5 Contrarian Theses On Where AI Is Going
A 10-year AI newsletter writer argues the consensus on AI's future is dangerously wrong.
A veteran AI industry analyst with a decade-long newsletter, Rob from Investing in AI, has published a provocative set of five theses arguing that the current consensus on AI's trajectory is fundamentally mistaken. The piece, titled "The Great AI Contraction: 5 Contrarian Theses," challenges optimistic narratives around agentic AI and investor returns, instead predicting a period of consolidation and shifting value.
The first thesis warns of an impending "trust recession" caused by the unpredictable actions of AI agents, systems that can autonomously take actions like making purchases or sending emails. This erosion of trust could slow adoption. Secondly, the analysis suggests physical assets (like data centers and chip factories) will see their valuations outpace those of pure AI software assets, flipping the current investment script.
The third thesis posits that AI will lead to a re-bundling of software, reversing the trend of best-in-class point solutions as AI assistants consolidate workflows. Fourth, practical inference economics—the cost and speed of running models like GPT-4 or Llama 3—will become more important than raw benchmark performance for commercial success. Finally, the author contends that most AI-driven efficiency gains will be competed away, with the value accruing primarily to consumers through lower prices and better services, not to company investors.
- Predicts a 'trust recession' from unpredictable autonomous AI agents, slowing adoption.
- Argues inference costs for models like GPT-4 will trump benchmark scores for commercial success.
- Contends most AI value will go to consumers, not investors, in a 'Great AI Contraction'.
Why It Matters
Challenges trillion-dollar investment theses, suggesting a shift from software to infrastructure value and consumer benefits.