AI News Roundup —October 2019
by Gabriella Runnels and Macon McLean
The Opex AI Roundup provides you with our take on the coolest and most interesting Artificial Intelligence (AI) news and developments each month. Stay tuned and feel free to comment with any stories you think we missed!
_________________________________________________________________
Smart Compose, Grammarly, and AI Writing
Have you seen an ad for Grammarly lately? Have you used Smart Compose, the relatively new Gmail feature that’s designed to freak you out by correctly guessing what you’re gonna type next? Did you hear about GPT2, the language model that was so powerful that it was only partially released to the public for fear of misuse? Then you’re aware of the power of natural language processing, a field that mixes computer science, linguistics, and other domains that focuses on the digital analysis of natural language. This excellent piece by The New Yorker discusses some essential concepts in and the current state of natural language processing (including a fun feature that shows the text that GPT2 predicts should follow each section of the article).
Rubik’s Robot
OpenAI, the AI research firm behind GPT2 (the landmark language model mentioned above), has hit the headlines again with their most recent development in AI: training a robotic hand to solve a Rubik’s Cube using deep reinforcement learning. To do so, they used a technique called Automatic Domain Randomization (ADR), a process that changes the environment in which the problem is solved in unexpected ways (e.g., changing the size of the Rubik’s Cube, or tying two of the robotic hand’s fingers together) after a performance plateau. However, some observers caution that this is less of a step forward than it seems, as the research primarily concerned teaching the hand to manipulate the cube efficiently, not how to solve a Rubik’s Cube in a novel way.
The World Is Turning Turtle
In our March roundup, we discussed how the US Department of Defense recently made some major investments in AI. Officials in the US and beyond view AI as “the next big military advantage,” and global powers like the US, China, and Russia are racing to be the world’s foremost superpower in the field. Recent developments, however, suggest that the way to win the AI race isn’t necessarily to have the best AI algorithms: the most effective way to get an edge over the competition is to figure out how to thwart their machine learning algorithms through minuscule alterations of relevant data. In one particularly bizarre example, such an AI was fooled into mistaking a turtle for a rifle. With national security and human lives at stake, these AI systems shouldn’t be deployed if they can be so easily deceived.
Try, Learn, Repeat
In the 21st century, most children around the world are exposed to some degree of formal schooling. One researcher who studies hunter-gatherer communities in the Congo, however, says that children who haven’t been exposed to formal education learn new concepts in ways that are remarkably similar to how “new-generation machine-learning technologies” learn as well. For example, children in these communities are often given the materials they need to complete a new task, but they aren’t given explicit information about the best way to solve the problem at hand. Instead, they learn what works and what doesn’t through trial and error — and this is directly comparable to the processes that machine learning algorithms use to take in feedback and adjust their behavior accordingly.
Not Weird at All
In classrooms across China, students as young as five don headbands and wristbands equipped with biosensors to measure their focus, participation, and performance in school. Currently, hundreds of millions of students are being monitored by the Chinese government, and their data is being collected to fuel AI algorithms. The headbands, used to actively monitor each student’s attentiveness, were actually developed by a Massachusetts-based startup company. While many see this use of technology on young students as invasive, such unprecedented access to personal data helps China in their quest for global AI dominance.
That’s it for this month! In case you missed it, here’s last month’s roundup with even more cool AI news. Check back in November for more of the most interesting developments in the AI community (from our point of view, of course).
_________________________________________________________________
If you liked this blog post, check out more of our work, follow us on social media (Twitter, LinkedIn, and Facebook), or join us for our free monthly Academy webinars.
