In the back-office automation example is:
- Assessing applications for a mortgage or a business loan
Really? Are we really giving these things as prime examples of what LLMs can automate.
I do not want my mortgage application or business loan rejected because some un-auditable, non-deterministic, inference decided so.
The tech upper classes really are sheltered from the real impact on real people that trusting IT can have.
There's countless scandals of faulty algorithms (which are tractable) causing people real pain, often with suicides resulting.
Using a black-box inference machine instead is MUCH worse.