2021-10-31-布鲁盖尔研究所-是什么阻碍了人工智能在欧洲的应用_（英）_19页_778kb

报告摘要

Artificial Intelligence (AI) is pivotal for future economic growth but adoption in Europe is lagging.
Current measurement of AI adoption is inaccurate due to inconsistent definitions and data collection.
The study focuses on the adoption stage of AI—integrating technologies into business operations.

AI advancement involves three distinct stages: research, development, and adoption.
Barriers differ by stage:
- Research: Lack of public funding and venture capital.
- Adoption: Less addressed by standard policies (e.g. tax deductions differ from research funding).
Non-response bias and varying taxonomies cause wide estimates (Eurostat 7% vs EC 42%).
Clusters by sector: Primary/secondary sectors use AI for operational tasks; tertiary uses NLP/rec engines.

Adoption follows Rogers’ Diffusion Model, showing slow uptake.
- Innovators (2.5%) quick to adopt; Laggards (16%) resist.
Country-level variations: Adoption rates significantly differ by nation (e.g., Malta 2%, Germany 19%).
Correlation with firm size: Larger EU firms adopt more; SMEs lag due to fewer resources and infrastructure.
Technology readiness: AI builds upon existing data-driven tools; start with incremental changes.

Technological readiness is key; hierarchical nature of AI needs stable data, computing, etc.
Organizational factors: Decentralized structures and top-level support drive innovation.
Environmental factors: Competition and financial availability amplify adoption potential.

Skills: 80% cite lack of internal labor skills.
Financial constraints: High technology & process adaptation costs; SMEs struggle with access.
Data & infrastructure: IT deficits hamper non-tech firms.
Regulatory uncertainty is a barrier for non-adopters (liability, reputational risk).

EU lags behind US/China in financing and translation of research talent into business skills.
Public vs private data used differently: External data for R&D, internal data for adoption.
GDPR may establish regulatory leadership; China and US data policies are less comprehensive.

Skills: Improve labor market matching via education, adult learning.
Financing: Adapt tax incentives or direct subsidies for SMEs. Increasing public R&D spending could go further.
Data: Promote open data initiatives; alleviate legal barriers in existing data usage.
Regulation: Create a regulatory sandbox to ease technological triability.
Metrics: Develop standardized definitions to track AI adoption.

展开完整摘要