Comparing AI's current state to the internet in 1997 highlights that while potential is clear, most practical applications are yet to be built. It is premature to declare winners like OpenAI, much like betting on Excite over Yahoo in the early web days.
Sticking your head in the sand or complaining about how 'evil' AI is provides a feeling of moral superiority but offers no practical benefit. The only helpful action is to fully engage with the tools, understand their capabilities, and figure out how to use them to become more valuable.
Contrary to the belief that AI will eliminate consulting, labs like OpenAI are acquiring consulting firms. This is because large companies need significant human-led projects to integrate AI into existing systems and workflows, a task they aren't staffed to handle internally.
The idea that companies will fire everyone after buying ChatGPT is naive. Enterprise software sales cycles are 18+ months long, and integrating new tech into core systems takes years. This inherent inertia means AI's impact on jobs will be a gradual evolution, not an overnight revolution.
With AI lowering the barrier to building software, getting user attention is harder than ever. This shifts the competitive advantage to distribution. Incumbents can spray a 'good enough' AI model across billions of users, establishing a default that's difficult for a superior startup product to displace.
Like the internet and mobile, AI will automate many jobs. However, this automation historically unlocks new types of work that don't exist yet. While there's short-term frictional pain, the long-term trend repeated over 200 years is job creation and increased prosperity.
A consultant isn't hired just to create a slide deck, which AI can mimic. They're hired for the strategy and politicking behind it. Similarly, the number of accountants has grown despite tools like Excel because automation frees them for higher-value, non-automatable work.
The backlash against AI isn't a single issue. It's a 'fuzzy mess' combining tangible economic anxiety, statistically insignificant environmental concerns (data center water usage is ~0.017% of U.S. total), and specific community grievances. This complexity mirrors the moral panic around social media.
Pre-AI, few would have guessed coding would be one of the most transformed roles. This proves that breaking jobs into 'automatable tasks' is unreliable. Unforeseen applications can disrupt jobs we assume are safe, like a personal trainer being replaced by an AI that analyzes form via phone camera.
Sam Altman's analogy of selling AI like electricity is flawed because utility providers are low-margin businesses. Without strong differentiation, model labs will face price competition, becoming a commodity. The real value will be captured by applications built on top, just as apps, not telcos, captured mobile's value.
The initial phase of any new technology is applying it to old problems. The transformative phase is creating things previously impossible. The key question for AI is not 'How can we do X faster?' but 'What new thing can we do now?', similar to how Spotify redefined music beyond an online CD store.