Latest news with #JoanCros
Forbes
03-06-2025
- Business
- Forbes
Revolutionizing Eye Tracking With Event-Based Sensing
HoloLens 2, a AR headset designed by Microsoft, exhibited during the Mobile World Congress, on ... More February 28, 2019 in Barcelona, Spain. (Photo by Joan Cros/NurPhoto via Getty Images) The Integrated Visual Augmentation System (IVAS) was a $22B project awarded to Microsoft in 2018 to develop Augmented Reality (AR) headsets for the US Army. As part of the funding, Microsoft was on contract to deliver 120,000 units of its HoloLens equipped headsets, but discontinued its production in 2024. In early 2025, Anduril Industries, a defense-tech startup, announced it would takeover the program. Anduril specializes in AI driven hardware technologies with a mission of 'Transforming US & allied military capabilities with advanced technology'. Palmer Luckey is CEO of Anduril. Funded by a suite of notable venture capital investors, it boasts a range of defense products spanning terrestrial, air, underwater and outer space defense capabilities. Palmer Luckey developed Oculus, a leading AR/VR platform, and sold it to Facebook (now Meta) in 2014. He left Meta in 2017 and started Anduril. Meta has evolved the Oculus platform into the Meta Quest headset family. Equipped with inward facing sensors, it enables eye tracking features which have various advantages. Meta is active in research on eye-tracking using event-based cameras from Prophesee. Use of event based cameras (versus conventional frame based cameras) provide significant benefits in speed, sampling rate, accuracy, background and noise rejection. It also dramatically reduces size, weight and heat generation, critical features for user comfort and battery life as commercial and military deployments of these headsets accelerate. Going full circle, Anduril and Meta recently announced a collaboration to work together to build to extended reality (XR) devices for the U.S. military. This effort will leverage the IVAS program that Anduril assumed charge of in early 2025. Mark Zuckerberg and Palmer Luckey finally see eye to eye! Prophesee, a pioneer and market leader of event-based neuromorphic vision technology recently announced a collaboration with Tobii, the global leader in eye tracking and attention computing. The goal is to deliver a next-generation event-based eye tracking solution tailored for AR/VR and smart eyewear applications. Given the discussion above on the fusion of IVAS, Meta and Anduril capabilities for AR/VR headsets, this is an exciting development. Event based sensing asynchronously reads out only the pixels which record intensity changes above a user defined threshold, resulting in lower latency, power, compute, storage and thermal management requirements. Prophesee ('predicting and seeing where the action is'), based in France, uses its event-based cameras for industrial automation, AR/VR, security, healthcare and AoT™ (Autonomy of Things) applications. Founded in 2014, the company closed its C round funding of $50M in October 2022. To date, it has raised $127M. Prophesee's focus is on providing its sensor chips, SDK (Software Development Kit) and camera reference designs to end users like Tobii and others who develop imaging systems for various applications. It's next generation MetaVision GenX 320 QVGA sensor (320 x 320 pixels, 6 μm pixel pitch, 4 mm x 3 mm footprint) is designed for eye tracking in gaming and wearable AR/VR headsets. Figure 1 highlights some of the capabilities: Figure 1: Metavision GenX 320 QVGA Sensor Capabilities Luca Verra is the CEO of Prophesee. A key initiative underway is the development of an even leaner event sensing chip than the MetaVision QVGA for smart glass applications. Reducing the pixel pitch by 50% (< 3 um) realizes a 5X reduction in the chip area. While the MetaVision sensor consumes ~25 mW of power at a system level (includes the sensing readout, signal pre-processing and a 900 um wavelength LED), the dedicated chip for smart glasses aims to reduce power consumption by > 10X, and provide data speeds approaching 1 KHz (similar to that reported in the Meta paper above). Compared to this, a CMOS based framed camera solution consumes ~250 mW (not compatible with smart glass applications) and data speeds of ~120 Hz. Data speeds are important in eye tracking since it provides finer resolution, important for various psychological aspects like emotional state, intent and interest. Mr. Verra is excited about the collaboration with Tobii: "Event-based vision is a perfect match for the growing demand for low-power, always-on sensing in next-generation wearable devices. This partnership builds on a strong foundation of collaboration between our teams and expands our joint capabilities into the eye-tracking space. With Tobii's proven leadership in eye tracking, combined with our experience in event-based sensing, we aim to set a new standard for what's possible in AR/VR and smart eyewear.' Tobii, founded ~ 20 years ago, is headquartered in Sweden, and bills itself as a leader in eye tracking (delivered the world's first remote eye tracker). It serves thousands of enterprises, universities, and research institutes around the world, and is a pioneer in attention computing (Figure 2) which it defines as 'technology that understands human attention and intent'. Listed on the Swedish NASDAQ, it has 600 employees and annual revenues of ~$90M. Figure 2: Factors Going Into Attention Computing Using Eyeball Tracking Tobii uses machine learning, artificial intelligence, and advanced signal processing to decode head and eye movements, translate microscopic gestures into accurate gaze signals, and generate insights to reveal what captures a person's attention and interpret intent. Applications fall into four broad verticals: 1) Consumer research and user experience: transitions the market research process from a formal survey-based one (in which questions posed to respondents invariably create bias) to a emotional one in which subtle cues gained from eye tracking like gaze, pose and presence can provide valuable information to modify products and packaging. The data can be gathered in formal or informal settings (like when customers shop and are looking at alternate products). 