【會員投票 🗳️】 Deepseek
Added 2025-01-30 17:01:01 +0000 UTC你認為 Deepseek 能夠構成中國對美國人工智能企業的有效挑戰？
Comments
No because Deepseek won't have the means to scale up to millions of GPUs like the big techs planning to do. Here is a good analytic report to understand Deepseek better: https://semianalysis.com/2025/01/31/deepseek-debates/
Tony
2025-01-31 21:09:43 +0000 UTCThe money is in the use cases, not foundational model. Hence a company of their size doing foundational model research is uncommon, particularly in China. Kudos to them. Monetisation in the end requires platform and data, another name for BigTech. On a technical level the industry has already been moving in this direction — to be more than a very good autocomplete — it still requires good engineering to put everything together in a way that works. Strategically all DeepSeek has done here is announce that they belong and are now in the game. The only real moat is continuous innovation. This move will help them attract more talent and funding. This is just the first move in a very long game. At the moment they are not trying to upset the big fellas, just putting a foot in the door. These guys are patient, capable and ambitious, they will be around. When a country has more IP of their own, they will have more respect for IP.
lyk
2025-01-31 17:29:06 +0000 UTC挑戰在於DeepSeek提供了一個嶄新方法蒸餾思考方法，讓美國為主導的AI模型重回演算法研究更新更好更省電的路徑。 原本DeepSeek挑戰不會成功，偏偏西方世界太多蠢人，視資安如無物，有些甚至是政府提出資安風險警告，還主動提供更多個資給中國⋯⋯ 免費往往就是最貴，切記。
Tai-bo Cheung 張大寶
2025-01-30 20:17:16 +0000 UTCAnyway i like your name
Gary Lee
2025-01-30 18:14:17 +0000 UTCAI唔係咁絕對嘅嘢嚟嘅。簡單啲解釋Deepseek 係優化咗既有嘅AI模型。就唔算係啲咩新嘢嚟嘅。當然優化咗之後個成本就梗係低咗㗎啦 。但呢要明白AI之所以有價值呢就係佢嗰個思考能力同埋解決問題嘅能力，即係佢嘅答案有幾準呢？就好似讀書考試咁攞90分就唔難嘅，但90分之後嘅每1分呢都非常之難而家AI嘅發展呢暫時就喺呢度。投入嘅資源就一定係大㗎啦，未優化到咪唯有用硬件更加強個計算能力囉。但其實一般平時用係唔係已經可以唔追求咁高個精確？開始優化呢？ 明顯呢真係要睇吓你要嚟做乜嘢睇吓。好簡單我問你答幫我起下稿嘅呢啲梗係得啦。不過如果你話係軍事部署生產複雜科技搞埋啲生物學啊 物料學啊 嗰啲複雜嘢似乎就未得啦
Good Year
2025-01-30 17:41:15 +0000 UTCIt doesn’t matter. If it isn’t created by Deepseek in China, some other mad man in Silicon Valley or in India will have done it. Now that is open-source, the trajectory of development is there, who is the developer of a certain stage would be trivial. The very crux of this is the fact that AI is proven to be low cost and accessible- to a much greater extend than we would have thought. Originally we would have thought it remained something expensive whereby the cost is slowly cutting down, like airplane tickets and electric cars, but now it reduces exponentially. The implications of such a cost reduction means that wide range job cuts of all sectors are imminent. I’m not suggesting that AI can perform better than humans (in some aspects they already are). But imagine this - the unemployment rate will be 20% if the AI can assist in an increase in productivity of a labour by 25% and thus substituting the work for 20% of the people. Are the governments prepared? If they are, they should be starting to slash income tax rate, as you could not have tax a factor of production that will soon not to be applied. Are we prepared? Are those people who defined their own value with their jobs prepared?
simonishandsome
2025-01-30 17:12:12 +0000 UTC
