There are a form of important questions relating to the affect of the efficiency features from the improved DeepSeek techniques.
Deepseek improved the Reinforcement learning.
Deepseek also immediately accessed the Nvidia chips. They did NOT express Nvidia CUDA. CUDA turned into combating them from doing what they valuable to discontinue with the chips.
The Deepseek analysis paper gave a listing of suggestions to Nvidia on how Nvidia needs to give a lift to CUDA.
The price of innovation is reducing.
I don’t fetch too too worthy to add on top of this earlier put up on V3 and I focus on it applies to R1 too (which is the more moderen, thinking the same).
I will train that Deep Learning has a legendary ravenous appetite for compute, love no tons of algorithm that has ever been developed… https://t.co/mX5kiQEJPX
– Andrej Karpathy (@karpathy)”https://twitter.com/karpathy/status/1883941452738355376?ref_src=twsrc%5Etfw” goal=”_blank” rel=”noopener”>January 27, 2025
This document is prolonged nonetheless very factual.
“With R1, DeepSeek basically cracked one among the holy grails of AI: getting models to goal step-by-step without counting on huge supervised datasets. Their DeepSeek-R1-Zero experiment confirmed one thing outstanding: the express of pure reinforcement…
– Chamath Palihapitiya (@chamath)”https://twitter.com/chamath/status/1883579259769462819?ref_src=twsrc%5Etfw” goal=”_blank” rel=”noopener”>January 26, 2025
A few significant questions/comments merely about my mind as I read more about DeepSeek. Itemizing them right here:
1) Let’s give 1% likelihood to your total conspiracy theories upfront so we can address it and stir on. Whether it is a long way attainable for China/Chinese language companies to make express of shell companies in… https://t.co/x7O28jc4gW
– Chamath Palihapitiya (@chamath)”https://twitter.com/chamath/status/1883571504153309377?ref_src=twsrc%5Etfw” goal=”_blank” rel=”noopener”>January 26, 2025
An obvious, “we are so motivate” 2nd in the AI circle in a system grew to alter into into “it’s so over” in mainstream.
> unbelievable shortsightedness
> the vitality of o1 in the palm of each coder’s hand to ogle, explore, and iterate upon
> tips compound
> the tempo of compounding hastens…– Jim Fan (@drjimfan)”https://twitter.com/DrJimFan/status/1883880955943002484?ref_src=twsrc%5Etfw” goal=”_blank” rel=”noopener”>January 27, 2025
Whether or no longer you’re eager on it or no longer, the model forward for AI is rarely any longer going to be canned genies managed by a “safety panel”. The manner forward for AI is democratization. Every web rando will whisk no longer refined o1, nonetheless o8, o9 on their toaster pc. Or no longer it is a long way the tide of history that we should always surf on, no longer swim… pic.twitter.com/VPmgV8wR1C
– Jim Fan (@drjimfan)”https://twitter.com/DrJimFan/status/1882799254957388010?ref_src=twsrc%5Etfw” goal=”_blank” rel=”noopener”>January 24, 2025
Brian Wang is a Futurist Belief Leader and a most neatly-liked Science blogger with 1 million readers per month. His blog Nextbigfuture.com is ranked #1 Science Files Weblog. It covers many disruptive technology and trends along with Station, Robotics, Man made Intelligence, Medication, Anti-rising older Biotechnology, and Nanotechnology.
Known for figuring out cutting edge technologies, he is at the 2nd a Co-Founding father of a startup and fundraiser for prime possible early-stage companies. He’s the Head of Compare for Allocations for deep technology investments and an Angel Investor at Station Angels.
A frequent speaker at companies, he has been a TEDx speaker, a Singularity College speaker and guest at a form of interviews for radio and podcasts. He’s birth to public speaking and advising engagements.
