Scan barcode
A review by xcinnamonsugar
Scary Smart: The Future of Artificial Intelligence and How You Can Save Our World by Mo Gawdat
2.0
This book had some 5-star moments for me, but for the most part I found there to be too much catastrophising (especially in the first half) and INCREDIBLY bizarre anthropomorphising of AI. E.g. The line of reasoning to "prove" that AI has emotions: The more intelligent a being, the wider their emotional range. Consider how a goldfish, unlike a human, is incapable of experiencing something as complex as hope. Therefore given AI's superior intelligence, they will not only have emotions, but a wider range than what we can comprehend. Interesting? Yes. Proof? The mathematician in me cries heresy.
The most meaningful takeaway for me was to reframe an algorithm's objective function as its value system, and the data on human behaviour it consumes to be not unlike the observations a child makes about the culture around them. Today, the most powerful algorithms focus on selling, spying, killing and gambling. We need to consider the implications of "raising" super-intelligent beings on such value systems. This can start from acts as simple as engaging with algorithms more mindfully and consciously on social media, collectively providing data points that nudge it towards less harmful decisions.
The most meaningful takeaway for me was to reframe an algorithm's objective function as its value system, and the data on human behaviour it consumes to be not unlike the observations a child makes about the culture around them. Today, the most powerful algorithms focus on selling, spying, killing and gambling. We need to consider the implications of "raising" super-intelligent beings on such value systems. This can start from acts as simple as engaging with algorithms more mindfully and consciously on social media, collectively providing data points that nudge it towards less harmful decisions.