#AI R&D costs are far too high for startups. Discuss.

https://qz.com/google-spend-100-billion-ai-development-deepmind-ceo-1851412787

Google, Microsoft and others are spending big money on AI. Can startups still compete? Or are they better off waiting for the big boys to train these models and focus on the application of these, regardless of which model wins at the end? Kind of like the cloud platforms, does it even matter if you choose AWS, Azure or Google Cloud anymore? They all largely offer the same service.

But of course, the other side of this argument is that AI is no longer limited by training, that’s what Tesla/Elon is saying. But he didn’t really elaborate on what he thinks is the next limiting factor.
https://www.benzinga.com/news/23/08/34128510/elon-musk-affirms-training-compute-will-not-be-a-constraint-for-teslas-full-self-driving-improvement

Maybe it is an exponential problem. Getting 80-90% is hard enough, but getting 95-99% might take much much longer. Referencing ImageNet progress, https://paperswithcode.com/sota/image-classification-on-imagenet