新兴技术未来报告：更新我们的人工智能展望（英）2024

Emerging Tech Future Report: Updating Our Generative AI Outlook

Introduction

In May 2023, the team published perspectives on how Generative AI (GenAI) was poised to impact various industries and technologies. This note revisits those perspectives, highlighting how GenAI has manifested itself and evolved since then. The rise of ChatGPT in early 2023 marked a pivotal moment, as AI became understood as an adaptable technology with broad application potential. Since then, investment in related technologies and startups has skyrocketed, driven by intense competition in the foundation model space.

Key Findings

1. Rise of GenAI:

   • Investment in GenAI-related technologies and startups has surged, with significant competition in the foundation model space.
   • The GenAI-infrastructure landscape is stratifying across several use cases, such as on-device inference, domain-specific knowledge, and raw computational power.

2. Blockers to Adoption:

   • Despite intensified efforts toward AI transformation, several blockers remain, including high compute costs, data availability, data security, and overall system complexity.
   • Application-level startups face a more challenging fundraising environment due to near-term pressures to demonstrate commercial viability.

3. Impact on Enterprise Applications and AI & ML:

   • GenAI has transformed the existing AI & ML vertical in limited but fundamental ways.
   • New LLMs represent the future of the industry, but legacy approaches continue to play a role.
   • GenAI software will contribute about 14% of AI software spending in 2024, growing to 32.3% by 2028 (IDC estimates).
   • Companies building legacy ML models have seen their estimated valuations plunge, particularly in the face of GenAI disruption.

4. Real-World Progress:

   • Enterprise rollouts of GenAI have progressed more slowly than initially forecast.
   • Infrastructure layers, including model architecture labs and semiconductor startups, have achieved 25 out of 39 unicorn valuations in the LLMOps space.
   • Semiconductor advancements, including those by Astera Labs, Cerebras, and Groq, have shown the need for a new software stack.
   • The AI agent space has become crowded but is expected to be disrupted by more action-oriented model capabilities.

5. Key Recent GenAI VC Exits and Talent Acquisitions:

   • Notable exits and acquisitions include Character.AI, Adept AI, Clickable, Argilla, Uizard, Miro, DarwinAI, and Inflection AI.
   • Major acquirers include Alphabet, Amazon, Hugging Face, NVIDIA, and Microsoft.

6. Crypto Impact:

   • The anticipated impacts of AI within the crypto space were seen as potentially transformative, especially with the rapid rise of GenAI.
   • AI is expected to significantly enhance smart contract development and audit processes.

Summary

GenAI has transformed the AI & ML landscape in limited but significant ways, with investments and competition in the foundation model space driving rapid advancements. However, several challenges remain, particularly around compute costs and data security. Enterprise applications and infrastructure layers are seeing substantial growth, but the adoption rate remains slower than expected. The AI agent space is crowded but expected to evolve further. Key startups and talent acquisitions highlight the ongoing interest and investment in GenAI.