There are 2 big engineering problems yet to be solved before the robots can give us psychotherapy er whatever:
Firstly, faster, more sophisticated computing architecture. The processors we rely on now are fast but not fast enough for human-like AI/MI writ large. Solutions such as quantum computing are in development but have a long way to go. This video explanation is excellent.
Secondly, power storage is still a problem for our street walking robot pal, although there are some very promising potentials out there right now.