2) Developmental psychology and healthcare research: Eye tracking data offers valuable insights into human behavior and cognitive processes, making it a versatile technology across a wide range of research domains like psychology, medical diagnostics and education. 3) Consumer technologies in gaming and wearables: System-on-chip solutions for eye tracking provide seamless interactions between users and their devices in wearables, smart glasses and AR/VR headsets. Sony is one of Tobii's customers (Sony PlayStation product lines). 4) Automotive, driver monitoring systems (DMS): Tobii Autosense enables OEMs to build for tomorrow, pushing the boundaries of what's possible, and achieve true in-cabin differentiation and safety monitoring. Tobii's hardware comprises a suite of wearable eye trackers. Software integration and ensuring that the devices work robustly across all populations (age, skin color, eye color, illumination conditions, vibrations, facial and eye-makeup, etc.) are critical challenges. To date, the company has sold a few million units. According to CEO Anand Srivasta, the eye tracking solutions market is in its infancy. He expects this number to grow as the adoption of this technology accelerates, especially in the consumer wearables, industrial productivity and DMS markets. Event based sensing can deliver significant disruptions in certain verticals where low size, weight, power consumption and heat generation are critical (wearables) and high data rates can enhance the eye tracking resolution to provide more insights into consumer intent and emotion. Other applications may continue to use framed CMOS cameras (when full images are needed, like in DMS). He expects the collaboration with Prophesee to 'complement to our existing camera technologies, giving our customers even more options and flexibility when designing their products. Smart eyewear is one of the most demanding segments for eye tracking—requiring ultra-low power, high performance, and seamless integration into a standard glasses form factor." As technology accelerates, data generation volume increases exponentially. In many cases, most of this data is useless, non-eventful and non-actionable. Attempting to analyze all of it slows down decision making and insights, and risks missing events that are important. Event based sensing has a significant role to play in lowering compute, storage and thermal management costs, providing higher actionable data rates and opening up new markets and technological acceptance.
Forbes
02-05-2025
- Business
- Forbes
Top 5 Decentralized Data Collection Providers In 2025 For AI Business
Adam Selipsky CEO of Amazon Web Service (AWS), speaking at the Keynote: Delivering a new World, ... More Barcelona, Spain, on March 01 2022. (Photo by Joan Cros/NurPhoto via Getty Images) The world runs on data, and businesses increasingly rely on it. However, traditional data sourcing methods often present challenges related to diversity, transparency, privacy, and cost. This article reviews the current state of decentralized data collection and outlines key steps for wisely selecting a decentralized data provider—along with a shortlist of top options to consider. Traditionally, centralized data collection involves gathering data from various sources—such as apps, devices, or websites—and sending it to a single central server or database controlled by one organization. This data is collected via APIs, sensors, tracking tools, or manual input. The biggest bottleneck of this model for AI's future and for businesses is the inability to collect truly 'global' and 'diverse' data from different regions and cultures. Decentralized data collection addresses this by leveraging blockchain technology. It enables small-scale cross-border payments which encourages global users to contribute data voluntarily in exchange for incentives—something that centralized or Web2 platforms cannot achieve. Another key aspect is transparency. Centralized AI and data collection are often criticized for operating as " black boxes," lacking transparency and accountability. People have no idea how and where they collect these data for their business. Furthermore, it's difficult to verify whether data is collected lawfully and ethically. In contrast, decentralized data collection enhances transparency by recording the data collection process on blockchain and storing data across multiple independent nodes rather than under a single authority. This blockchain-powered structure allows users to trace how and where their data is used efficiently, reduces the risk of hidden manipulation, and ensures that no single party can alter or monopolize the data without broad consensus. As a result, decentralized solutions are emerging as a strong alternative for businesses seeking more robust data strategies. By leveraging blockchain technology, decentralized data collection enhances both data diversity and verifiability, opening access to new, previously untapped data sources. Businesses interested in exploring decentralized data collection should: Below are five noteworthy platforms operating in the decentralized data