1 December: Associate Professor, Department of Computer Science, Stanford University, recentlyAndrew NgOn the personal social media account, he published his opinion on AI The idea of foam as a hot topic。
Now, with OpenAI’s 1.4 trillion-dollar capital plan and the short-lived 5 trillion-dollar mark on Britain’s market value, more and more voices are beginning to wonder whether speculation and speculation have pushed AI’s investment to unsustainable levels
IN RESPONSE, U NDA GAVE HER OPINION THAT AI IS NOT A PIECE OF IRON, BUT THAT THE LEVEL OF FOAM VARIES FROM ONE FIELD TO ANOTHER. IT SETS OUT THE FOLLOWING POINTS:
AI APPLICATION LAYER: UNDERINVESTMENT. ITS POTENTIAL IS FAR GREATER THAN MOST PEOPLE KNOW. U NDA PREDICTS THAT MANY NEW AI-BASED APPLICATIONS WILL NOT BE COMPLETED IN THE NEXT DECADE. "IT IS ALMOST POSSIBLE TO ASSERT THAT AI-BASED INFRASTRUCTURE/TECHNOLOGY-BASED APPLICATIONS (E.G. LLM API) MUST BE MORE VALUABLE THAN INFRASTRUCTURE. I'M SORRY
AI Logic Infrastructure: Substantial investment is still needed. U Nda stated that, at this stage, despite the current low penetration rate of AI, infrastructure providers have difficulty meeting the need for processing capacity to generate token。
AI MODEL TRAINING INFRASTRUCTURE: PRUDENTIAL OPTIMISM, BUT FOAMS MAY EXIST. U N’DALTAN SAID THAT IF PART OF THE AI TECHNOLOGY WAREHOUSE (PERHAPS IN THE TRAINING INFRASTRUCTURE) COLLAPSES AS A RESULT OF OVERINVESTMENT, IT COULD LEAD TO A BROADER NEGATIVE MARKET SENTIMENT TOWARDS AI AND AN IRRATIONAL OUTFLOW OF RATIONAL INTEREST IN INVESTING AI, DESPITE THE STRONG FUNDAMENTALS OF THE FIELD AS A WHOLE。