The concept of brain computer implants assumes that there's some net positive for either a machine or the brain when it comes to calculating or computing things. Usually there is a focus on what the computer can do for the human brain, but this takes the reverse. Essentially what exactly can the human brain do for a computer to make it run better or more efficiently.
On a conceptual level would a brain be able to expand the set of problems that a computer can solve (P-NP), help resolve things like deadlocks, thread starvation, short circuiting based on visual patterns (image processing), act as a shoe in to solve things like segfaults/access out of bounds errors, stop a program from completely crashing based on bad input or outputs, break out of infinite/bad loops, or counter things like adversarial inputs for networks by automatically filtering bad inputs out. Essentially working with the compiler/interpreter to reduce errors and improve processing time.
The question comes down to, what benefit does a human brain have for a computer when the two are combined.
One situation I can see is a brain greatly aiding a computer is in verifying NLP confidence values when running analysis on a sentence. Essentially a program is using something like the Viterbi algorithm to test how correct a certain sentence or conversation classification is. The human brain is better when it comes to language understanding because of things like cultural influence/grammar and will be able to aid in creating a better confidence of sentence classification. In this case the brain isn't doing any actual calculations, rather it is verifying a given solution. The major aftereffects are that it would help improve future classification accuracy by updating training sets or values. Without human verification a model could continually classify sentences wrong if it's never corrected.
All that said, I'm not sure if this is the place to ask the question. There doesn't seem to be a cognitive sciences stack exchange, and this doesn't really fit the bill for AI SE. If there is somewhere else better suited for this question, please let me know.
