AI in the Wild: Industry Emergence

Lecture 3: AI in the Wild and the Emergence of the Industry

• The focus of this lecture is on on how AI is used to generate and capture value through strategy and business models
• The material also discusses the impact of AI on employment, work tasks, and recent use cases
• The initial exploration is how AI is taking root outside of computer science research labs: how it is produced and adopted

AI in the Wild

• AI development involves compute resources like cloud services or supercomputers using GPUs and TPUs
• Skilled research scientists and engineers work on large-scale systems as part of AI expertise
• Data creation, collection, and preparation are crucial, with pre-training emphasizing quantity and fine-tuning emphasizing quality
• Real-time data feeds into AI inference
• Foundation models are pre-trained with large, general models for broad applicability
• Fine-tuned models undergo additional training for particular use cases or behaviours
• AI deployment can be in-house, closed source, or open source with model hubs for sharing
• User-facing apps deploy AI models for user interaction and incorporate user feedback data
• Plug-ins offer tools and data, often from third parties, for downstream FM services, utilizing relevant data for context in inference

AI Value Chain

• Players in the AI value chain differ in levels of vertical integration
• Google distinguishes itself as the most vertically integrated player, from TPU accelerators to Gemini
• The AI industry supply chain, as noted by Casilli and Tubaro in 2021, involves an "AI last mile" and heteromation

Growing Role of Industry in AI Development

• Industries are increasingly involved in AI development
• There's a rise in private sector involvement in AI research and development
• There's a shift towards private capture in AI-related policies
• There is a growing trend in the percentage of AI PhDs being hired by industry.
• There is a growing trend in the AI research faculty hired by industry
• Computing power usage in image classification models shows a shift from academia to industry

AI Market Dynamics

• According to the AI Index 2024, AI job postings as a percentage of all job postings vary by sector in the United States
• The information sector leads with 4.63% (but decreased by 3.67% from 2022)
• Professional, scientific, and technical services at 3.33% (decreased by 9.52% from 2022)
• Finance and insurance at 2.94% (decreased by 11.71% from 2022)
• Manufacturing at 2.48% (decreased by 18.81% from 2022)
• Total investment in AI has declined but is pushed up by GenAI
• Mergers and acquisitions signal industry consolidation
• The US is dominating the landscape of investments and startup numbers in AI, followed by the EU, China, and the UK
• Focus is on infrastructure and NLP

AI Use

• As per Rammer et al. (2021), the use of AI in German firms is limited and primarily concentrated in large firms
• Adoption tends to be at the final user stage
• It is essential to view AI as a societal backbone rather than as solely a productive backbone
• The use of AI differs among the various sectors

Public Investments in AI

• In the US, the growth in AI contract spending is mostly driven by the Pentagon and NASA
• Strategic asset investments are next to industry consolidation
• The rise of an "AI arms" race reinforces this trend

Typology of AI Firms

• AI giants have end-to-end integration across the AI stacks and can produce AI for internal and external use
• AI creators can produce or customize AI and rely on AI giants
• AI-powered operators use AI in their day-to-day operations and leverage both AI giants' services and internal capabilities
• AI takers leverage only off-the-shelf solutions to enable critical business functions

Summary

• AI is an industry and requires addressing structure-related questions
• Increasing specialization in roles is taking place in a world-spanning supply chain
• Vertically integrated actors aim at market tipping and drive M&A waves
• Adoption is limited upstream but widespread for final users/consumers
• AI is a strategic asset for national interests and advances are driven by both companies and government procurement