Same trend as from special-purpose computers in mid-20th century to universal computers today.

This is dubious: there are many trends in computing towards specialisation today, e. g. TPU, however, NVIDIA bets on more universal GPUs. Models are much more malleable, so there is no problem in creating a more specialised model. But I don't have a sufficient intuition to judge whether universal pre-trained models or more specialised ones will work better.

Hawkins' arguments for universal AI:

• Cost effectiveness (Damme argument as for personal use computers in) —- this doesn't make sense, given that models are not hardware and new model designs will be searched for by other AIs.
• Some of the most important applications for AI will require universality — this could be true, but this doesn't mean that most AIs will be universal. Maybe, universality Willbe too expensive to run, so it will be reserved for the most important applications.

Related:

• AI leveraging computation is more effective than leveraging human knowledge
• ‣

References

A Thousand Brains