How today's policy will shape computing for the next 100 years

The explosion in artificial intelligence (AI) over the last decade has led to headline-worthy feats across various fields, such as computers beating humans in the game of Go, enabling autonomous vehicles, solving protein folding, and promising brain-like intelligence. Underlying these advances are two key factors: scaling graphics processing units (GPUs) and increased budgets. The two main drivers of computing sucess - increased comuting budgets and manufacturing investments - are running against prohibitive barriers. On the one hand, the end of Moore's law, which essentially stonewalls manufacturing at advanced process nodes, necessitates different forms of post-digital computing. This situation has led to a flurry of exploratory ideas (e.g., quantum computing, neuromorphic computing, analog computing, etc.), which are, at best, modestly mature, and urgently need prioritization at the policy level. On the other hand, there is an exponential and insatiable demand for compute, fueled by the AI revolution. How do we both meet the societal demands and also pave the path for the next type of computer? The answer lies in both understanding the options in front of us and also in udnerstanding the budget demands, which presents a landscape that is very different from the era of digital computing.