A curious thought came to mind. I answered my own annoyance at the 30 minutes it took for me to search and calculate the electric-only and gasoline-only energy cost of chevy volt with an imagined google that just answers the question for me when I asked “what is the electric only mileage of a 2018 chevy volt?”
The marketing folks and the EPA mucked it up with “mixed mileage” which is useless to me to decide whether the charger I’m sitting at is cheaper or more expensive than filling up at Costco with my gas rebate Visa card.
Now, my demand is pretty special, but I am protected as a human being and economic agent to optimize my expenditure. The EPA and car makers want to sell cars, and is therefore free to express the information which ever way they want to.
In the future, my AI can certainly read their publication and answer my question(including the answer, “it cannot be determined from available information). But one wonders what happens to our language and culture, when one party has so much freedom (and incentive) to bias the represented ideas that it becomes effectively impossible or at least economically irrational for a second party to take on the expense of understanding what is said to my own advantage. In fact, the training that goes into this have awesome sophistication, the communication is produced in good faith to offer help information according to our social standards, they stand ethically unchallenged and is profitable for the producers to produce them. I.e. what I earn in 30 minutes (plus subsequent time used to complain about it) is more than the 2 cents that I saved for the duration of my ownership of this car at this particular charger at the current Costco gas price provided BoA/AAA doesn’t terminate my gas rebate card program.
There are a lot of people trained in this kind of communication. They include branding folks, they include sales people, they include public relations people. They also include those skilled at encoding it into the HTML my browser received. In the future, the marketers may have the skills to create an AI-document, a product2vec, or advertisement2vec, if you will. My own AI will be compatible with that standard of communication(like my browser can read html), and it interprets the marketing vector it receives and understand it and presents it to me in place of the browser. My AI of course understands my economic needs and my preferences. It will therefore dig for things I need and want.
Since obfuscation occurs in human languages and expressions, one wonders how much obfuscations will be embedded in those future AI marketing vectors? Will it be economically feasible for humanity to figure out the right amount of obfuscation to allow?
Alternatively, this might be the fall of AI, if our consensus comes to that we are all very unhappy and this whole social order built on internet and computer technologies should just fall. Technology will just fail to unite humanity and move us onward. I would have no problem with that. People have to change for the system to change. Not everyone can be like RBG and change the system before people changes wholeheartedly.