• 72987930 darkthunder 樓主 : 首席軟件工程師在此，AI是一個騙局，它不過是一個經過美化的文本自動補全工具——雖然看起來感覺挺酷——但只有在提供足夠的數據和上下文時才有效，而生成這些數據和上下文需要花費數百萬美元。目前有大量資金投入到AI基礎設施中，但AI的盈利能力還未能達到相應的水平（儘管目前盈利情況還不錯，但其與思科在互聯網泡沫時期的情況驚人地相似，並且正朝着類似的崩盤方向發展）。

  舉報

  回復
  取消回復

• 72987930 Goblinn : Yea, bunch of old monkeys in office think AI is the way because salesmen, not engineers, make up 90% of american lobbies. Would be nice if AI was actual AI but for now it's just a highly resource intensive text/pixel guesser. The actual "AI" that benefits and helps people  (and is significantly more profitable) would be predictive AI (ie. breast cancer prediction AI that is out right now) because it actually uses known patterns and concepts to come to accurate predictions rather than try to invent things from scratch (which it CANNOT do, at all) because generative AI is similar to a car salesman trying to invent a new car based on his knowledge of cars he sold rather than being an actual car designer

  舉報

  回復
  取消回復

• 72987930 Goblinn : Im saying generative AI doesn't know anything and isn't helpful because it sucks, the only reason companies (in America particularly) are going all in is because the current cost of advertisement/marketing makes up about 1/3rd of all company expenditure for most retail/entertainment companies, imagine if you can somehow cut down that cost by 80%, that's basically the selling point, that got expanded to "AI can create anything for you at 1/4th the cost", which companies are slowly starting to realize is untrue because of pricey usage and data training contracts for half assed results, currently only proven ok with Palantir (surveillance state) and Google (global data farming and predicting/controlling)

  舉報

  回復
  取消回復

• 72987930 darkthunder 樓主 : PHP? I work with Rust/PostgreSQL/GraphQL/Typescript, lot of LLM algorithms are written in C++/Python, Vector databases are insane technology but the sad truth is the upper limit of "AI" currently is LITERALLY just text/pixel guessing, that's the entire concept of generative AI, researchers at Google will tell you the same thing because they have yet achieved actual "independence", we've JUST got to the "reasoning" ability after BILLIONS of dollars and 8 years since the emergence of transformers, so it's close (say give or take 3-5 years), but currently not enough to match the P/E. You can live in delusion all you want because you want to make money off the stock going up, but that's just the truth, also most stocks in the AI sector have hit their upper band ceiling for a while, so they were due for a selloff, AVGO's dogshit guidance was the spark

  舉報

  回復
  取消回復

• 72987930 Goblinn : That's one of MANY chinese AI btw, Qwen (from Alibaba) is reportedly being used at Meta, they're smuggling chips because the US opened trade, China still in talks to ban NVIDIA chips anyway (they dont give a shit), and it did reduce the cost of training and was a more highly efficient model, the point of capitalism is competition to drive innovation, NVIDIA started pushing for efficiency and cosy from that but as a downside the consumers are being punished, now a couple months down the line they're no longer including VRAM on chips and instead relying down the supply chain to supply it while Micron for example completely exists the RAM industry to focus on AI but no clear profits yet from either of those decisions. At the end of the day shareholders and traders need a good profit margin to justify P/E, profit margins not good enough? Sell. Profits good enough but no clear future? Sell. It's how you maximize gains

  舉報

  回復
  取消回復

查看更多評論...

• 72987930 樓主 74858079 : 英偉達尚未證明有顯著利潤來匹配其目前在自身供應鏈和客戶中進行的萬億美元循環融資，人工智能/芯片領域沒有重大突破，而且Rub ins產品尚未銷售，而中國正在迎頭趕上，看起來這將是一個沒有明確前進動力的糟糕季度。

  舉報

  回復
  取消回復

• 72987930 樓主 72959495 : 來源？

  舉報

  回復
  取消回復

• 72987930 樓主 : 你是不是用chatGPT寫的這個，哈哈，你在這句話裏用conflate這個詞不對，如果你確實是想這麼用的話，那就有矛盾了，因爲我根本沒有把長期估值和短期盈利匹配起來。英偉達在每次的業績中「勉強」表現得更好一點，這反而導致大規模拋售然後再回升，因爲每次都只是「超出」了盈利預期，而「尚可」的指引——考慮到他們在基礎設施上投入的大量現金，這根本不夠好。循環融資並不是鬆散定義的，他們確實與自己的供應鏈和客戶簽署了長期的、數萬億美元的合同協議，以此來「推動增長」，這沒問題，但利潤從何而來？每份未解答的業績，以及成本/效率方面的「突破」——英偉達正在爲基礎設施投入大量現金卻沒有保證利潤，我確定只要他們能保持領先（根據黃仁勳自己所說，他們正在逐漸失去優勢），未來肯定會賺更多錢，但這只是如果他們能夠保持領先的話，他們的市盈率仍然遠高於其收入。

  舉報

  回復
  取消回復

• 72987930 樓主 74858079 : NVIDIA is cutting off VRAM from its chips due to memory crunch from the AI boom, NVIDIA continues to market AI as "trillion dollar profitability" while not making a trillion dollars but having a 5 million dollar market cap instead of letting the technology emerge and sell itself as it comes (which is guaranteed profit anyway), which is exactly what happened with the dot com bubble, which means there is no realized value yet, you can think of it as unrealized/realized gains and losses, the pretty number looks better but until that's raw cash in your bank account then it doesnt matter and is always subjected to volatility

  舉報

  回復
  取消回復

• 72987930 : 最大反彈幅度可能在1%-1.5%之間，但短期內Advanced Money Destroyer的看跌動能較強，支撐位在200。如果跌破200，務必拋售或持有，因爲此前的支撐位是170。

  舉報

  回復
  取消回復

• 72987930 樓主 酷冷至尊雷霆650W : 在177重新買入

  舉報

  回復
  取消回復