More data doesn't help if the additional data points don't add information to the dataset. At some point it's better to add features than simply more rows of o

More data doesn't help if the additional data points don't add information to the dataset. At some point it's better to add features than simply more rows of observations. Arguably text and images are special cases here because we do self supervised learning (which you cant do for self driving for obvious reasons). What TSLA should have done a long time ago is keep investing in additional sensors to enrich data points, rather than blindly collecting more of the same.