A rapidly growing trend in the business world is the growth of virtual teams. Some companies are allowing teleworking, which means that more and more workers have to coordinate their activities with other team members through the phone or internet. Even among workers who are working from their office, more and more companies are having employees work in teams with employees who work at other locations who they may or may not have ever met in person. New technologies have been developed to help facilitate virtual teamwork, but challenges still remain.
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