Rethinking The Way We Benchmark Machine Learning Models

Do machine learning researchers solve something huge every time they hit the benchmark? If not, then why do we have these benchmarks?

Link to Full Article: Read Here