Tom Goldstein, a computer scientist at the University of Maryland, and his colleagues added a twist to how they presented numbers to a transformer that was being trained to add, by embedding extra “positional” information in each digit. As a result, the model could be trained on 20-digit numbers and still reliably (with 98% accuracy) add 100-digit numbers, whereas a model trained without the extra positional embedding was only about 3% accurate.

So they had an idea on how to fix a completely dysfunctional method and their result is, that a computing task that we can perform automatically and with 100% accuracy for 80 years or so, using a different and well understood method, can now be done at much higher energy costs using said inadequate method with 98% accuracy.

Why are these people still receiving research money?

So now they’re awash with money, and this has become the topic DU jour, is that why they’re now pivoting to the 3 pillars of ai thing? I.e. reasoning, planning and external control?