Deep Learning Amnesia

I just listened to the TED interview with Yann LeCun, and as he explained how Deep Learning works...it became apparent we are throwing away the most important data we are generating. He says we write small programs, feed them into neural networks, which then spit out some probabilities...BUT, all the patterns and simulations that were generated are essentially thrown out...it becomes a black box. In other words, he isn't researching AI, he is generating a black box oracle. All the critical insights into WHY answers are generated, the interconnected UTILITY of those answers, and the EFFICIENCY at which those answers were generated are discarded...what!? Those would seem to be the most important parts. We need to be able to "distill" some algorithmic wisdom from these "neural networks." No surprise he works for Facebook, a company that doesn't give two sh*ts about improving the quality of society, but instead extracting all the value from it. Sad to learn we are not growing the field of AI, we are creating data abusing machines.