I've been hearing the Neuro-symbolic cheerleading for 5 years now. I remember Yoshua bengio once debating against it and seeming dogmatic about his belief in pure learning and in how neurosymbolic systems wont solve all the limitations that deep learning has. I have yet to see any results and don't expect to see any. My guess is that transformers continue to scale for 5 more years at least and we will stop asking questions then about what paradigm shift needs to take place because it will be obvious that the current paradigm will do just fine.