturers are interested in knowing how their equipment is used and degrades in the field. Instrument-level data will also help them to decide where to send sales reps. Hospitals are interested in instrument-level data because it will help them operate more efficiently by improving instrument-level tracking and instrument inspection. Currently, most hospitals only track at the tray-level, but the industry wants to be able to track at the instrument level. Image Credits: Katakem I’m starting to sense a theme emerging here — one more healthcare robotics firm from my time at the Showcase stage. Kyle’s headline really says it all here: “Katakem is developing a robot to automate drug development.” The firm has developed what it deems a “robot chef,” designed to create chemical reactions. It tells ZebethMedia: The production of a chemical product is strictly regulated and standardized. [But] the development phase between discovery and production is still carried out manually and no significant data is extracted. Through data, we can help companies develop new life-saving drugs faster and, of course, this means higher revenues and better margins for them … Data [from OnePot] is reliable, clean and immediately usable. Image Credits: Jasper Montreal-based Jasper is taking a unique approach toward a market controlled by the likes of Seamless, DoorDash and Uber Eats. The firm’s play revolves around the deployment of a proprietary chain of automated ghost kitchens designed to dramatically speed up food delivery. The robotics aspect comes in through the kitchen, allowing for minimal or no staff for the food preparation process. “Having good meals at home is expensive or time consuming … Food delivery is highly inefficient — restaurants or ghost kitchens prepare meals worth a few dollars and then pay someone to ship them across town,” CEO Gunnar Froh told ZebethMedia. “While most customers aren’t aware of this, about half of their dollars are spent on platform fees and delivery costs. By running robotic kitchens in or next to residential high-rises, Jasper eliminates labor and delivery inefficiencies to offer residents freshly prepared gourmet meals at the cost of home cooking. Jasper meals are plated on porcelain, which allows its clients to cut up to a third of their household waste.” Swap Robotics at ZebethMedia Startup Battlefield at ZebethMedia Disrupt in San Francisco on October 18, 2022. Image Credits: Haje Kamps / ZebethMedia A couple of robotics-focused firms made it onstage for the Battlefield pitch-offs as well. Swap has developed an electric mower specifically designed to cut vegetation around solar farms. “Right now, there are a couple of main challenges when cutting all of the vegetation in solar fields,” the company tells

Want to know how a dress looks on you? AIMIRR has your back… and front • ZebethMedia

By zebethcontrolPosted on October 20, 2022Posted in AIMIRR, Commerce, Pritesh Kanani, Robotics & AI, Startups, virtual try onNo Comments

With many people buying clothes online versus in-store where they can try them on in the dressing room, finding the right fit can be a challenge. AIMIRR is taking on this challenge by bringing the dressing room to the customer with its real-time garment rendering technology that overlays images of clothing on a live video of the individual. Founder and CEO Pritesh Kanani was exhibiting Seattle-based AIMIRR’s technology as part of the Battlefield 200 at ZebethMedia Disrupt and announced that the company closed on an exclusive partnership with Chicago custom clothing marketplace Balodana Inc. for its fitting room technology. The company is also officially launching its first product and service in November after closing over 10 partnerships in Chicago and Seattle, including a collaboration with Fashionbar at Chicago Fashion Week, taking place this week. AIMIRR’s core virtual try-on technology shows the garment in 3D down to the size, shape and texture, including showing how the garment will fit as the individual moves. “We are not just designing apparel filters, we are developing a graphical digital fitting room experience that remains true to a shopper’s body over any online shopping website,” Kanani told ZebethMedia. He got the idea for the company in 2020 while he was getting married. His grandmother wanted to pass down her wedding dress to his fiancee, but then the global pandemic hit. With his grandmother in India and his fiancee in the United States, it was difficult to get the dress there and to know if it would fit. Kanani recalls looking for options to help and decided instead to leverage his seven years in the computer vision and graphics industry building vision video creation tools to start AIMIRR. He honed the idea while part of the University of Chicago’s Polsky Accelerator program through which he got $120,000 to develop the technology. The company has been operational for about four months now and has been offering a $49-per-month trial with a group of retailers to provide the technology on 10 of their garments. The clothing brands host the technology on their websites and are able to gather data about the fit and popularity of the garments from the app. Currently, customers scan a QR code with their phone to activate the technology using their device’s camera. Kanani said the next iteration will involve an embedded link to create the experience on a laptop. The company has largely been bootstrapped so far, but he has plans to attend two more accelerator programs and will raise a seed round in 2023. “Our next steps will be increasing revenue and getting from the small business segment to our enterprise partnership,” Kanani said. “In the next six to nine months we will complete our shipping to production on the partnership that we have, and then finish off with the partnerships we are targeting. Beyond that, we will acquire funding to get into an enterprise beyond those 10 garments.”

