DeepTech Explorers

1. Startups vs. Big Tech: Silicon Valley's ethos of innovation through disruption by startups is largely being replaced by a trend where startups are often acquired by big tech companies instead of growing to become independent challengers. 2. Co-opting Disruption: Big tech companies use their resources and influence to neutralize potential startup competitors by integrating them into their ecosystems, often redirecting their innovative potential towards reinforcing the status quo rather than disrupting it. 3. Strategies of Influence: These companies deploy strategies such as taking board seats in startups, providing them with funding, and granting access to critical data, all aimed at steering startups away from paths that could threaten their dominance. 4. Impact on Innovation: This trend is stifling meaningful innovation and competition, as startups focus on less ambitious projects that align with the interests of big tech, slowing the pace of technological advancement and economic growth. 5. Regulatory Response: The article suggests that current regulatory approaches might be insufficient to address these subtler forms of anti-competitive behavior and calls for a reevaluation of antitrust perspectives to foster a more competitive and innovative tech landscape. https://lnkd.in/dk7ypyNP

Visualization of only 20 of the infinite possible solutions to the three body problem. https://lnkd.in/gw5i22-T

🤖 Over the last couple of weeks I’ve been watching this series of videos on Physics Informed Machine Learning, and I can’t get enough of it! ✈️ Lots of ideas buzzing around my head now from learning dynamic stability modes of an aircraft by feeding SINDy* telemetry/FTI data to extending classical performance models using discrepancy models (and perhaps more importantly, why you might want to do any of this… 🤔). Huge thanks to Steven Brunton for this excellent series 👏🏼👏🏼👏🏼 https://lnkd.in/edwymZMc *SINDy: Sparse Identification of Non-Linear Dynamics

𝐃𝐫𝐢𝐥𝐥𝐢𝐧𝐠 𝐨𝐧 𝐭𝐡𝐞 𝐌𝐨𝐨𝐧'𝐬 𝐋𝐮𝐧𝐚𝐫 𝐒𝐨𝐮𝐭𝐡 𝐏𝐨𝐥𝐚𝐫 𝐑𝐞𝐠𝐢𝐨𝐧: 𝐏𝐑𝐎𝐒𝐏𝐄𝐂𝐓'𝐬 𝐐𝐮𝐞𝐬𝐭 𝐟𝐨𝐫 𝐋𝐮𝐧𝐚𝐫 𝐖𝐚𝐭𝐞𝐫 𝐚𝐧𝐝 𝐑𝐞𝐬𝐨𝐮𝐫𝐜𝐞𝐬   The European Space Agency - ESA is in the testing stage of an innovative payload package called 𝐏𝐑𝐎𝐒𝐏𝐄𝐂𝐓 (𝐏𝐚𝐜𝐤𝐚𝐠𝐞 𝐟𝐨𝐫 𝐑𝐞𝐬𝐨𝐮𝐫𝐜𝐞 𝐎𝐛𝐬𝐞𝐫𝐯𝐚𝐭𝐢𝐨𝐧 𝐚𝐧𝐝 𝐢𝐧-𝐒𝐢𝐭𝐮 𝐏𝐫𝐨𝐬𝐩𝐞𝐜𝐭𝐢𝐧𝐠 𝐟𝐨𝐫 𝐄𝐱𝐩𝐥𝐨𝐫𝐚𝐭𝐢𝐨𝐧, 𝐂𝐡𝐚𝐫𝐚𝐜𝐭𝐞𝐫𝐢𝐳𝐚𝐭𝐢𝐨𝐧 𝐚𝐧𝐝 𝐓𝐞𝐬𝐭𝐢𝐧𝐠). This comprehensive system aims to analyze volatiles in the lunar regolith and demonstrate In-Situ Resource Utilization (ISRU) techniques. The PROSPECT payload is planned to land on the Moon's south polar region in 2027.   Water is a key target of the mission: there may be concentrations of frozen water at or below the surface especially in the lunar polar regions – something suggested by recent measurements from orbit. Information on how much water is present and how accessible it is would help plan future missions using local resources.   PROSPECT consists of two main components:   1. 𝐏𝐫𝐨𝐒𝐄𝐄𝐃: A drill capable of extracting samples from depths up to 1 meter below the lunar surface. It includes integrated temperature sensors and a permittivity sensor to measure soil properties along the borehole. Leonardo, the overall lead, developed the ProSEED drill.   2. 𝐏𝐫𝐨𝐒𝐏𝐀: The miniaturised laboratory, ProSPA, will receive the samples from the drill via a carousel with multiple ovens, seal them, and heat them to extract the cold-trapped volatiles. The ProSPA instrument will then measure the nature and abundance of lunar volatiles using the gases released from the sample. ProSPA will also test specific processes which could be applied for resource extraction in the future.  ProSpa was developed by The Open University with support from RAL Space.   The system incorporates imaging capabilities, with cameras on both ProSEED and ProSPA providing visual context and sample characterization.   Originally developed for Russia's Luna-27 mission, PROSPECT has been adapted for accommodation on different lander platforms. It is now slated to fly on a Intuitive Machines lander under the NASA Commercial Lunar Payload Services (CLPS) mission to the lunar south polar region in 2027.   PROSPECT represents a significant step forward in lunar science and exploration capabilities. By enabling detailed in-situ analysis of lunar samples and demonstrating ISRU techniques, it will provide crucial data to support future long-term lunar missions and commercial activities on the Moon.   Image Source: Open University – ProSEED and ProSPA components of PROSPECT   #LunarExploration #ISRU #LunarResources #MoonWater #LunarDrill #LunarLab

