Smart city deployment needs a powerful AI accelerator to performance, total cost of ownership and bottlenecks. Read our case study to see how our #tsunAImi acceleration card + @NeuReality_AI
's NR1 provides power for deployment with low impact. https://
tinyurl.com/2uyx8uyh
AI HARDWARE
-
AI Accelerator Card Powers Smart City Deployment Efficiently
By
–
-
AI Computing Power Grows Faster Than Moore’s Law
By
–
50 years of Moore’s Law has just happened in 10 years in AI. ai training compute for the largest runs has increased 10x every year for 10 yrs. it took Moore’s law ~50 years to see 10 orders of magnitude increase in transistor density. and this is just the beginning
-
Drones and AI Innovation in Ukraine War
By
–
This is a good (not paywalled) summary of the current state of drones in Ukraine. Thanks to @scienceisstrat1 – which btw is an excellent account if you like great stats / trends in science and tech. https://
washingtonpost.com/world/2023/07/
26/drones-ai-ukraine-war-innovation/
… -
US China Accelerate Military AI Integration Drone Swarms Race
By
–
The U.S. and China are accelerating research on how to integrate artificial intelligence into their militaries https://
wsj.com/world/china/ch
ina-u-s-test-intelligent-drone-swarms-in-race-for-military-ai-dominance-db361265?st=gcjjawvl29u5onz
… via @WSJ -
Real-time GPU rendering and diffusion models performance evolution
By
–
By 2018 or so, on a dedicated GPU, I could get 10-20 fps https://
vimeo.com/382088640#t=58s Current SOTA are all diffusion models, which are multi-step, hence a bit slower again. I bet there will be real-time variants in a year or two. -
Drones: The Asymmetric Power Reshaping Modern Warfare Economics
By
–
worth remembering how important drones are in Ukraine a drone costing $5-10k can take out a tank costing $5-10m… it might take ten or 20 or even 100 attempts but the cost advantage is still huge this is the asymmetric power of the coming wave
-
Design Principles for Lifelong Learning AI Accelerators
By
–
Uncovering Design Principles for Lifelong Learning AI Accelerators by Kudithipudi, Daram, Zyarah, Zohora, Aimone, Yanguas-Gil, Soures, Neftci, Lomonaco, Thiem, Epstein + @tenstorrent VP of AI Matthew Mattina. Read the full paper here –> https://
tenstorrent.com/research/uncov
ering-design-principles-for-lifelong-learning-ai-accelerators/
… -
Transformer Math 101: GPU Training Optimization Deep Dive
By
–
pod on Transformer Math 101! with @QuentinAnthon15 of @AiEleuther everything you ever wanted to know about the nitty gritty of training and making the gpu go brrrrrrrrrrrr
-
Open Challenges in Large Language Model Research Today
By
–
Open challenges in LLM research The first two challenges, hallucinations and context learning, are probably the most talked about today. I’m the most excited about 3 (multimodality), 5 (new architecture), and 6 (GPU alternatives). Number 5 and number 6, new architectures and
-
Xiaomi Cyberdog 2: Advanced Robotic Pet With AI Sensors
By
–
Xiaomi ha presentado Cyberdog 2, una nueva y ambiciosa versión de su mascota robótica.
— Juan Merodio (@juanmerodio) 16 août 2023
Equipado con los más avanzados sensores, como cámaras, sensores de fuerza, micrófonos y LiDAR y con una capacidad de procesamiento mejorada, la marca china pretende hacerse con el mercado… pic.twitter.com/2wLdiiQprSXiaomi ha presentado Cyberdog 2, una nueva y ambiciosa versión de su mascota robótica. Equipado con los más avanzados sensores, como cámaras, sensores de fuerza, micrófonos y LiDAR y con una capacidad de procesamiento mejorada, la marca china pretende hacerse con el mercado