For the majority of people, artificial intelligence won’t be here for at least the next decade, despite all the hype. But the truth is it has already become a necessity for many businesses who work with data and it’s been widely in use today.
While some of us are still trying to figure out the difference between artificial intelligence and machine learning, AI is fast progressing. This breakthrough technology has already become accessible for any software developer; tech giants are currently competing to dominate the field of artificial intelligence; China has taken serious steps to become the leader in AI; some jobs might soon be automated, and we’ve seen some unprecedented advances in deep neural networks.
AI is taking the world by storm. If you don’t want to fall behind, here are the latest developments to keep an eye on.
1. AI is becoming more accessible.
A few years ago artificial intelligence was a prerogative of big tech companies like Amazon, Google, and Microsoft. They can hire an army of the best tech talent and invest a truckload of money into innovations. Today any company can capitalize on AI.
Artificial intelligence is becoming affordable because of a variety of open source software that allows for building advanced self-learning systems. TensorFlow built by Google is one of them.
TensorFlow is used by Dropbox, eBay, Intel, Twitter, Uber, and a great number of companies who provide AI consulting services. It’s one of the most well-maintained frameworks for machine learning. Keras is another library for building deep learning models. It is built on top of TensorFlow and allows to quickly build and test a neural network with minimal lines of code.
Amazon, Google, and Microsoft are currently in a rush to build AI-based platforms for any company to use. They’re working on creating AI utilities for their AWS, Google Cloud, and Azure cloud computing services.
Because machine learning models get better with more data, customers who started using one vendor are very unlikely to change it. That’s why the competition for dominance is heating up. The company who wins the race for AI can become the operating system of the future with the revenue that will be twice as much as the current $260 billion cloud market.
2. China might soon be leading the way in AI
A few years ago Chinese technology entrepreneurs were focused on repeating Western success stories. Chinese social network Renren clearly copied Facebook. Alibaba is often described as the “Amazon of China.” And WeChat started as a copy of Whatsapp.
Today it looks like the era of “Copy-to-China” is over. In fact, we might be entering the new era called “Copy-from-China.”
Chinese tech industry has set off to become the world leader in AI and machine learning. Last year, the total global investment into AI-focused startups amounted to $15.2 billion. Chinese investors poured 48% of this money, which is 10% more than the United States (the current leader in the field) who invested only 38%.
What’s more, the Chinese government wants to significantly improve education to build a strong AI talent pool. And they’re also planning to invest $15 billion in artificial intelligence companies.
Some of the largest tech players working on AI innovations include Baidu, the dominant Chinese internet search engine company, Tencent who owns WeChat, and Didi the ride-sharing giant who bought Uber’s Chinese operations.
3. AI is automating routine work, not taking people’s jobs. At least for the next couple of years
“AI is coming for your job” is one of the most discussed AI-related topics on the internet. The advancements in artificial intelligence do look like computers are much better than people at performing some tasks. For example, computers can go through hundreds of scanned legal documents in a few seconds and retrieve the needed data for a lawyer. But does this capability of an AI application mean that lawyers will soon disappear? Absolutely not.
Everybody heard about how AI conquered chess. But very few people know that what can beat an AI chess champion is a freestyle chess or centaur chess where a human player and an AI program play chess as a team. The same with lawyers. An AI algorithm can automate the process of legal document review thus making a lawyer much more productive.
According to PwC’s job automation study, only 3% of jobs are at potential risk of automation by 2020. This doesn’t mean we can all just relax and take it easy. The same study claims that by mid-2030s 44% of workers with low education at risk of automation. The future of the job market will belong to people with creative minds who invest in lifelong learning.
4. AlphaGo’s victory is one of the largest landmarks for AI.
Since we’ve already touched upon that significant chess game where machine won, let’s think about what it means for the future of AI.
When Google’s AlphaGo crushed Lee Sedol, one of humanity’s best Go players, it made a case for reinforcement learning.
Reinforcement learning is a technique in machine learning where a computer learns how to behave in a given situation without any instructions. It learns by performing different actions that lead to positive and negative outcomes. Think of golf, for example. Using reinforcement learning, a computer program can learn how to hit the ball into the hole in one stroke by trial and error. Once the ball gets into a hole the program will get a positive reward (which is simply feedback) and will remember the actions that led up to it.
Unlike golf, winning a chess game – and especially the game of Go – is a hard problem to solve. The AlphaGo’s victory means that reinforcement learning can be set to analyze any type of situation.
Machine learning algorithms that power intelligent systems today need to be trained on large labeled datasets. But supplying training data for a specific situation isn’t always possible, and the process itself is rather expensive. With reinforcement learning, you don’t need all that data. Reinforcement learning opens up great opportunities for building AI applications with general-use deep learning algorithms. AlphaGo proved that it’s realistic.
6. Dueling neural networks bring imagination to AI.
Machines learning systems can’t create their own things because they don’t have imagination. But it seems like the solution has been found.
In 2014 Ian Goodfellow, a Ph.D. student at the University of Montreal was having an academic argument in a bar. He came up with the idea of a generative adversarial network or GAN which might well be the solution that data scientists have been looking for.
The idea of this approach is to train two neural networks on the same dataset and make them play a so-called “real or fake” game. Here are the rules: let’s assume our neural networks have been trained on the images of cats. One of these networks needs to create variations on images it has seen (for example, it can add an extra tail to a cat in the picture). The other network gets to decide which of these images is like the one it has been trained on (the real image) and which one is created by the generator (the fake image). With time the neural network that needed to create variations on images will learn how to do it really well so the other network couldn’t spot the difference between real and fake.
Researchers have been doing experiments using GAN and they have already achieved some great results. For instance, in one experiment, a neural network could create credible faces of people who don’t exist.
Dueling neural networks open an opportunity for data scientists to create entirely synthetic datasets that can be used for training machine learning models.