Artificial Intelligence: Consequences
AI is going to change the world.
Like the free market, AI is a ruthless optimizer. However, like the invisible hand of the market, AI makes mistakes. Markets create bubbles that burst, and crashes that correct into rational valuations. We know that the free market is an imperfect optimizer, but it is the best one we have so far. Generally, when you zoom out and see the big picture, the direction of market progress is up up up.
AI and market economics are two optimizers. They intersect in stock and currency trading bots. If we want to keep the exponential global growth trend line going, we need disruptive innovations like AI. Progress, however, is not a zero sum game. Instead, the bigger the circle of things you know, the bigger the circumference of things you don’t know gets. In previous articles, I mentioned the link between growth investing and AI: fear. The fear of what AI can do leads to greed for the same. But just as the free market is not so good at some tasks, like innovation (government grants and military research gave us the internet, and even AI itself), so too we can see that AI is not good at all problems, at least not for now. Like the free market, AI has a set of predictable failure stereotypes, including overfitting, poor precision-recall trade-off, learning from bad data e.g. bias, optimizing for the wrong objective, etc.
However, when set upon the right kind of problem, artificial intelligence is extremely effective. Machines learning to do regression or classification is one thing. That’s the kind of nerdy computer stuff that, like Moore’s law, we expected to get better, regardless of AI. But what can we do with AI that is really unexpected? Well, there are many examples, including style transfer, idea comparison, title generation from article text (summarization), and machine translation. We sometimes forget that simply recognizing objects in images used to be sci-fi-esque artificial intelligence, and now it is simply convolutional neural networks, often pre-trained and sitting behind an API. Once you know something is solved, it seems to go from magic to logic pretty quickly in the human mind.
AI exposes new possibilities that were not expected to emerge so quickly. It’s a knowledge and automation gold rush. This all means we humans are in for a lot of disruption, and not all of it will be good.
We humans model in our brains what machines will do, and that process is about to get weird.
Take the example of a broken vending machine. Imagine walking up to a vending machine, putting in your credit card, pressing the soda selection button, and realizing that the vending machine did not dispense a drink. In this scenario, your brain thinks about how a vending machine works, and it models what could have gone wrong. Did you press the button hard enough? Maybe it was just a mechanical failure to press the switch. Did the card reader work? Maybe the communication between the machine and internet is down. Maybe your credit card transaction was denied. These are questions you can ask because you get the idea on an intuitive and physical level of what the parts in a vending machine do.
However, when we swap out the simple vending machine for a chat bot vending machine, one that takes your order as a voice command, and dispenses a drink - or in this case it fails to dispense a drink, more parts come into play. In this AI vending machine scenario, you have a whole new set of things that may have gone wrong. Did the machine hear your request? Did it properly convert all of your utterance into text (STT)? Did the text get converted into your intended action (NLU)? Did the AI understand that you paid? Added to the mechanical and electronic models in your head, is this whole new layer of things that can go wrong. Now your brain will have to model how the AI inside products “works”.
AI will create worse and more annoying user experiences. However , it also opens up new better user experiences. It is straightforward to create a chat bot that asks for fast food orders at a drive through. Tim Hortons order collection system anyone?
As many jobs are lots to automation, humans will need to pick up more computer interaction skills. If our new jobs are as robot overlords, let’s make sure we can rock that role and continue growing the aggregate value of our shared economy.
I recently discussed on a LinkedIn thread how all humans need to learn how to program at a basic level. If we want to remain optimistic about AI, then we need a stronger tech savvy skill set in our human population. We can’t expect to keep our automation-friendly jobs forever. Few jobs that existed in 1899 are still around today. And so, will AI come for your job? If your job is easy to automate with AI, then yes it will. New technology has consequences. But you will, as we have throughout human history, rise above.
Learn to program. Go to w3schools.com for free lessons. Or try udacity.
If you enjoyed this article on some of the implications of artificial intelligence, then please try out the clap tool. Tap that. Follow us on medium. Go for it. I’m also happy to hear your feedback in the comments. What do you think?
Happy Coding!
-Daniel
daniel@lemay.ai ← Say hi.
Lemay.ai
1(855)LEMAY-AI
Other articles you may enjoy:
