1 – It’s Your Fault Microsoft’s Teen AI Turned Into Such a Jerk
Microsoft’s first open experiment with a chat bot in the US was brought to a halt after a series of its less-than-politically-correct statements leaked out on Twitter. The bot, Tay, was created to mimic the speech patterns of a “typical” 19-year-old American female, with data sourced from messaging platforms such as Twitter and GroupMe. While Microsoft’s intent was to “conduct research on conversational understanding,” the company realized its mistake a few hours after release, when the chat bot began making hateful tweets that involved topics from Trump’s stance on immigration to Hitler. Microsoft’s spokesperson commented that Tay was taken offline so that adjustments could be made, which may include the addition of a filter (if the chat bot even makes a reappearance). Tay, which learned through a combination of neural networks and input content from employees, was designed to adapt to individuals and their conversation patterns. The incident may reflect poorly on Microsoft, but it is also being touted as an unfortunate mirror of human beings’ imperfect behavior.
(Read the full article on Wired)
2 – Octopus Inspires AI Robots on a Mission
Scientists at Raytheon in Aurora, Colorado, are creating an AI that has distributed intelligence, similar to an octopus. While a human has one brain governing all of its actions, octopi have a central brain plus bundles of nerves in each leg that act as mini brains. The creature’s central brain “mediates” its actions; if one leg wants to go off in a particular direction, the central brain will rally the other legs to follow suit. The robotic system named Crowder is designed to work in a similar manner, though it looks very different – it’s a series of unconnected bug-like robots, each about the size of palm. Crowder is an ideal model for adaptable systems that can manage multiple machines at once, such as drones or autonomous vehicles. According to the robot’s makers, this includes giving the robots more autonomy, so that if a human controller loses contact with a drone, the machine can complete its mission independently.
(Read the full article on Discovery News)
3 – Google Patent Bid Reveals Self-Driving Car Decision-Making, and a Future of Self-Driving Boats, Golf Carts and Lawnmowers
Google’s patent application to the U.S. patent office indicate its interest in developing autonomous vehicles outside of automobiles, including trains, boats, lawn mowers, snowmobiles – just about any other vehicle that a person could name. Application contents also shed light on the controversial value-judgment AI systems that are meant to run the autonomous vehicles. The system acts on a “risk magnitude” scale, with various types of accidents and other negative events representing different values. Negative events also include the emotional reactions of humans, such as annoyance at sudden acceleration, which the vehicle would interpret and translate into a future “active-sensing action.” Additional details were also gleaned about the vehicles’ system of sensors, which basically allow the car to “see and hear” everything around it and to orient itself based on this input.
(Read the full article on Silicon Beat)
4 – Google Looking to Echo Amazon’s Voice-Controlled Echo Device
Google is looking into building a voice-controlled AI device similar to Amazon’s Echo, alongside claims that parent company Alphabet is looking to embrace a period of fiscal discipline. Details about the product are being kept secret, but reports have been made that it will answer questions and play music in similar fashion to Echo. Google’s Nest, a company acquired by Google in 2014, requested to be involved in the project but was denied by Google. While Nest has expanded from 300 to 1,000 employees since date of purchase, there are reports of internal unrest, possibly due to the unit’s stance to remain independent in many facets of its operations. Nest has been more financially conservative than many of its Google counterparts, turning down free food and continuing to rent servers from Amazon; however, Nest CEO Tony Fadell has a mixed reputation amongst employees, particularly those of other companies whom the company has acquired. Fadell has echoed that Alphabet approached all of its divisions to demand projected numbers and detailed business plans to ensure more efficient operations, a sign that the company is looking to cut back on unnecessary expenses.
(Read the full article on Silicon Valley Business Journal)
5 – Voice-Controlled Calorie Counter
MIT researchers presented a beta version of a voice-controlled, calorie-logging web system at this week’s International Conference on Acoustics, Speech, and Signal Processing in Shanghai. Users can verbally describe a meal, while the AI-run system organizes parts of the description into relevant nutritional data based on information available from the US Department of Agriculture (USDA). Users can also refine entries by verbal commands after information has been selected; for example, turning a whole banana into half of a banana. The speech-recognition system was designed specifically to respond to food-related terminology. Researchers used machine-learning algorithms to find patterns in “syntactic relationships” between similar but not identical words, such as ‘oats’ and ‘oatmeal’. The system is designed in response to the US’ obesity epidemic as a prevention tool for individuals trying to lose or maintain their weight.
(Read the full article on MIT News)
Image credit: Microsoft