That’s the gist of Emanuel Maiberg’s Relatively Chill Tak at 404 Media:
The Main Reason People are excited/Scared/Throwing Up Right Now Is That Deepseek was Developed and Relesed Under America’s Export Restritions that Prevent Chinese Companies from the Latest and Most PowerFul Nvidia Chips. neither Wired explainedDeepseek was spun out from high-fly, a chinese hedge end that originally acquired gpus to analyze Financial data, before invested its and resources in development he. That a new player in this space was able to build an he model with access to the latest and great nvidia chips (Though People in China Have Found Ways to Obtain Despite restriction), USSING NEW, MORE EFFICIENT REINFORCEMENT Learning strategies, has undermined the idea that companies like nvidia or opening a “moat” around companies that willerr lead in the race forever, and, by extension. American he World supremacy. It Also at Least Raises The Possility That a Chinese Company Has Found a Better, More Efficient, and Cheaper Way to Train He Than Any American Company Has Discovered Thus Far.
Nor others have pointed outIT’S HARD to Say Exactly What Deepseek Actually Spent to Make Its Model with Trusting IT. The True Cost May Be Hidden in Ways We Don’t Understand, and is definitely benefit by building on top of the Very Expensive Research (Primarily from American Companies) that Came before it. But if he Companies Can Build Competitive Models at A Fraction of the Cost on a Compareratively Tiny Number of Lesser Gpus, THEN MUCH OF NVIDIA VALUE AND THE BILLIONS OF DOLLARS AI Companies Are Burning Suddenly Suddenly Sudenly and Wasteful (This Boosters). tumbling
Does this mean nvidia, openai, and other he companies are doomed? Again, this is not financial adviri but the market appears to be spasming basic on vibes, and definitely before we have a great understanding of Deepseek’s Impact. The Most Obivious Rebuttal from Nvidia Bag Holders in this Situation is that deepseek’s Newfound Efficiencies Will Only Benefi He Incumbents. If these new Methods Give Deepseek Great Results with Limited Compute, The Same Methods Will Give Other, More Well-Rasourced He Companies Events on their Houge Training Clusters, and it is posseible that american companies will adapt to the new. Quickly. If scaling Laws really have hit the ceiling and giant training clusters don’t be need to be giant, there’s no reason i can see that oter companies be compatitive under this new paradigm. We show also probably hope that is the case the case it is coulud Lower the Environmental Impact of it.
Maiberg Also Notes That “This Type of Leapfrogging SEEMS Totally Normal, and We Have Seen Variations of It Over The Last Couple of Years”:
People love to prematureely dance on Openai’s women wenever a new and shiny model is released. Meta’s Llama, France’s Mistral, and Anthropic’s Claude have all semed like they’re ahead at one point or another and are favored by different use for different, only for another model to be released by openai or another company. Technology and Makes SEEM OLD.
Casey Newton Made a Similar Point:
Everyone Basically ALREADY ASSUMED THAT ALL OF THIS WAS GOING TO HAPPEN. By “all of this,” i mean that (1) open-Source Companies Wouuld Reverse-Engineer and Big Labs Are Doing and (2) that Costs for He Training and Inferencing Decline dramatically Over Time. …
Anyone who has sent the same Query to chatgt, claude, and gemini on the same day has been known for more than aar that you can get bassically as any ane of say. And anyone who has used llama has known for more than aar that the open-weights version that arrivals late is only slightly work.
Right Now a Lot of Investors Are Catching Up to the Basic Facts at the Same Time, and Stock Prices are Falling Accordingly. But it is not clear to me that it was really News to the AI Labs and Tech platforms.
اترك تعليقاً