Of course, it's not guaranteed that AI innovations will diffuse throughout society. At some point perhaps governments will take control, in the style of the Manhattan Project, and they'll keep the advances secret. But even then, I expect that the internal advances by the research teams will add cognitive abilities in small steps. Even if you have a theoretically optimal intelligence algorithm, it's constrained by computing resources, so you either need lots of hardware or approximation hacks (or most likely both) before it can function effectively in the high-dimensional state space of the real world, and this again implies a slower trajectory. Marcus Hutter's AIXI(tl) is an example of a theoretically optimal general intelligence, but most AI researchers feel it won't work for artificial general intelligence (AGI) because it's astronomically expensive to compute. Ben Goertzel explains : "I think that tells you something interesting. It tells you that dealing with resource restrictions -- with the boundedness of time and space resources -- is actually critical to intelligence. If you lift the restriction to do things efficiently, then AI and AGI are trivial problems." 1
Should those methods fail, the next step is to seek the specialized Behavioral Intelligence services that reference OSINT to aid in establishing the veracity of subjects during the forensic interview and is used to create a behavioral profile. OSINT data is correlated with interview data to include a variety behavioral patterns; ., a list of daily personal contacts, habits of activities, visited places of interest, vehicles used, favorite group involvements, etc. According to the Director, psychologist and forensic interviewer at MN-Behavioral Intelligence Agency, (2016) OSINT data base has to be critically filtered and analyzed before it can be applied within investigative interviewing and interrogation.