collection space, outlining their core functionalities and potential business applications. Core offering: Decentralized data marketplace for AI and ML datasets. Strengths: Best for: Anyone looking to buy/sell datasets or run compute-to-data workloads. Example: access a specific medical imaging dataset to train a diagnostic AI, with the data provider maintaining control over the data itself. Website: Core offering: Decentralized knowledge agent platform and AI data marketplace. Strengths: Best for: AI developers looking to build autonomous agents trained on community-owned or enterprise-specific knowledge bases. Example: Collect a large and diverse dataset of user reviews to train a sentiment analysis AI agent. Website: Core Offering: Decentralized data collection and labeling solution for AI. Strengths: Best For: Enterprises needing diverse, real-world, and structured datasets to train or fine-tune AI models. Example: Collect a 50-language and high-quality dataset for a specialized natural language processing AI. Website: Core offering: Decentralized platform for users to control, monetize, and pool personal data for AI. Strengths: Best for: Building AI models with ethically sourced, user-consented personal data, especially in social, health, and lifestyle domains. Example: Users can leverage Vana to own, control, and monetize their personal data by contributing it to community-led AI projects Website: Core offering: Real-time data network for decentralized data streams. Strengths: Best for: AI systems that rely on live data feeds like autonomous vehicles, smart cities, or trading bots. Example: If your AI business focuses on predicting traffic patterns, you could use Streamr to access real-time data feeds from connected vehicles and sensors. Website: As AI continues to scale, the true bottleneck won't be algorithms—it will be data. Success in the coming wave of AI innovation hinges on timely access to high-quality, well-labeled, and diverse datasets. Yet, efficient data collection infrastructure remains in its infancy. Forward-thinking organizations that invest in scalable, ethical, and AI-ready decentralized data collection solutions now will be the ones leading the industry tomorrow. The age of intelligent data sourcing isn't a trend—it's the next mainstream.
Forbes
28-04-2025
- Business
- Forbes
Visa's Earnings Could Impress, But Tariff Risks Loom Large
The Visa logo, the American multinational payment card services corporation and the world's ... More second-largest card payment organization, appears on a MacBook in Barcelona, Spain, on February 10, 2025. (Photo by Joan Cros/NurPhoto via Getty Images) Visa (NYSE:V) is scheduled to announce its earnings on April 29. According to consensus estimates, revenues are anticipated to increase by roughly 9% compared to the previous year, reaching $9.55 billion, while adjusted earnings are projected at approximately $2.68 per share, slightly above last year's figure. This growth is expected to be fueled by increasing payment volumes and higher cross-border volumes, which are generally more profitable. However, investors will pay closer attention to Visa's outlook. The U.S. economy may encounter several challenges in the near future due to tariffs introduced by President Donald Trump on key trading partners. This situation could significantly affect Visa, considering that its operations are closely linked to consumer spending and international travel volumes. Visa currently holds a market capitalization of $580 billion. Over the past twelve months, revenue totaled $37 billion, and the company has been operationally profitable, reporting $24 billion in operating profits and a net income of $20 billion. Therefore, for those seeking potential gains with less volatility than specific stocks, the Trefis High Quality portfolio offers a viable alternative, having outperformed the S&P 500 and achieved returns exceeding 91% since its launch. See earnings reaction history of all stocks Here are some insights about one-day (1D) post-earnings returns: Additional information regarding observed 5-Day (5D) and 21-Day (21D) returns following earnings is summarized alongside the statistics in the table below. A relatively lower-risk strategy (though it may not be effective if the correlation is weak) involves understanding the correlation between short-term and medium-term returns after earnings, identifying pairs with the strongest correlation, and executing the appropriate trades. For example, if 1D and 5D demonstrate the highest correlation, a trader could take a "long" position for the next 5 days if the 1D post-earnings return is positive. Below is some correlation data based on a 5-year and a 3-year (more recent) history. Please note that the correlation 1D_5D refers to the correlation between 1D post-earnings returns and the following 5D returns. V Correlation Between 1D, 5D, and 21D Historical Returns Post-earnings stock reaction can sometimes be influenced by peer performance, and this pricing-in might even start before the announcement. Below is historical data comparing Visa stock's past post-earnings one-day (1D) returns with the 1D post-earnings returns of peers that reported just prior to Visa, allowing for a fair comparison. V Correlation With Peer Earnings Discover more about the Trefis RV strategy which has outpaced its all-cap stocks benchmark (a combination of the S&P 500, S&P mid-cap, and Russell 2000) to deliver strong returns for investors. Alternatively, if you desire potential growth with a steadier experience than an individual stock like Visa, explore the High Quality portfolio, which has outperformed the S&P and achieved returns exceeding 91% since its inception.
