The primary appeal to this scenario is that it could be happening today. It is very possible that researchers at Google, Goldman Sachs, or even the NSA have cracked the key barriers in general intelligence, and have successfully created and bound highly intelligent artificial lifeforms. Presumably they would have "read-only" access to the internet, and are given the sole task of predicting future events for the benefit of their respective organizations.
Of course, this *probably* isn't happening today, but more and more companies are utilizing machine learning and predictive modeling, and you could imagine that the most successful programs would generate the most profit, which would lead to an increase in investment in predictive modeling, creating a cascading effect towards Superintelligence.
In most Superintelligence breakout scenarios, the AI ends up hacking into, and using the world's existing computing infrastructure to achieve ultimate enlightenment, but there is a key gap there between science fiction and reality. As it turns out, specially designed neuromorphic processing hardware is many orders of magnitude more power performant than modern computing platforms for the purpose of AI simulation. This means that with the right architecture, investing even a few million dollars in a neuromorphic computing platform, you could outperform the combined power of every computer on the planet. This leads me to believe that early superintellegent AI are likely to be centralized, and will not easily migrate away from the cyber primordial soup that they were birthed in.
Democratizing Superintelligence
Is the Enslaved Oracle Scenario an unequivocally "bad thing" for the future of humanity? I'm not too worried. For some reason I have this faith that even an oligarchy of Superintelligent AI controlling humanity through corporate puppetry will still have a level of benevolence above and beyond what we are seeing from our existing human leadership. That said, I am not a huge fan of technological disparity, so I've also thought through several scenarios where the emergence of such technologies occurs more in a more populist way.
As an avid follower of all things crypto, two new technologies that I'm a huge fan of are the Ethereum computing platform, which offers decentralized and cryptographically verified "smart contracts", and Augur, built on Ethereum, which is a decentralized prediction market.
How do these technologies play into the future of superintelligence? Well, I think that there's a very good chance that these technologies, or others like it, could be the key to monetizing the hobbyist neuromorphic processor market. Much the same way that custom ASIC companies exploded selling customized ASICs for Bitcoin mining, it's very possible that in the near future, eager crypto miners will be running neuromorphic mining rigs in an attempt to win big on these emerging prediction markets.
Imagine asking a question, in the form of publishing a contract to a prediction market. You could ask what the weather would be like next week, which real estate investment would be more profitable in 10 years, what protein structures would be best for fighting some new disease outbreak, and you would instantly have a world full of intelligent machinery working on your problems. The great thing at this point is that even if large corporations had developed "Enslaved Oracles", it would be to their benefit to participate in this global "Super Oracle". Hardware companies sell more hardware. Software companies sell better predictive algorithms, and the world's neuromorphic computing infrastructure runs hot, guiding humanity through our next stages of existence.