RIF Robotics powers robots that inspect and organize surgical equipment • ZebethMedia

By zebethcontrolPosted on October 20, 2022Posted in Battlefield 200, Biotech & Health, Disrupt, RIF Robotics, Robotics & AI, TechCrunch Disrupt 2022No Comments

Several years ago, Kevin DeMarco’s aunt was an operating room nurse who asked DeMarco — knowing that he programmed robots for a living — if there was a robot that could prepare surgical equipment. After investigating the problem with a colleague at Georgia Tech, where DeMarco was working as a research faculty member, he decided to leave his position to create the robot that his aunt once mused about. In this quest, DeMarco ended up co-founding RIF Robotics, one of the startups in the ZebethMedia Disrupt Battlefield 200. Led by DeMarco, Sergio García-Vergara and a third co-founder, Collin Farill, who’s an industrial designer by trade, RIF Robotics seeks to use a combination of AI and robotics to relieve healthcare workers of the burden of mundane tasks so they can focus on clinical work. Image Credits: RIF Robotics Sterile processing — the cleaning of medical equipment — is also tough on technicians performing it, who have to spend hours each day inspecting and cleaning tools. Some equipment requires over 100 steps to disinfect, and the pace in busy hospitals can be relentless. The costs can add up, too. One study estimates that just 20 instrument errors that end up creating delays in the operating room can cost a hospital as much as $3,385. Extrapolating out to a year, the cost to the hospital would be about $48,000, the research found. RIF isn’t tackling cleaning. But the startup claims its prototype product, which was developed in less than two months, can save surgeons time by identifying, classifying and manipulating four different instruments and assembling a small surgical tray. Two machine learning systems — an image segmentation system and an object classifier, trained on sets of both real and synthetic images of surgical tools — help a robotic manipulator arm grasp and move the instruments. “The major challenges that the sterile processing industry is facing are a lack of experienced surgical technicians, instrument-level tracking, infection traceability and cost traceability,” DeMarco told ZebethMedia in an interview. “Medical device manufacturers are interested in knowing how their equipment is used and degrades in the field. Instrument-level data will also help them to decide where to send sales reps. Hospitals are interested in instrument-level data because it will help them operate more efficiently by improving instrument-level tracking and instrument inspection. Currently, most hospitals only track at the tray level, but the industry wants to be able to track at the instrument level.” Image Credits: RIF Robotics Future prototypes will be able to recognize more tools and determine if there’s any leftover “bioburden” (i.e., blood and bone) on instrument surfaces and evaluate instruments’ sharpness and overall condition. But even in its current form, DeMarco believes that RIF has built a product hospitals would use. “Three Atlanta hospitals and the Veterans Affairs are interested in our product,” he said. “We have a collaborative research and development agreement with the Veterans Affairs, which allows us to conduct customer discovery and pilot studies at their facilities … [We’ll deploy] three alpha versions of our systems at local Atlanta hospitals, where we already have existing connections.” RIF is currently bootstrapping — DeMarco claims that the company has a burn rate of less than $1,000 per month. But the team isn’t naïve about the long road ahead. RIF is going after an $800,000 debt pre-seed round and hopes to hire a medical device industry expert after the round concludes. The company, which is pre-revenue, also expects to require three rounds of funding and close to four years before it reaches profitability. RIF Robotics’ co-founders pose for a photograph in scrubs. Image Credits: RIF Robotics There’s also competition from vendors like RST Automation, which sells a semi-automated medical tool identification and organization system. Steris and R-Solution Medical — two other rivals, albeit not direct ones — are developing robots to transport and store surgical trays and equipment. DeMarco claims that RIF’s solution is more capable. But the proof will be in the pudding — RIF aims to turn its prototype into a manufacturable product by fall 2023. “The healthcare industry is starving for innovation,” DeMarco said. “We are protecting ourselves from the potential headwinds by developing products and solutions that are directly asked for by the industry and the end users.”

