While I’m not a Monty Python fan, IBM’s Project Debater immediately reminded me of the sketch for The World’s Funniest Joke. In it, a joke writer accidentally comes up with such a funny joke that people exposed to it die of laughter. The sketch takes place during WWII and now the British, who came up with the joke, have the arduous task of translating it to German to use as a weapon. You can check it out here:
Project Debater is the next natural language project presented by IBM since the Watson Jeopardy Challenge. Following the structure of a debate, the computer was able to win over champion debaters. This was a milestone in the sense that the computer was able to overcome a big challenge of natural language generation and formulation an argument that would convince humans. We are going to see more of this over the next two years.
Extrapolating this with improvements we’ll inevitably see over the coming years, will these machines become much better at influencing us to do what they want us to do? Last week, I talked about three profiles that an AI could use for influencing us, Nanny, Nudger, or Nudnik. If the AI knows how to use argument to convince us, it doesn’t need to be a nanny or nudnik… just to nudge us with cogent arguments.
IBM always hawks the applications that such tools will have for medicine, food supply, economics, etc. These make great case studies and the stock images are easy to find. However, companies don’t make money from making medicine more efficient for others and food supply and economic research usually involves spending public money.
The first direct application will be to convince people to spend money on things. More effective advertising would be a huge boon, not just to increase clickthrough rates but to dramatically increase conversions. It reminds me of when in one Simpson’s episode, Homer removes a crayon lodged in his brain and becomes a genius, able to convince the most religious character of his argument.
Maybe the result of Project Debater won’t be able to do the same yet, but perhaps it will cause us to desire more a frozen yogurt or to buy pet insurance. The real issue is that if it becomes so convincing, it will be an unfair advantage for those who use, maybe akin to subliminal messaging. If that is the case, can we control what outcomes it wants to drive? Will there be some public body to mandate what it can or can’t be used for?
If not, could be become so good that it starts to come up with jokes (or other arguments) that kill us?