You’ve probably heard of artificial intelligence by now. It’s the technology powering Siri, driverless cars, and that creepy Facebook feature that automatically tags your friends when you upload photos.
But what is AI, why are people talking about it now, and what will it mean for your everyday life? SunTrust recently released a report that breaks down how AI works, so let’s tackle those questions one at a time.
What is AI?
AI is what people call computer programs that try to replicate how the human brain operates. For now, they only can replicate very specific tasks. One system can beat humans at the complicated and ancient board game called Go, for example. Lots of these AI systems are being developed, each really good at a specific task.
These AI systems all operate in basically the same way. Imagine a system that tries to identify whether a photo has a cat in it. For a human, this is fairly easy, but a computer has a hard time figuring it out. AI systems are unique because they are set up like human brains. You feed a cat photo in one end, and it bounces around a lot of different checkpoints until it comes out the other end with a yes or no answer — just like your eyes passing your view of a cat through all the neurons in your brain. AI is even talked about in terms of neurons and synapses, just like the human brain.
AI systems have to be trained, which is a process of adjusting these checkpoints to achieve better results. If one checkpoint determines whether there is hair in the photo, training an AI system is like deciding how important the presence of hair in a photo is to decide whether there is a cat in the photo.
This training process takes a huge amount of computer process to fine tune. The better a system is trained, the better results you can get from it and the better your cat photo system will be able to determine whether there is a cat in a photo you show it. The huge amount of processing power required to run and train AI systems is what has kept AI research relatively quiet until recently, which leads to the next question.
Why are people talking about AI all the time now?
There is a famous AI contest where researchers pit computers against humans in a challenge to correctly identify photos. Humans usually are able to identify photos with about 95% accuracy in this content, and in 2012, computers were able to identify about 74% of photos correctly, according to SunTrust’s report. In 2015, computers reached 96% accuracy, officially beating humans for the first time. This was called the “big bang” of AI, according to SunTrust.
The big bang of AI was made possible by some fancy new algorithms, three specifically. These new algorithms were better ways of training AI systems, making them faster and cheaper to run.
AI systems require lots of real-world examples to be trained well, like lots of cat photos for example. These cat photos also had to be labelled as cat photos so the system knew when it got the right answer from its algorithms and checkpoints. The new algorithms that led to the big bang allowed AI systems to be trained with fewer examples that didn’t have to be labelled as well as before. Collecting enough examples to train an AI system used to be really expensive, but was much cheaper after the big bang. Advances in processing power and cheap storage also helped move things along.
Since the big bang, there have been a number of huge strides in AI technology. Tesla, Google, Apple and many of the traditional car companies are training AI systems for autonomous driving. Google, Apple and Amazon are pioneering the first smart personal assistants. Some companies are even working on AI driven healthcare solutions that could personalise treatment plans based on a patient’s history, according to SunTrust.
What will AI mean for your life?
AI technology could be as simple as making your email smarter, but it could also extend your lifespan, take away your job, or end human soldiers fighting the world’s wars.
SunTrust says AI has the capability to change nearly every industry. The moves we are seeing now are just the beginnings, the low hanging fruits. Cities can become smarter, TSA might be scanning your face as you pass through security and doctors could give most of their consultations via your phone thanks to increased AI advancements.
SunTrust estimates the AI business will be about $US47.250 billion ($A61 billion) by the year 2020. Nvidia, a large player in the AI space thanks to its GPU hardware and CUDA software platform, is a bit more conservative. It only sees AI as a $US30 billion business, which is four times the current size of Nvidia.
There is no doubt AI is a huge opportunity, but there are a few companies you should watch if you’re an investor looking to enter the AI space, according to SunTrust.
- Nvidia (NVDA): The company makes the GPUs which are the backbone to running AI processes quickly, and has developed industry-standard software called CUDA that lets programmers better user the companies hardware. An entrenched company culture of innovation has been their secret sauce for years.
- Google (GOOGL): Google is at the forefront of AI research. It’s Alpha GO system made headlines for conquering a hugely complicated board game, but it’s also working on driverless cars, smart assistants and so much more. The company recently developed what it calls a Tensor Processing Unit, TPU, that should make waves in the AI hardware markets.
- Intel (INTC): The company primarily is focused on CPUs, which aren’t great for training AI systems, but are good for running the AI systems once they have been trained. They also are the core of a majority of the servers and computers the AI software runs on.
- ARM (bought by SoftBank in 2016): The chipmaker is focused on bringing AI closer to the consumer, according to SunTrust. The company produces low-power consumer chips that they want to use to bring AI off the potentially insecure cloud and let users interact more directly with it.
- Qualcomm (QCOM): Qualcomm, like ARM, is working on small versions of AI systems. They want to be the chip maker in drones and other internet-of-things devices, according to SunTrust.
- So many others: Remember, AI will probably be a part of every company eventually.
One is for sure. AI is exciting, sometimes scary, but ultimately, here to stay. We are just starting to see the implications of the technology.