Adobe’s AI prototype pastes objects into photos while adding realistic lighting and shadows • ZebethMedia

By zebethcontrolPosted on October 19, 2022Posted in Adobe, AI, image editing, Media & Entertainment, Robotics & AINo Comments

Every year at Adobe Max, Adobe shows off what it calls “Sneaks,” R&D projects that might — or might not — find their way into commercial products someday. This year is no exception, and lucky for us, we were given a preview ahead of the conference proper. Project Clever Composites (as Adobe’s calling it) leverages AI for automatic image compositing. To be more specific, it automatically predicts an object’s scale, determining where the best place might be to insert it in an image before normalizing the object’s colors, estimating the lighting conditions and generating shadows in line with the image’s aesthetic. Here’s how Adobe describes it: Image composting lets you add yourself in to make it look like you were there. Or maybe you want to create a photo of yourself camping under a starry sky but only have images of the starry sky and yourself camping during the daytime. I’m no Photoshop wizard, but Adobe tells me that compositing can be a heavily manual, tedious and time-consuming process. Normally, it involves finding a suitable image of an object or subject, carefully cutting the object or subject out of said image and editing its color, tone, scale and shadows to match its appearance with the rest of the scene into which it’s being pasted. Adobe’s prototype does away with this. “We developed a more intelligent and automated technique for image object compositing with a new compositing-aware search technology,” Zhifei Zhang, an Adobe research engineer on the project, told ZebethMedia via email. “Our compositing-aware search technology uses multiple deep learning models and millions of data points to determine semantic segmentation, compositing-aware search, scale-location prediction for object compositing, color and tone harmonization, lighting estimation, shadow generation and others.” Image Credits: Adobe According to Zhang, each of the models powering the image-compositing system is trained independently for a specific task, like searching for objects consistent with a given image in terms of geometry and semantics. The system also leverages a separate, AI-based auto-compositing pipeline that takes care of predicting an object’s scale and location for compositing, tone normalization, lighting condition estimation and synthesizing shadows. The result is a workflow that allows users to composite objects with just a few clicks, Zhang claims. “Achieving automatic object compositing is challenging, as there are several components of the process that need to be composed. Our technology serves as the ‘glue’ as it allows all these components to work together,” Zhang said. As with all Sneaks, the system could forever remain a tech demo. But Zhang, who believes it’d make a “great addition” to Photoshop and Lightroom, says work is already underway on an improved version that supports compositing 3D objects, not just 2D. “We aim to make this common but difficult task of achieving realistic and clever composites for 2D and 3D completely drag-and-drop,” Zhang said. “This will be a game-changer for image compositing, as it makes it easier for those who work on image design and editing to create realistic images since they will now be able to search for an object to add, carefully cut out that object and edit the color, tone or scale of it with just a few clicks.”

Makersite lands $18M to help companies manage product supply chains • ZebethMedia

By zebethcontrolPosted on October 19, 2022Posted in Climate, environment, funding, Makersite, Planet A Ventures, Robotics & AI, startup, Startups, supply chain, sustainabilityNo Comments

