prediction
The future of AI lies not only in pushing more billions of parameters into hot GPUs, but in a smart combination of classic transformers and LLMs with event-driven neuromorphic architectures.
A prediction by מייק. Statements are as reported by the cited publishers — not verified by DecisionGateway.
העתיד של ה-AI לא נמצא רק בלדחוף עוד מיליארדי פרמטרים לתוך כרטיסי מסך לוהטים, אלא בשילוב חכם של טרנספורמרים ו-LLMs קלאסיים עם ארכיטקטורות נוירומורפיות מונחות-אירועים
Discussion
Related claims
On neuromorphic chips like Intel's Loihi 2, which physically co-locate memory and compute, LLMs of 370 million parameters run without any matrix multiplications at all, and provide throughput 3.5 times higher relative to energy consumption compared to Edge GPUs.author · statementIn transformer-based language models and architectures like Mamba, every token prediction requires activating billions of parameters in parallel through the attention mechanism.author · statementAI is moving beyond individual apps toward infrastructure, compute and real business use cases.Prompt Talks Ep. 2 · predictionInstitutions are optimistic about the application opportunities that will come from AI moving from model capability to scenario implementation next year.机构 · predictionAnthropic stated that its internal estimates show the pace of AI capability progress may accelerate further.Anthropic · predictionIn a deep and very enjoyable review, Volker Kempa presents the alternative gaining engineering momentum: Neuromorphic AI based on Spiking Neural Networks (SNNs).author · statement
More claims by מייק
In event-driven neuromorphic architectures, parts of the network that do not need to work simply stay at rest.statementIn event-driven neuromorphic architectures, memory is adjacent to the processor.statementIn event-driven neuromorphic architectures, time is a computational feature.statementEBMs promised a revolution 20 years ago, and the problem was never elegance — it was training at scale.statement'The end of diffusion' is hype; the better question is whether it is worth paying for more expensive training to get built-in verification and flexible compute at inference time — a real and interesting trade-off.statement
All claims by מייק