> simply saying "we should make this better thing" isn't enough.

Besides the references to his company which has customers and a product that already works on these principles the literature currently shows that this is very much possible if you dig into the correct niches. Besides the SOTA in few-shot and meta-learning it is possible to smartly choose the correct few samples for the network that yield the same results.

It has also been my primary focus for the past 5 years and the core of the company I founded.

> it is possible to smartly choose the correct few samples for the network that yield the same results.

And then, someone is using pretrained 500B model, and fine-tuning your few examples, and getting new SOTA.

They might get new SOTA because the metric is accuracy, but if the metric was accuracy weighted by sample efficiency, then SOTA would look a lot less impressive.

Simplest way to weigh by sample efficiency: multiply accuracy by ratio of test set to training set sizes. Everyone's training/testing on 80/20 splits, so everybody's SOTA would go down by 3/4s.

It's more of "this direction seems higher ROI than that direction", in particular quality vs quantity of data.

Already in 2018 SenseTime reported that for face recognition, clean dataset surpasses accuracy of 4x larger raw dataset.

https://arxiv.org/abs/1807.11649

«Small data /ai/» is not "impossible", it is actually necessary: AI, opposed to this ML, implies perfectioned digestion of the input data.

Only, the article seemed to show a very conservative Ng about the algorithms, a focus on data management - so it's still ML.

I would say that Andrew Ng has some credibility in putting practice to his preaching.

Atleast someone's working on it.