In 2018, Neil D’Souza, a software engineer by trade and previously the VP of product development at Thinkstep, came to the realization that his ten-plus-year effort to solve enterprise product challenges in the areas of sustainability, compliance and risk were having little impact. The way he saw it, they took too long, which minimized their influence on product design choices. “For example, analyzing a car’s life cycle assessment can easily take an automotive company an entire year,” D’Souza told ZebethMedia in an email interview. “Speed matters, otherwise the analysis just becomes a meaningless report.” That frustration was the genesis of his startup, Makersite, which aims to produce near-instant impact assessments in the areas of sustainability, compliance and risk to inform corporate-level decisions. Makersite, D’Souza says, is an attempt to bridge the gap between experts who know what “good” looks like from an environmental, cost, compliance or risk perspective and decision makers with control over the product supply chain. With over 30 customers including Microsoft, Cummins and Vestas and a balance sheet showing profitable operations over the last few years, Makersite is beginning to attract investor attention, this week securing $18 million in a Series A round with participation from Planet A Ventures. D’Souza says the tranche — Makersite’s first besides “a few convertible notes”; the company was bootstrapped until now — will be put toward work with integrators and resellers and expanding the size of Makersite’s team. “There are many companies out there that specialize in solving cost, compliance, risk or sustainability challenges. The problem is they each sit in siloes and the data they use is specialized to the people who work in those fields,” D’Souza said. “That’s what makes our solution different. We’re unique in the space as we’re the first to solve the challenge of bringing multi-criteria decision analysis to non-experts.” Using AI, Makersite maps a company’s product data against a material and supply chain database, generating automated reports. The idea is to help companies meet their sustainability goals while minimizing costs and keeping compliance at the forefront. The aforementioned database — which D’Souza says is among the largest of its kind — allows Makersite to identify contextual relationships to build a model of products and their supply chains automatically. The models cover not just what a product is made out of, but how every component or ingredient is manufactured — all the way from the mining resources to the factory floor. “[Makersite] enables a customer to drop in a bill of material for, say, a wind turbine, tell the AI that it’s a wind turbine, answer a few questions (e.g., about power output), and the system will automatically build a ‘cradle-to-grave’ model of that turbine that’s localized to where it’s made and where it’ll be erected,” D’Souza explained. “That allows you to optimize designs of specific elements of the turbine — like the tower and nacelle — to locally available resources and infrastructure, such as recycling facilities, and understand trade-offs across the lifecycle and criteria, like cost, risks and regulations.” As Makersite grows its headcount from around 40 employees to over 100 over the next 12 months, D’Souza says that the focus will be on building out the company’s sales and marketing teams to grow business particularly in the U.S. and Europe. On the integration side, Makersite’s investing capital in connectors to software like Autodesk to deliver cost and environmental insights within computer-assisted design platforms. “There is a paradigm shift towards sustainable products which are driven by regulation, competition, customer demand and investments,” D’Souza said. “For that, Makersite enables procurement and product design professionals to make day-to-day decisions without the need for compliance, sustainability, cost or risk experts.”

Deep Render believes AI holds the key to more efficient video compression • ZebethMedia

By zebethcontrolPosted on October 19, 2022Posted in Battlefield 200, Deep Render, Disrupt, machine learning, Media & Entertainment, Robotics & AI, TechCrunch Disrupt 2022No Comments

Chri Besenbruch, CEO of Deep Render, sees many problems with the way video compression standards are developed today. He thinks they aren’t advancing quickly enough, bemoans the fact that they’re plagued with legal uncertainty and decries their reliance on specialized hardware for acceleration. “The codec development process is broken,” Besenbruch said in an interview with ZebethMedia ahead of Disrupt, where Deep Render is participating in the Disrupt Battlefield 200. “In the compression industry, there is a significant challenge of finding a new way forward and searching for new innovations.” Seeking a better way, Besenbruch co-founded Deep Render with Arsalan Zafar, whom he met at Imperial College London. At the time, Besenbruch was studying computer science and machine learning. He and Zafar collaborated on a research project involving distributing terabytes of video across a network, during which they say they experienced the shortcomings of compression technology firsthand. The last time ZebethMedia covered Deep Render, the startup had just closed a £1.6 million seed round ($1.81 million) led by Pentech Ventures with participation from Speedinvest. In the roughly two years since then, Deep Render has raised an additional several million dollars from existing investors, bringing its total raised to $5.7 million. “We thought to ourselves, if the internet pipes are difficult to extend, the only thing we can do is make the data that flows through the pipes smaller,” Besenbruch said. “Hence, we decided to fuse machine learning and AI and compression technology to develop a fundamentally new way of compression data getting significantly better image and video compression ratios.” Deep Render isn’t the first to apply AI to video compression. Alphabet’s DeepMind adapted a machine learning algorithm originally developed to play board games to the problem of compressing YouTube videos, leading to a 4% reduction in the amount of data the video-sharing service needs to stream to users. Elsewhere, there’s startup WaveOne, which claims its machine learning-based video codec outperforms all existing standards across popular quality metrics. But Deep Render’s solution is platform-agnostic. To create it, Besenbruch says that the company compiled a dataset of over 10 million video sequences on which they trained algorithms to learn to compress video data efficiently. Deep Render used a combination of on-premise and cloud hardware for the training, with the former comprising over a hundred GPUs. Deep Render claims the resulting compression standard is 5x better than HEVC, a widely used codec and can run in real time on mobile devices with a dedicated AI accelerator chip (e.g., the Apple Neural Engine in modern iPhones). Besenbruch says the company is in talks with three large tech firms — all with market caps over $300 billion — about paid pilots, though he declined to share names. Eddie Anderson, a founding partner at Pentech and board member at Deep Render, shared via email: “Deep Render’s machine learning approach to codecs completely disrupts an established market. Not only is it a software route to market, but their [compression] performance is significantly better than the current state of the art. As bandwidth demands continue to increase, their solution has the potential to drive vastly improved commercial performance for current media owners and distributors.” Deep Render currently employs 20 people. By the end of 2023, Besenbruch expects that number will more than triple to 62.

