'Ex Machina', here we come: A new algorithm helps computers learn the way we do

In a 'visual Turing test,' human judges couldn't tell them apart

Machine learning is all about getting computers to "understand" new concepts, but it's still a pretty inefficient process, often requiring hundreds of examples for training. That may soon change, however, thanks to new research published on Friday.

Aiming to shorten the learning process and make it more like the way humans acquire and apply new knowledge based on just a few examples, a team of researchers has developed what they call a Bayesian Program Learning framework and then used it to teach computers to identify and reproduce handwritten characters based on just a single example.

Whereas standard pattern-recognition algorithms represent concepts as configurations of pixels or collections of features, the BPL approach learns by “explaining” the data provided to the algorithm -- in this case, the sample character. Concepts are represented as probabilistic computer programs and the algorithm essentially programs itself by constructing code to produce the letter it sees. It can also capture variations in the way different people draw a given letter.

The model also “learns to learn” by using knowledge from previous concepts to speed learning on new ones, so it can use knowledge of the Latin alphabet to learn letters in the Greek alphabet more quickly, for example.

Most compelling of all is that the algorithm allowed computers to pass a sort of "visual Turing test." Specifically, the researchers asked both humans and computers to reproduce a series of handwritten characters after being shown just a single example of each; in some cases, subjects were asked to create entirely new characters in the style of those originally shown. Bottom line: human judges couldn't tell the results apart.

The researchers have applied their model to more than 1,600 types of handwritten characters in 50 writing systems, including Sanskrit, Tibetan, Gujarati and Glagolitic. They even tried it on invented characters such as those from the television series "Futurama."

A paper describing the research was published Friday in the journal Science. Its authors were Brenden Lake, a Moore-Sloan Data Science Fellow at New York University; Ruslan Salakhutdinov, an assistant professor of Computer Science at the University of Toronto; and Joshua Tenenbaum, a professor at MIT in the Department of Brain and Cognitive Sciences and the Center for Brains, Minds and Machines.

“It has been very difficult to build machines that require as little data as humans when learning a new concept,” said Salakhutdinov. “Replicating these abilities is an exciting area of research connecting machine learning, statistics, computer vision, and cognitive science.”

Join the newsletter!

Or

Sign up to gain exclusive access to email subscriptions, event invitations, competitions, giveaways, and much more.

Membership is free, and your security and privacy remain protected. View our privacy policy before signing up.

Error: Please check your email address.
Show Comments

Latest Videos

Conversations over a cuppa with CMO: Microsoft's Pip Arthur

​In this latest episode of our conversations over a cuppa with CMO, we catch up with the delightful Pip Arthur, Microsoft Australia's chief marketing officer and communications director, to talk about thinking differently, delivering on B2B connection in the crisis, brand purpose and marketing transformation.

More Videos

Dear buyer/ buyer Mandate,We are direct end seller refinery company for the supply of Petroleum products.We have all available in tank, J...

Russian oil & gas trading llc

Starwood VP of marketing: Managing customer expectations requires emotional credit

Read more

Having an effective Point of Sale system implemented in your retail store can streamline the transactions and data management activities....

Sheetal Kamble

​Jurlique’s move to mobile POS set to enhance customer experience

Read more

Thanks for another excellent post. Where else could anybody get that type of info in such an ideal way of writing? In my opinion, my seek...

Stephan jordan

6 Mobile Marketing Trends to Leverage in 2014

Read more

Hey there! Very interesting article, thank you for your input! I found particularly interesting the part where you mentioned that certain...

Martin Valovič

Companies don’t have policies to disrupt traditional business models: Forrester’s McQuivey

Read more

I too am regularly surprised at how little care a large swathe of consumers take over the sharing and use of their personal data. As a m...

Catherine Stenson

Have customers really changed? - Marketing edge - CMO Australia

Read more

Blog Posts

Highlights of 2020 deliver necessity for Circular Economies

The lessons emerging from a year like 2020 are what make the highlights, not necessarily what we gained. One of these is renewed emphasis on sustainability, and by this, I mean complete circular sustainability.

Katja Forbes

Managing director of Designit, Australia and New Zealand

Have customers really changed?

The past 12 months have been a confronting time for marketers, with each week seemingly bringing a new challenge. Some of the more notable impacts have been customer-centric, driven by shifting priorities, new consumption habits and expectation transfer.

Emilie Tan

Marketing strategist, Alpha Digital

Cultivating engaging content in Account-based Marketing (ABM)

ABM has been the buzzword in digital marketing for a while now, but I feel many companies are yet to really harness its power. The most important elements of ABM are to: Identify the right accounts; listen to these tracked accounts; and hyper-personalise your content to these accounts to truly engage them. It’s this third step where most companies struggle.

Joana Inch

Co-founder and head of digital, Hat Media Australia

Sign in