Nanobots and fighting spam
One of my favorite fiction writers is Michael Crichton. His books often center around technology and how when humans try to control or influence the physical environment, things go awry. In Jurassic Park, humans recreated dinosaurs and attempted to profit from it. In Sphere, they discover a 3-dimensional circle at the bottom of the ocean and try to weaponize it, only for nearly everyone to die. The central theme of these books is that the natural environment is still a longs ways away from being controlled by science.
The book I just picked up this weekend is Prey. I'm only about 1/4 of the way through it, so please, no spoilers in the comments section. There are two things so far that have jumped out at me.
First, some humour. At one point, the protagonist Jack describes his career. He is a manager of engineers and programmers. He says at one point that he used to program, but that programming is a young person's job and he doesn't do that anymore. I laughed out loud when I read that. I don't program anymore either. I manage projects and do research. Am I really getting that old?
Secondly, Crichton often winds real facts into his stories. In this story, it talks about molecular manufacturing, or what most normal people would call nanotechnology. The main character has done some nanotechnology work where little microbots learn and evolve. One of the techniques that they learn to understand is natural language processing; Crichton doesn't go into further details, but I take this to mean that nanobots create an artificial life form that understands real language that people use.
What immediately jumped into my head was using such a technology to fight spam. While molecular technology and spam fighting is probably near the bottom of the list of possible uses, it really could be kind of cool. Rather than a spam analyst having to think of patterns of phrases and text in languages that indicate spamminess, a neural network could learn to understand what patterns of text occur in messages which mean that it is spam.
The spam that we come across is still primarily based in the English language. Spam analysts need to be fluent in English. This is the case because English language spam uses a lot of slang (so does the regular English language). In sexually explicit spam, words that can be used in normal language, when combined together, only occur when spammers are hawking a product. Other types of spam like mortgage refinancing or Viagra are equally good examples.
What if a neural network containing molecular particles worked as a spam filter? This filter would not need to be continuously populated with new streams of good mail and bad mail. Instead, it would learn to recognize spam based upon previous patterns and then make predictions on new types of spam that it hadn't seen in the past. One of the weaknesses of current filters is that is spammers change spam content from new bots, spam filter effectiveness drops. With a neural antispam network, it wouldn't matter. The spam filter would "know" that the new type of message is spam because it has learned to recognize what spam is (and more importantly, what it is not). As humans, we can do this pretty easily. Current spam filters struggle. The spam filter of the future will not.
I'm just typing down some random thoughts but I think it's kind of a neat idea to think about.
