[PeterisP]

ASCII digits do not always imply base-10 numbers, they can also be identifiers (e.g. phone numbers), parts of words (IPv6, Log4j), and used in various 'written slang' such as g2g, 4ever, m8 for mate, etc, etc.

And, crucially, I'd argue that for in "chatbot" tasks those other uses are more common than arithmetic, so arbitrary focus to specifically optimize arithmetic doesn't really make sense - the bitter lesson is that we don't want to bias our architecture according to our understanding of a specific problem space but rather enable the models to learn the problem space directly from data.

[uoaei]

You're missing the picture again.

Stepping one level out in the metacognition hierarchy is the key. "Learning to learn" as it were. It is only the relative ease of implementation and deployment of feedforward models like Transformers that makes it seem like we have reached an optimum but we desperately need to move beyond it before it's entrenched too thoroughly.

[PeterisP]

Okay, but it does seem that this hack is in the entirely opposite direction; a pure transformer is more towards "learning to learn" than any special preprocessing to explicitly encode a different representation of numbers.

We probably do have to move beyond transformers, but not in the direction of such hacks, but rather towards even more general representations that could encode the whole class of all such alternate representations and then learn from data which of them work best.

[uoaei]

You seem to be making my point just fine. What was your confusion, then?