We're in the middle of the second renaissance where proof of work replaces the resume, an audience replaces the employer, and taste replaces credentials.
A simple framework for identifying high-value applications of AI:
1. AI can partially or fully automate a task, resulting in 10x better, cheaper, and/or faster results;
2. Humanity currently pays >$10B per year to perform this task;
3. Usage of the product generates proprietary data/feedback that becomes difficult to replicate elsewhere as it scales.
I disagree.
Every AI provider that isn't winning the data network effect is incentivized to counterposition, telling you a story about why the network effect is bad.
But a model trained on my information, without the information of millions of other users would be awful.
It is generally the best thing for the world for us to put our combined knowledge into these models rather than siloing inferior models.
Furthermore--this is simply unstoppable. Every network effect business has competitors that try to convince you that you are better off without the network effect:
* Social networks that tell you you should own your social graph (failed)
* Desktop linux users telling ordinary consumers they should use linux rather than Windows/MacOS (largely failed)
* OpenStack which said you should run your own cloud computing infrastructure rather than AWS or Azure (not sensible to the vast majority of companies)
But the network effects always win in the end. You deleting your Instagram account does little to Meta. You deleting your Claude account does little to Anthropic.
There are hundreds of millions of users contributing knowledge to these models every day whether you do or not, and it's going to be almost impossible to get everyone to stop.
If you over-focus on proprietary models you are largely going to be left behind by the snowballing intelligence of foundation models.