Swap Robotics is paving the way for electric solar vegetation cuts and sidewalk snow plowing • ZebethMedia

By zebethcontrolPosted on October 19, 2022Posted in Battlefield, battlefield 2022, Disrupt, Robotics & AI, Startups, Swap RoboticsNo Comments

Swap Robotics, a company that manufactures electric grass-cutting and snow removal robots, presented today at ZebethMedia Disrupt Startup Battlefield to detail how it’s making sustainable outdoor work equipment. For the next few years, 95% of the startup’s focus will be on facilitating robots that cut grass and vegetation on 1,000+ acre utility-scale solar farms. The company’s secondary focus is sidewalk snow plowing. The startup was founded in October 2019 by CEO Tim Lichti, CTO Mohamed H. Ahmed, Machine Design Lead Spencer Kschesinski and Electrical Design Lead Adonis Mansour. Lichti, Kschesinski and Mansour all attended the University of Waterloo together and then got to know Ahmed during their first year. The team originally planned to develop a robotic cutting solution for sports fields, but kept hearing from landscapers that cutting 1,000+ acre utility-scale solar installations was a challenging job that could use a modern solution. Tim Lichti, CEO at Swap Robotics pitches as part of ZebethMedia Startup Battlefield at ZebethMedia Disrupt in San Francisco on October 18, 2022. Image Credits: Haje Kamps / ZebethMedia The team decided it would be their mission to create a solution that could sustainably cut grass in a controlled environment. Swap Robotics was aware that solar vegetation cutting comes with its challenges, as it requires a unique type of cutting deck that is able to get underneath solar panels, and recognized that a robotic solution could address the problem. “Right now, there are a couple of main challenges when cutting all of the vegetation in solar fields,” Lichti told ZebethMedia in an interview. “The way it’s done is unsustainable. It’s done by gasoline or diesel-powered equipment, so there’s obviously a big carbon footprint there. There’s also a high cost from gasoline and diesel itself. The equipment is also going through rough terrain, so there’s a lot of equipment breakdown and costs associated with that. Since what we’re doing is 100% electric, it’s a lot more sustainable. There are also way fewer parts, so it’s not going to break down nearly as often.” Image Credits: Swap Robotics The robots have built-in hydraulics that move the grass cutting blades and the snow plow attachment. The attachments have a “quick swap” system, hence the name Swap Robotics, to make it easier and quicker to switch attachments. The robots’ batteries can also be swapped in five minutes, which allows for nearly 24/7 operation. The robots can also hold more than 1,000 pounds. Within 60 days of debuting its robots in mid-2022, Swap Robotics had over $9 million of signed agreements for solar vegetation cutting. Swap Robotics says it has developed the world’s first 100% electric cutting deck to reach the grass and vegetation underneath solar panels. The company also says it has developed the world’s first 100% electric “rough cut” deck that can easily cut down vegetation up to two-inches in diameter. Lichti says Swap Robotics currently has several robots in commercial operation in Texas, but is unable to disclose which companies are currently using the robots. The startup is also in the midst of releasing a batch of 10 new robots and has ordered supplies for the next batch of 10 robots. The company anticipates additional sales in the future as a result of its new relationship with SOLV Energy. As for the company’s business model, Swap Robotics charges a price per acre. Lichti says the model is convenient because customers are already familiar with paying a price per acre for grass cutting done by humans. The price per acre can vary depending on factors such as the size of the site, frequency of cuts and the terrain. The startup’s goal is to provide customers with 15% to 20% in savings when compared to their current cutting costs per acre. Image Credits: Swap Robotics The startup also announced that it received an investment from SOLV Energy, the largest utility-scale solar building company in the United States, but is unable to disclose the amount. Lichti says the funding is part of its pre-Series A round that it plans to close at the end of October. A large portion of the investment will go toward commercialization of the startup’s robots. The company plans to ramp up operations to have dozens of robots in service. In addition, some parts of the investment will be used for capital expenditure. Prior to this investment, Swap Robotics raised $3 million in the three years after its launch from angel investors and SOSV. The funding was used to get Swap Robotics’ initial batch of a dozen robots into commercial operation. The investment was also used for software, mechanical and electrical development. Swap Robotics at ZebethMedia Startup Battlefield at ZebethMedia Disrupt in San Francisco on October 18, 2022. Image Credits: Haje Kamps / ZebethMedia Swap Robotics plans to have a larger Series A round in 2023. “Long term, we would love Swap Robotics to be an outdoor robotics platform for work,” Lichti said. “We’ve developed a form factor that is compact, extremely strong and robust and has a built-in hydraulics system that can have dozens of different use cases. I think this makes it an ideal platform for heavy use cases, especially those that sometimes may be seasonal.” Lichti reiterated that the startup’s main focus will largely be on solar vegetation, and that the potential for its additional use cases is part of its longer term vision. As for what these use cases could look like, Lichti noted that the robots could potentially be used for street sweeping or reforestation efforts.