What a day for the quantum industry. 🌟 Google Quantum AI has just proven, beyond any reasonable doubt, that Fault-Tolerant Quantum Computing is possible. For the first time ever, a logical qubit meets all criteria, demonstrating that Quantum Error Correction is effective. 🚀 There are three key standards for a logical qubit to serve as a viable component in a useful quantum computer: ✅ Below Threshold: Integrating more physical qubits within a logical qubit significantly reduces errors. The more, the merrier! ✅ Breakeven: The performance of the logical qubit surpasses that of the best physical qubit involved. ✅ Repeatable and in real time: error correction cycles are executed in sequence on the same chip using the same qubits while errors are tracked as they occur. All benchmarks have been impressively achieved, marking a global first. 🌍 This advancement highlights the rapid pace of quantum computing; just two years ago, Google's researchers had a system that was barely at the threshold and had not yet achieved breakeven. It also underscores the leadership of superconducting circuits in Quantum Error Correction. 🏆 To the folks at Google Quantum AI: Keep up with the fantastic work! Truly inspiring for us all! Just don’t rest on your laurels, cats can sprint fast! 😏 Read the paper (which is beautifully written, as always!): https://lnkd.in/eGBk-3XC #LogicalQubit #QuantumComputing #FaultTolerantQC #FTQC #GoogleQuantumAI

My brain doesn't want to believe this is real. China is stepping things up. Enter Unitree's G1: the $16,000 robot that's mass production ready. Here's the breakdown: • 1.32m tall, 35kg weight • 23 degrees of freedom for human-like movement • 2m/s walking speed (it can keep up with you) • 3D LiDAR and depth camera • Voice command capability • 2-hour battery life The G1 robot can also handle a payload of 10kg. I can’t wait for a pair of these to rock up when I next move house. My take: We had the browser in 1994, the smartphone in 2008, and generative AI in 2022. AI is now getting a body. I wonder what this price point will be when Figure 02 or Tesla's Optimus walk into the home. Competition means better products for us as the consumer. P.S. Follow me Alex Banks for daily AI highlights & insights, hit the 🔔 to never miss a post.

This short video is incredible.... What's next in AI in the coming 2-3 years? Eric Schmidt has a mind-blowing answer. Watch until the end 👇 #genai #llm #ai

2.5 million short in just 10 years. In the next decade, the US and Europe combined will be short of more than 2.5 million construction workers. That is why robotics is one of the most exciting frontiers for construction and traditional industries in general. When Owen and Martin asked me to host a robotics podcast for Bricks & Bytes it took me nanosecond to jump onboard. I'll be interviewing the brightest minds and the most successful entrepreneurs building robotics for the built world, starting with Troy Demmer at Gecko Robotics. BitBuilders is out on all major platforms on August 21st! Enjoy the trailer!

𝗠𝗶𝗱𝗷𝗼𝘂𝗿𝗻𝗲𝘆 𝗶𝘀 𝗶𝗻 𝗛𝗨𝗚𝗘 𝘁𝗿𝗼𝘂𝗯𝗹𝗲!🚨 German AI lab Black Forest Labs just dropped their new 𝗼𝗽𝗲𝗻-𝘀𝗼𝘂𝗿𝗰𝗲 (!!!) AI model FLUX.1 and it completely outperforms Midjourney. All you need to know: ⬇️ → The German-based company was founded by the original team that launched Stable Diffusion and recently left Stability AI (e.g. Robin Rombach, Andreas Blattmann, and Dominik Lorenz). → The company's first product features their text-to-image FLUX.1 models that are built on what it calls a "hybrid architecture," combining transformers and diffusion techniques. The model is available in 3 variants: 1️⃣ Flux 1 Pro: [API Access](https://docs.bfl.ml/)   2️⃣ Flux 1 Dev: Open weight, but non-commercial license   3️⃣ Flux 1 Schnell: Efficient, 4-step diffusion model with an open-source Apache 2 license → The quality looks amazing (see pictures below) and FLUX.1 seems very strong at generating human hands, which was a weak spot in earlier image-synthesis models like Stable Diffusion 1.5 due to a lack of training images focused on hands. → Black Forest Labs also announced with their launch a $31 million Series Seed funding round led by Andreessen Horowitz. Along with that the company brought on high-profile advisers, including entertainment executive and former Disney President Michael Ovitz and AI researcher Matthias Bethge. → Though text-to-image generation is Black Forest’s current focus, the company plans to expand into video generation next (competing with OpenAI’s Sora, Runway’s Gen-3 Alpha, and Kuaishou’s Kling). In my view this is crazy to see that they were stealth-building this in just a few months and already outperforming Midjourney, DALL-E 3, etc. It’s unbelievable how fast things are moving. Must be a huge team effort to build up clusters, infrastructure and datasets to launch so fast. In addition, another contribution towards the open-source AI ecosystem. After Stability imploded, open-source AI has been missing a good image-gen company. Kudos to the Black Forest Labs team! As a German/European, I’m especially happy to see that we get this innovation here! Excellent work, guys. 🇩🇪 🇪🇺 🏎️ 👉You can have a play with it here: https://lnkd.in/eYuPV6hg 👉 You can find the weights & inference code here: https://lnkd.in/d_bPiYq7 #midjourney #llm #genai #flux

Forget RAG, welcome Agentic RAG