“Until people see robots going down the street killing people, they don’t know how to react because it seems so ethereal,” he said. “AI is a rare case where I think we need to be proactive in regulation instead of reactive. Because I think by the time we are reactive in AI regulation, it’s too late.”
There’s been a lot of talk recently about regulating AI, whether it be autonomous vehicles or more far-reaching regulations on AI in general, as MP Matt Hancock discussed at CogX. Although many agree with Musk’s position that regulation is critical, others (such as François Chollet, the creator of the deep neural net platform Keras) are more skeptical.
Tim Simonite, from Wired, discusses the problems that Apple is facing as it simultaneously commits itself to innovation in AI and to being uninterested in collecting user data.
The challenge, in a nutshell, is that it might be difficult for Apple to commit in the AI market if they remain to continue to eschew the cloud. “The iPhone is beefy as mobile device goes, and it’s a good bet Apple will add dedicated hardware to support machine learning. But it’s tough for anything it puts in your hand to compete with a server—particularly one using Google’s custom machine learning chip.”
Ian and Alexey, Google Brain researchers who are top-level influencers in machine learning, are having a session on Quora today (10AM PDT) and will take questions on adversarial machine learning research, their competition in Kaggle on the same topic as well as about Google Brain.
The in-house brand of subscription fashion startup Stitch Fix, Hybrid Designs employs a data science team that works with the company’s order data to predict which clothes customers will want to wear. The team identifies viable gaps in the company’s inventory—clothes that people would buy but a designer hasn’t made yet, says Stitch Fix chief algorithm officer Eric Colson.
“If we could be accurate enough to buy and hold inventory, could we be accurate in what isn’t available to buy, what doesn’t exist?” Colson asked. “Now when something is ostensibly missing from the market, we fill it in with our own algorithmically generated designs.” Take the algorithms tour on their site…really cool stuff!
Freeway merging in congested traffic is a significant challenge toward fully automated driving. Merging vehicles need to decide not only how to merge into a spot, but also where to merge.
In this paper, researchers from Toyota present a method for the freeway merging based on multi-policy decision making with a reinforcement learning method called passive actorcritic (pAC), which learns with less knowledge of the system and without active exploration. The method selects a merging spot candidate by using the state value learned with pAC. They evaluate their method using real traffic data. Their experiments show that pAC achieves 92% success rate to merge into a freeway, which is comparable to human decision making.
Senator Maria Cantwell, D-Wash., just drafted forward-looking legislation that aims to establish a select committee of experts to advise agencies across the government on the economic impact of federal artificial intelligence.
AI could open up opportunities for citizens to work and engage with government processes and policies in a way that has never been possible before. New AI tools that include voice-activated processes could make areas of government accessible to people with learning, hearing and sight impairments that previously wouldn’t have had the opportunity in the past. To this end, our Head of Product, Julian, designed an open source Alexa Skill for pension age calculation for DWP. He sees Alexa and brethren — voice only hands free computing — as a big opportunity to break help reduce digital exclusion.
Rob May, co-founder and CEO of Talla, is very skeptical about calling winners and losers in the AI race at this point. It’s like calling the winners of the Internet in 1996. You would have picked Yahoo to win, and Amazon not to, and you would have been wrong.
To explore how ML can learn subjective concepts, Google introduced an experimental deep-learning system for artistic content creation. It mimics the workflow of a professional photographer, roaming landscape panoramas from Google Street View and searching for the best composition, then carrying out various post-processing operations to create an
aesthetically pleasing image.
Their virtual photographer “travelled” ~40,000 panoramas in areas like the Alps, Banff and Jasper National Parks in Canada, Big Sur in California and Yellowstone National Park, and returned with creations that are quite impressive, some even approaching professional quality — as judged by professional photographers.