Sixty years of Moore’s Law gave us the computational horsepower to process exaflops of data. Four decades of the internet, accelerated by COVID, provided trillions of tokens worth of training data. Two decades of mobile and cloud computing put a supercomputer in the palm of everyone’s hand. In other words, decades of technological progress have accumulated to create the necessary conditions for generative AI to take off .
The rise of ChatGPT was the spark that lit the fuse, unleashing a density and enthusiasm of innovation we haven’t seen in years — possibly since the early days of the internet. The breath-taking excitement was especially intense in “Cerebral Valley,” where AI researchers became rock stars and hacker houses were packed every weekend with new autonomous south korea mobile database agents and companion chatbots. AI researchers went from literally “hackers in garages” to special forces commanding billions of dollars of computing power. arXiv’s print output was so high that researchers jokingly called for a moratorium on new publications so they could catch up.
But soon, excitement about AI turned to near hysteria. Suddenly, every company was an “AI co-pilot.” Our inboxes were flooded with indiscriminate pitches like “AI Salesforce,” “AI Adobe,” and “AI Instagram.” $100 million in seed funding returned. We found ourselves in an unsustainable frenzy of fundraising, talent wars, and GPU purchases.