Comment by MHard
1 year ago
I think the idea would be to compete in different segments. Startups often focus on problems where hypergrowth is at least possible with massive investments. But the examples listed in the article advocate for ideas that might generate millions but not billions in revenue.
Some of the ideas the author is proposing are already happening in Germany. There are grants to develop games for instance which pay you a monthly salary for 1 year and a while back there was a similar one for open source projects.
There is a huge gap in the market, I'd say between capital investment of $50,000 and $5,000,000, where a ton of ideas which can make 20%+ ROI per year once mature live. But because they will never hyper scale they can't be financed by startups, and banks are too scared and stupid to invest.
I've done the math for a document segmentation pipeline that every RAG system needs and you'd be able to get one off the ground for around $2,000,000 USD and be rolling in cash instantly. Good speaker diarization is another family of models that will print money once solved, but again, there isn't the possibility for hyper scale.
Meanwhile at least a dozen companies I've spoken to have wasted on the order of $10,000,000 trying to solve these problems in house and failed.