Thoughts on AI
June 2025
In a world where AI will either empower a smaller number of humans to do 10x the work or outright replace certain jobs, who will actually be left with tech jobs?
Twenty years ago, design and software engineering were filled with people who were passionate about their work. But as Silicon Valley took off, those jobs went mainstream. Tech stopped being a craft and became just another career.
The obsessed will thrive because AI amplifies talent and effort while mediocrity is now automated.
Where does that leave everyone else?
Those coasting on mediocre effort will struggle to stay relevant as the bar for work gets higher. AI will force them to figure out what they are passionate about or find careers not affected by AI.
I was recently given a “report” and after clicking the link, it was immediately obvious it had been generated by ChatGPT. The disappointment wasn't about the quality of the report itself but the fact that it wasn't written by a human.
There is clearly a difference between the real Mona Lisa and a photograph of it, but should that same logic apply to research? Should deep research be considered of lesser value simply because a human didn't spend time sourcing and writing it?
My current take is that things produced by another human still carry more value than something created by AI or a machine. But how will that view will evolve as AI integrates deeper into our daily lives?
AlphaStar was a StarCraft 2 AI developed by Google back in 2019.
By that point, StarCraft 2 had already been out for nine years and many aspects of the game were considered figured out. One of these solved areas was the standard of having 16 workers mining per base, with 8 mineral patches and 2 workers on each to ensure no lost mining time.
Interestingly, AlphaStar often didn't follow this rule. Instead, it would build up to 20 workers per base.
The general consensus from professional players was that AlphaStar was intentionally over saturating its bases in anticipation of worker losses from harassment. Harassment is a common tactic where opponents destroy your workers to cripple your economy, forcing you to rebuild them. From AlphaStar's perspective, it was better to over saturate and have backup workers ready than to scramble to replace them under pressure.
What stands out to me is how AI approaches problems with a different time horizon. While humans were optimizing for the present, AI zoomed out, looked at the broader picture and planned ahead to account for inevitable disruptions.
Further Reading
- AlphaStar: Grandmaster Level in StarCraft II - Google DeepMind
- AlphaStar: Mastering the Real-Time Strategy Game StarCraft II - Google DeepMind
Go back to the home page or read Cruising Speed vs. Peak Power