Inside | Real news, curated by real humans
Inside AI

Inside AI (Aug 21st, 2019)

1. During the Hot Chips conference on Tuesday, Tesla chip designers showed how the company's AI chips are dramatically better in performance compared to the earlier Nvidia chips. The fine-tuned AI chips - which have 6 billion transistors apiece - are "smart" enough to power Tesla's full self-driving abilities in the future, according to the company. Their performance has improved by a factor of 21, compared to the earlier Nvidia chips. Ganesh Venkataramanan, one of the chip designers and a former AMD processor engineer, said that in order to meet "performance levels at the power constraints and the form factor constraints we had, we had to design something of our own." The chips, optimized for self-driving cars, run at 2GHz and perform 36 trillion operations a second. - CNET

  • Email gray
  • Permalink gray

2. Google Brain project founder Andrew Ng announced that he opened a second office for two of his projects - Deeplearning.ai and Landing.ai - in Medellin, Colombia. While it may seem like an odd choice, according to TechCrunch, Medellin is safer than it has been in years past and has been a hub for startups, thanks to incubators such as Ruta N. Ng, the former chief scientist at Baidu and CEO of Coursera, said he sees "early signs of momentum" that will make Colombia "a talent magnet both regionally and globally." Ng, who hired people from Poland, Egypt, Chile, and other locations for the new offices, says it's important to create AI hubs outside of China and Silicon Valley because, in part, "they’ll provide a different perspective." He plans to expand the AI Medellin team to between 150 and 200 employees within the next two years. - TECHCRUNCH

  • Email gray
  • Permalink gray

3. Presidential candidate and Sen. Bernie Sanders says that he will ban police from using facial recognition software if elected. Sanders (D-Vermont) also called for a moratorium on the criminal justice system using algorithmic risk assessment tools to try and predict which criminals will re-offend, saying that “we must ensure these tools do not have any implicit biases that lead to unjust or excessive sentences.” Sanders says he wouldn't allow the tools to be used until they passed an audit. While Sanders is the first presidential candidate to voice support for an outright ban on facial recognition use by police, Sens. Kamala Harris (D-CA) and Cory Booker (D-NJ) has also expressed concerns about the technology, including its racial bias. - CNN BUSINESS

  • Email gray
  • Permalink gray

4. Google is using AI to make smarter spell check and grammar suggestions in Gmail. The feature, which will only be available for G-Suite customers at first, is rolling out over the next several weeks and resembles the Gmail autocomplete function. The AI autocorrects emails while the person types, and also offers suggestions for grammar mistakes, highlighting them with a blue line. In a blog post, Google wrote that the features "can also help you write and edit with more confidence if you’re a non-native speaker," noting that the "AI-first approach" helps people "communicate smarter and faster, without sweating the small stuff.” - MASHABLE

  • Email gray
  • Permalink gray

5. The co-founder of DeepMind, Mustafa Suleyman, was placed on leave without explanation. Suleyman, who has acted as a public face for Google's AI lab, "is taking time out right now after 10 hectic years,” a DeepMind spokeswoman said, without saying why he is going on leave. Suleyman, who oversees DeepMind’s “applied” division for practical uses in the AI lab, has faced controversy over the lab's work in the U.K. health sector, including receiving access to 1.6 million patient records illegally. He also served on a group that formed after the controversy surrounding Google's AI contract with the Pentagon. - BLOOMBERG

  • Email gray
  • Permalink gray

6. Researchers trained a machine-learning algorithm to autonomously distinguish between two types of bright (but similar-looking) galaxies: those that are merging together and those that are birthing large numbers of stars. Both these look very similar in telescopic images, appearing as bright lights in the very-distance universe. As described in the journal Monthly Notices of the Royal Astronomical Society, the researchers wanted to distinguish between the two types of galaxies, so they created fake images - based on what each looks like up close - and blurred them to better resemble what they look like in a telescope. They then trained an AI to distinguish between images of the two types. They hope to build a much larger database to further improve the algorithm. - LIVE SCIENCE

  • Email gray
  • Permalink gray

7. An editorial piece in the journal Nature argues that France and Canada's international AI ethics panel needs to be independent of outside groups. The opinion, which was not attributed to a specific writer, notes that governments have generally lagged when it comes to prioritizing the ethics surrounding AI. Amid this so-called leadership vacuum, Canada and France plan to launch an International Panel on Artificial Intelligence at the G7 summit in Biarritz, France, which takes place Aug. 24-26. The panel, which aims to create a network of experts to advise governments on AI issues, "should be supported and shielded from undue influence," the journal argues. - NATURE

  • Email gray
  • Permalink gray

8. Australia's Little Ripper Group is using its AI-based drone tech to monitor crocodiles in the northeast state of Queensland. Little Ripper and the University of Technology Sydney designed the technology, which can differentiate among 16 different types of marine life. The AI drone system includes an in-built siren and speaker system and flotation pod deployment to help rescue people during emergencies or alert them about crocs or sharks. More than 50 of the drones were deployed around Australia last summer. - ZDNET

  • Email gray
  • Permalink gray

9. An AI developed by the Defense Intelligence Agency can help map the production-and-distribution networks of opioids in the U.S. Brian Drake, DIA’s director of artificial intelligence for future capabilities and innovations, developed a program called SABLE SPEAR that uses more than 20 data sources to draw connections via AI, allowing researchers to better understand the network of fentanyl and other synthetic drugs. These include a billion satellite images, 8 million cargo receipts, and various data from 43 million websites. - DEFENSE ONE

  • Email gray
  • Permalink gray

10. The AI-focused unicorn UiPath Inc. opens new offices in Houston on Thursday. The New York-based AI and robotic process automation startup — which was valued at over $1 billion — has 5,187 square feet of space at the downtown Main&Co. mixed-use facility. It has 71 employees in Houston, with plans to hire more. - HOUSTON BIZ JOURNAL

  • Email gray
  • Permalink gray

Written and curated by Beth Duckett in Orange County. Beth is a former reporter for The Arizona Republic who has written for USA Today, Get Out magazine and other publications. Follow her tweets about breaking news and other topics in southern California here.

Editor: Kim Lyons (Pittsburgh-based journalist and managing editor at Inside).

  • Email gray
  • Permalink gray

Subscribe to Inside AI