There are two ways most people respond to a new technology. The first is uncritical enthusiasm. Everything new is going to change everything, solve every problem, and unlock a future that feels like science fiction. The second is reflexive fear. Every new technology is going to destroy jobs, privacy, democracy, or human connection. Both are forms of not thinking clearly. Here is a better framework.
The trap of both extremes
Hype and panic share a structural problem: they both substitute emotion for analysis. The enthusiast and the fearful person are doing the same cognitive move, just in opposite directions. They are pattern-matching to a pre-existing narrative rather than actually evaluating what they are looking at.
The enthusiast's narrative: technology is always progress, progress is always good, therefore this technology is good. The fearful person's narrative: new things disrupt existing things, disruption causes loss, therefore this technology will cause loss. Both are technically correct in some cases and completely wrong in others. Neither tells you anything useful about the specific technology in front of you.
The five questions worth asking
When a new technology appears, or when you want to evaluate a technology that is already here, five questions do most of the work.
What does it actually do? Not what the marketing says. Not what the critics say. What does it actually do, in the real world, today? This requires reading technical descriptions, watching actual demonstrations, and ideally using it yourself. Many technologies are discussed by people who have never used them.
What are its actual failure modes? Every technology fails in specific, predictable ways. A hammer will drive nails but it will also miss and hit your thumb. AI systems generate confident-sounding text but sometimes generate false information. Understanding failure modes is not pessimism. It is the minimum knowledge required to use something safely.
Who designed it and what did they optimize for? Technologies are built by people with specific goals, constraints, and incentives. Social media platforms were optimized for engagement, which turned out to mean they were also optimized for outrage. Navigation apps were optimized for the fastest individual route, which sometimes produces traffic jams that are slower for everyone. Understanding what a technology was built to maximize tells you a lot about what it will actually produce.
Who benefits and who bears the costs? Rarely are the beneficiaries and the cost-bearers the same people. The people who benefit from a technology are often the ones talking about it most. The people who bear the costs are often not in the room when decisions are made. Asking this question does not require you to oppose the technology. It requires you to have an honest picture of what it actually does in the world.
What does it make more likely to happen? Technologies do not determine outcomes. They change probabilities. A printing press makes mass communication more likely, which makes both mass education and mass propaganda more likely. Smartphones make both instant connection to loved ones and compulsive distraction more likely. Asking what a technology makes more likely, rather than treating it as inevitable or impossible, is the most useful frame.
Applying this to AI
Run AI through these five questions and you get a much more useful picture than either "AI will save us" or "AI will destroy us."
What does it actually do? It recognizes patterns and generates outputs based on those patterns. It can do this at remarkable scale and speed in certain domains.
What are its failure modes? It generates false information with false confidence. It reflects biases in its training data. It can be used to deceive at scale. Its outputs require human evaluation to be reliable.
What was it optimized for? Depending on the system, it was optimized for accuracy, engagement, helpfulness, or commercial adoption. These produce different systems with different tradeoffs.
Who benefits and bears costs? The beneficiaries of AI capability are currently the organizations and individuals with the technical sophistication to deploy it well. The costs in terms of labor market disruption and information environment degradation are more diffuse.
What does it make more likely? Significant productivity gains for people who use it well. Significant vulnerability for people who use it poorly or do not use it at all. A more complicated information environment. Faster progress in medicine, science, and engineering. Greater concentration of capability among well-resourced actors.
This is not a reason to be afraid of AI. It is a reason to understand it clearly.
The operating principle
The best operating principle for thinking about any technology: be curious, be specific, and be honest about what you do not know yet. The future is not determined by the technology. It is determined by the choices humans make about how to use it, govern it, and distribute its benefits.
That is a reason for neither panic nor uncritical enthusiasm. It is a reason to think carefully.
For a deeper version of this kind of thinking applied to AI, quantum computing, and the other forces reshaping civilization, everything published on this blog is free. The book collection goes further, and all titles are available on Amazon.
Understand today. See tomorrow.