Artificial intelligence (AI) researchers have a long history of going back in time to explore old ideas, and now researchers at OpenAI, which is backed by Elon Musk, have revisited "Neuroevolution," a field that has been around since the 1980s, and they've achieved state of the art results. The group, which was led by OpenAI's research director Ilya Sutskever, explored the use of a set of algorithms called "Evolution strategies," which are aimed at solving "optimisation" problems. Optimisation problems are just like they sound, think of something that needs optimising, such as your route to work, a flight plan, or even a healthcare treatment and optimise it. On an abstract level, the technique the team used works by letting successful algorithms to pass their characteristics on to future generations – in short, each successive generation gets better and better at whatever tasks they've been assigned. However, coming back into the present day, the researchers took these algorithms and reworked them so they'd work better with today's deep neural networks and run better on large scale distributed computing systems.

Because the model was trained to be generative, it was also able to output reviews with preset sentiments. The table below is pulled from the paper and shows a random assortment of examples for both positive and negative reviews. These results are cool, but if you're totally new to this, let's take a few steps back. Even before machine learning, engineers interested in classifying sentiment would employ relatively dumb heuristics like keyword search to get the job done. However, with these methods, a sentence like, "I hope you're happy," could easily be misinterpreted as having a positive connotation simply because it possesses the word happy.

Our system beats other approaches on Stanford Sentiment Treebank while using dramatically less data. The number of labeled examples it takes two variants of our model (the green and blue lines) to match fully supervised approaches, each trained with 6,920 examples (the dashed gray lines). Our L1-regularized model (pretrained in an unsupervised fashion on Amazon reviews) matches multichannel CNN performance with only 11 labeled examples, and state-of-the-art CT-LSTM Ensembles with 232 examples. We were very surprised that our model learned an interpretable feature, and that simply predicting the next character in Amazon reviews resulted in discovering the concept of sentiment. We believe the phenomenon is not specific to our model, but is instead a general property of certain large neural networks that are trained to predict the next step or dimension in their inputs.

You might think, based on the type of research they're doing, that the OpenAI office would be full of gadgets, full of wonder, full of weird experiments. There are no Faraday cages. Well, okay, there is a robot. And it's tucked away in a side room. It's surrounded by cobbled-together protective material so that it doesn't smash into itself if it starts flailing about due to a programming error.

AI research has a long history of repurposing old ideas that have gone out of style. Now researchers at Elon Musk's open source AI project have revisited "neuroevolution," a field that has been around since the 1980s, and achieved state-of-the-art results. The group, led by OpenAI's research director Ilya Sutskever, has been exploring the use of a subset of algorithms from this field, called "evolution strategies," which are aimed at solving optimization problems. Despite the name, the approach is only loosely linked to biological evolution, the researchers say in a blog post announcing their results. On an abstract level, it relies on allowing successful individuals to pass on their characteristics to future generations.

In 2013 a British artificial-intelligence startup called DeepMind surprised computer scientists by showing off software that could learn to play classic Atari games better than an expert human player. DeepMind was soon acquired by Google, and the technique that beat the Atari games, reinforcement learning, has become a hot topic in the field of AI and robotics. Google used reinforcement learning to create software that beat a champion Go player last year. Now OpenAI, a nonprofit research institute cofounded and funded by Elon Musk, says it has discovered that an easier-to-use alternative to reinforcement learning can get rival results when it plays games and performs other tasks. At MIT Technology Review's EmTech Digital conference in San Francisco on Monday, OpenAI's research director, Ilya Sutskever, said that could allow researchers to make progress in machine learning faster.

Ilya Sutskever, director of OpenAI, an independent research group, will describe what might be the next big breakthrough in artificial intelligence today at EmTech Digital, a conference organized by MIT Technology Review in San Francisco. Sutskever will describe research showing an approach in machine learning that can perform even better than methods that have produced huge breakthroughs recently. His technique may also prove far more scalable. In a blog post describing the work, Sutskever and colleagues describe using "evolutionary strategies" to have machines figure out for themselves how to solve a complex task. The researchers say the approach is distantly related to a decades-old approach that involves optimizing algorithms using a process of simulated evolution.

One thing you won't hear him championing is the unfettered rise of artificial intelligence, which he once described as the "biggest existential threat" to humankind. Musk's prejudice prompted him to donate millions to the ethics think tank OpenAI--and it's why he's urging other billionaire techies like Facebook's Mark Zuckerberg and Alphabet's Larry Page to proceed with caution on their myriad of machine learning and robotics experiments. OpenAI is both an ethics and a research institution. Its mandate (plucked from its website): "Because of AI's surprising history, it's hard to predict when human-level AI might come within reach. When it does, it'll be important to have a leading research institution which can prioritize a good outcome for all over its own self-interest."

Elon Musk is a well-known harbinger of the potential for ill held by artificial intelligence. The Tesla and SpaceX CEO also helped start OpenAI, a group with a broad mandate that focuses on developing AI out (as the name implies) in the open, rather than behind closed doors as the exclusive province of high-powered governments and secretive private contractors. Musk, it turns out, was in on the AI train early with an investment in DeepMind, which was later acquired by Google. Musk wasn't in DeepMind for a return, as is the case with most investments; he wanted access to greater insight regarding DeepMind's progress, and the progress of AI in general, according to a new feature in Vanity Fair. The enterprising CEO wanted to be able to see how fast AI was improving, and what he found was a rate of gains that he hadn't expected, and that he thought most people would not possibly expect. This was the insight that Musk needed to begin a campaign warning against the potential dangers of AI, and to develop his own efforts to responsibility develop the tech via OpenAI.

Even more amazing, the researchers never explicitly programmed this AI communication. Instead, it "evolved" as a response to a reinforcement learning problem. While the jargon can get a bit technical, the OpenAI blog does a decent job of parsing it. The important thing to grok is the language was never defined, but rather hit upon as a solution to a general problem of learning to communicate. This type of AI method is called reinforcement learning, and involves the use of a reward signal to continually guide the agent towards an optimum outcome.