Meet E-liza Dolls, the startup that’s building dolls to help young girls learn to code • ZebethMedia

By zebethcontrolPosted on October 19, 2022Posted in Battlefield 200, Disrupt, e-liza dolls, Robotics & AI, Startups, TechCrunch Disrupt 2022No Comments

E-liza Dolls, a Berkeley-based startup, is aiming to challenge the gender gap in STEM by helping young girls learn to code using dolls. The company, which exhibited as part of the Battlefield 200 at ZebethMedia Disrupt, builds dolls that include programmable computers that girls can code through an app. The startup was founded in 2021 by Eliza Kosoy, a Ph.D. student at UC Berkeley, who is focused on the intersection of child development and artificial intelligence. Kosoy originally came up with the idea for the dolls in 2017 while she was working at MIT in an AI lab that was mostly made up of men. Kosoy says she realized that if only a certain group of people were designing the future of AI and technology, it would only benefit that group, which is when she had the idea to come up with a way for young girls to learn to code. Kosoy wanted to find a way for girls to learn about coding without having to give up their interests, which is why she decided to combine dolls and technology. Regardless of what people may think about gendered toys, the purpose of E-liza dolls is to help girls feel confident when it comes to exploring STEM by giving them a product that is designed specifically for them. The market is filled with toys that are designed and marketed for and by males. Of course girls can play with these toys too, but some of them may prefer to play with something that is designed for them. “We want to expose young girls to technological concepts and encourage creative thinking through hardware and software, preventing girls from being influenced by generational stereotypes,” Kosoy told ZebethMedia. “Parents have so few options; they feel they need to force their daughters to play with STEM products designed for boys in order to get their daughters on a STEM path. We believe little girls don’t have to sacrifice their interests in order to play with educational STEM toys.” E-liza Dolls is currently in talks with manufacturers and plans to launch on Kickstarter in early 2023. Kosoy says the team is one prototype away from the Kickstarter launch, as the startup plans to add a few iterations to the dolls and enhance their design features. After the initial launch on Kickstarter, the company plans to release the product officially in mid-2023. The 18″ dolls operate via a piece of hardware embedded in each doll. The device has a screen and is Bluetooth-enabled to receive code via the doll’s companion app. Girls can plug in different sensors or use the built-in sensors to code the doll to do different things, such as building a security alarm for your room using a distance sensor or creating a truth detector using a heartbeat pulse sensor. Since launching, E-liza Dolls has received $100,000 in funding from AIX Ventures. The company is currently in the midst of raising a pre-seed round that consists of funding from several angel investors, including poet Rupi Kaur.