Coatue人工智慧(AI)長篇報告精要

https://www.coatue.com/blog/perspective/ai-the-coming-revolution-2023

# AI迭代特性與生產力提升將比前幾次產業創新進程快速

# 融資主力目前集中於模型層級

# 趨勢轉變：從建立更快的「計算器」到建立超級智慧的「大腦」。

# 人工智慧啟用新平台：智慧即服務 (IQaaS)。

# 研究人員的焦點：擴展人工智慧模型效能：更多參數和大型資料集是主流，而不是更多計算和更長的訓練。

# 資料稀缺是擴展的潛在障礙：高品質的文字資料可能很快就會耗盡，圖像和影片資料有較長期潛力。

# 解決人工智慧在"模型輸出、模式、可用性、企業部署"方面的挑戰將有潛力釋放巨大的機會。

Coatue long AI report take-aways:

• AI advancing much faster than previous technology waves with proving utility across domains.
• AI stack from bottom: Semi > Cloud > Ops > Models  APPs: Funding concentrated in "model layer" for now.
• Trend shift: From building faster “calculators to building super-intelligent “brains.
• AI enables a new platform: Intelligence-as-a-service (IQaaS).
• Researchers focus: Scaling AI model performance: more parameters and large datasets are mainstream vs more compute and longer training.
• Data scarcity is a potential wall to scaling : High-quality text data could be exhausted soon, images & video have longer runway.
• Solving AI’s challenges in Model output, Modalities, Availability, Enterprise deployment: has potential to unlock vast opportunities.