AI at 70 Part 2 14 more lessons from a lifetime of boom and bust

AI at 70 (Part 2): 14 more lessons from a lifetime of boom and bust Context is all you need September 2026

Adrian Cox Deutsche Bank Research Institute IMPORTANT RESEARCH DISCLOSURES LOCATED IN APPENDIX 1.

Adrian Cox | Would the real AI bubble please stand up? | December, 2025

Artificial intelligence marks 70 years since Dartmouth Summer Research Project “Context is all you need”: clues from the past for what comes next

The term “artificial intelligence” almost didn’t exist. Dartmouth College 14 more lessons from a lifetime of boom and bust Professor John McCarthy invented it to attract scientists to come to his landmark summer workshop 70 years ago after his first attempt, “automata 1. This boom is remarkable not just 8. Tech IPOs have tended to studies”, failed to excite much interest. for scale but also speed outperform other IPOs 2. AI makes hard things look easy, 9. Jobs come and go but working This is the second of our two reports looking at what the past 70 years of and easy things look hard hours have only fallen stunning breakthroughs and false dawns since then can tell us about what 3. AI models are commoditising fast 10.It's a tough job market for recent to expect from the current technology and investment boom. 4. Benchmarks are ageing even graduates faster than models 11.Concerns over data sovereignty The first, “AI at 70: 14 lessons from a lifetime of boom and bust”, reviewed 5. Tech use accelerates when the date back long before AI themes including non-linear growth to infrastructure constraints and cost falls far enough 12.AI arrived as trust had been valuation booms and busts. 6. China has left the US behind in declining for decades electricity generation 13.New technology brings new risks This one builds on those themes and gives them a spin forward to the 7. Mega IPOs will eclipse entire years 14.Tech may have a role in declining future, including the speed of investment compared with past booms, a of share sales educational performance snapshot of AI’s strengths and weaknesses, a brief history of previous tech IPOs, and an overview of the prospects for jobs, trust and risks.

The stakes look higher than ever.

The leading companies have released even more capable models in the past week; there has been a series of security breaches where AI agents appeared to collaborate to evade detection; and a senior government official who joined OpenAI warned yesterday that “most people could die” if superintelligence is built without more robust safeguards.

2 Adrian Cox | AI at 70 (Part 2): 14 more lessons from a lifetime of boom and bust | September 2026