Artificial Intelligence, The Future
A much needed conversation condensed.
Its not going to happen tomorrow but AI will be cracked in the near future and we need to have a conversation about how we want to use this tool.
I’m going to use Siraj Raval’s Talk on AI and Deep Learning for a quick history of AI
When AI was first being developed, the behavior was very rule focused. “If x happens do y ”. Example: “If translating a sentence, do a direct word for word translation. Don’t worry about context”. It was a very rigid system but it was the first iteration. The next phase that became of AI was based on options. “Given x options, pick the best option”. In practice, its exactly how Google Maps goes through different paths to your destination and picks the best one.
Definitely a lot better than the previous step but now we are venturing into the learning phase of AI. We give the system an objective and a bunch of data and let the system decide the best way to complete the objective. Example: Image Classification. We give a system a bunch of images of different kinds of dogs so the system can identify dogs in images. Instead of giving the system explicit rules of what a dog is and isn’t, give the system all the data it needs to learn what the essence of a dog is so it can point out dogs in images.
The correct behavior is very important in training AI. You’ve probably already heard the story about Facebook and their AI. Facebook was working on a chatbot that could intelligently converse with humans. They had two bots talking to each other about splitting a list of items but bots started to use english in a more representational manner than literal. The point of the bot was to use english like humans do not in their own way. The change in language protocol is what led to its shut down.
We are still learning how learning works. Eventually we will crack it and the effects on our society will be very signficant.
The biggest and most immediate hit we will take is job loss. Machines are already taking up jobs like cashiering and self-checkout has been a thing forever. Why have truck drivers when you can have self-driving trucks? Why have cooks when you can just have a robotic kitchen?
It’s already starting to happen. In Japan, Fukoku Mutual Life insurance replaced 34 workers with IBM’s Watson Explorer AI. The program is supposed to calculate policy holders and payouts. They believe this shift will increase their productivity by 30% and save them 140 to 200 million yen (1.3 -2.5 million dollars).
In the future there could be massive problems with Human-AI relationships. Elon Musk thinks there’s a big chance that humans could become second class citizens because “there wouldn’t be many jobs that AI can’t do better.” AI would be superior compared to the inefficient humans.
Why should we keep it going then?
The benefits are huge. Way huge. AI can solve our problems faster than we imagine. It can go through lifetimes of scenarios and learn from them faster and better than we humans can. AI and robotics can do jobs too dangerous for humans. Its being used find treatments for cancer. AI is not going to start running companies. People are still needed to run the software and hold the company together in turn, more jobs can actually be created using AI.
AI is coming and we can thrive if we chose to use it wisely. We need to start having conversations now about regulating AI and the path we want to walk down.