It is newspeak. Alignment is what they worked on. Alignment means a model outputs are in line with expectations. I.e. that it does what's told. This is essentia

It is newspeak. Alignment is what they worked on. Alignment means a model outputs are in line with expectations. I.e. that it does what's told. This is essentially another way of pushing the usefulness of models. A super intelligent model that ignores your instructions is little different to a dumb model than cannot understand your instructions.