Mingxing Tan - Waymo | LinkedIn (original) (raw)
Other similar profiles
Explore more posts
- Simon Lancaster ๐บ๐ธ๐จ๐ฆ๐ต๐น University of Waterloo โข 35K followers ๐ When I wrote that โthe LLM era is peaking, the SLM era is just getting started,โ it was a directional bet. Now, ๐ก๐ฉ๐๐๐๐ has doubled down with another paper: the future of AI isnโt about SLM vs LLM, itโs about the two working side by side. ๐ก Key takeaways: โข ๐๐๐ฏ๐ฟ๐ถ๐ฑ ๐ฏ๐ฒ๐ฎ๐๐ ๐บ๐ผ๐ป๐ผ๐น๐ถ๐๐ต๐ถ๐ฐ: Hybrid systems have the edge when it comes to cost efficiency, stability, and maintainability. โข ๐ฆ๐๐ ๐ ๐ฎ๐ฟ๐ฒ ๐ฑ๐ถ๐๐๐ถ๐น๐น๐ฒ๐ฑ ๐๐๐ ๐: We havenโt seen many natively designed SLMs yet. Once those appear, efficiency and reliability could jump another order of magnitude. โข ๐ฆ๐บ๐ฎ๐น๐น๐ฒ๐ฟ ๐บ๐ผ๐ฑ๐ฒ๐น๐ ๐ต๐ฎ๐น๐น๐๐ฐ๐ถ๐ป๐ฎ๐๐ฒ ๐น๐ฒ๐๐: Narrower scope means tighter alignment, perfect for structured agent tasks. โข ๐๐ฑ๐ผ๐ฝ๐๐ถ๐ผ๐ป ๐บ๐ถ๐ด๐ต๐ ๐ฏ๐ฒ ๐๐น๐ผ๐: Billions already sunk into LLM infra mean inertia is real. ๐ญ The insight for founders and operators: The next edge in AI wonโt come from running ever-bigger models in the cloud, it will come from designing modular systems that match model size to task complexity. Use the right tool for the right job. ๐ค So where do you see this goingโtoward a fusion of SLMs and LLMs, or just another round of monster-sized models? ๐ NVIDIAโs new report is here: https://lnkd.in/gvVWJ-uT
- Murugavel Ganesan SiliconAuto โข 4K followers While Waymo is busy perfecting the art of "Not Hitting a Trash Can in San Francisco," Indian startups are basically training their AI to survive an episode of Mad Max. Since they can't rely on the "polite rules" of Western driving, here is how local companies are technically tackling the "Final Boss" of roads: 1. Ditching the Maps (The "Mapless" Approach) In the US, Waymo relies on HD Maps (centimeter-accurate digital twins of the city). In India, if you rely on a map, you'll eventually drive into a newly dug trench or a spontaneous wedding tent. The Solution: Startups like Minus Zero and Swaayatt Robots (เคธเฅเคตเคพเคฏเคคเฅเคค เคฐเฅเคฌเฅเคเฅเคธ) use a "Mapless" philosophy. The Tech: Instead of following a pre-planned map, the AI uses Physics-Aware Vision. It looks at the road in real-time and calculates "traversable space" (where it can go) rather than "legal lanes" (where it should go). If thereโs a gap between a bus and a cart, and physics says the car fits, it goes. 2. Reinforcement Learning (The "Chaos Simulator") Western AI is often trained on "Orderly Data." Indian AI is trained on Stochastic Traffic Dynamics (fancy talk for "pure randomness"). The Startup: Swaayatt Robots (เคธเฅเคตเคพเคฏเคคเฅเคค เคฐเฅเคฌเฅเคเฅเคธ) (based in Bhopal) has been testing at high speeds in unstructured environments. The Logic: They use Deep Reinforcement Learning to train the car via millions of "games." The AI is rewarded for getting through a crowded intersection without stopping. It learns "adversarial" behaviorโhow to edge forward and "bully" its way through traffic, much like a local Rickshaw driver would. 3. Vision-First vs. Lidar-Heavy A single Lidar sensor can cost as much as a small car. In India, where a "minor scratch" is a daily ritual, putting $50,000 worth of spinning lasers on the roof is a financial nightmare. https://lnkd.in/gEaW7EyC The Solution: Minus Zero is building a Vision-Only stack (similar to Tesla, but for chaotic roads). Why? Cameras are cheap. By using "Nature-Inspired AI," they try to mimic how the human brain processes visual depth and motion without needing expensive laser pulses. Itโs about making the AI "smart" enough to understand that a moving pile of hay is actually a hidden tractor. 4. Narrowing the Scope (The "Industrial First" Strategy) Recognizing that public roads are a headache, many are starting in "Controlled Chaos." The Startup: Ati Motors and Flux Auto. The Strategy: Instead of a Robotaxi in Delhi, they are deploying autonomous trucks and tugs in mines, ports, and factory campuses. The Benefit: You still have unpredictable obstacles (forklifts, workers), but you don't have to worry about a 14-year-old on a dirt bike popping a wheelie in front of your sensors.
- Luis Oria Seidel Isapre Colmena โข 8K followers ๐ Tesla Robotaxi: The Battle for Autonomous Transportation The launch of Tesla's Robotaxi marks a milestone in autonomous mobility, directly competing with Uber, Lyft, and Waymo. A recent analysis reveals how prices and wait times define their strategies: Tesla bets on total autonomy to reduce costs in the long term, while rivals opt for hybrid models with human drivers for greater immediate scalability. ๐ Comparison of Prices and Waits - Tesla: Competitive initial prices at 0.30โ0.30-0.30โ0.40 per mile, with average waits of 5-10 minutes in urban areas, focusing on 100% autonomous fleets for efficiency. - Uber/Lyft: Rates of 1.00โ1.00-1.00โ1.50 per mile, waits of 2-5 minutes thanks to their existing network, but they depend on drivers to cover demand peaks. - Waymo: 0.70โ0.70-0.70โ1.20 per mile in limited zones, waits of 3-7 minutes, prioritizing safety with geofencing and massive AI data. This dynamic highlights Tesla's disruptive vision versus the pragmatic approach of the incumbents, promising a future where AI redefines urban transportation. For more information visit: https://enigmasecurity.cl #Tesla #Robotaxi #AutonomousVehicles #Waymo #Uber #AIinMobility #TechnologicalInnovation Support Enigma Security with a donation for more news: https://lnkd.in/er_qUAQh Connect with me on LinkedIn: https://lnkd.in/eXXHi_Rr ๐ Mon, 02 Mar 2026 17:47:16 +0000 ๐Subscribe to the Membership: https://lnkd.in/eh_rNRyt
Explore top content on LinkedIn
Find curated posts and insights for relevant topics all in one place.