Category: Technology

http://www.techradar.com/news/the-5-best-laptops-of-computex-2018

 In case you missed the news late last year, Ikea is teaming up with Sonos to create a new series of smart speakers that goes by the name SYMFONISK. 

Today, prototypes of the speakers were shown off at Ikea’s Democratic Design Day event in Sweden and they’re … well, interesting.

There will be a number of speakers in the range – which Ikea says will be more affordably priced than regular Sonos speakers – with the one pictured up above doubling as a shelf or hang beneath a kitchen cabinet without seeming out of place.

Ikea hasn’t nailed down a specific release date for the line-up (other than a vague “sometime after summer 2019” that it mentions in a blog post on its website) but claims that the first of these prototypes is within earshot.  

The motivation behind SYMFONISK is to create a series of speakers that “save space, get rid of cords, make clutter invisible, and bring sound and music into the home in a more beautiful way” according to  Björn Block, Business Leader for Ikea Home Smart.

It’s worth noting that, because these are prototypes, the final design might change before they release sometime next year – which is good considering there’s one that looks like it could be the unwanted sequel to Nintendo’s Virtual Boy console

SYMFONISK’s debut follows yesterday’s announcement of the Sonos Beam, a slim smart soundbar that conveniently fits underneath your TV that’s developed exclusively by Sonos that comes out on July 17.

A speaker for every room

While it’s still early days for the SYMFONISK, Ikea certainly has a creative idea driving the innovation. That idea, partnered with the design company’s penchant for fitting furniture together in a utilitarian way, bodes well for the final product.

Beyond the addition of Sonos system compatibility, Ikea says that the SYMFONISK will also support its line of Trådfris smart home line of lights and switches.  

The SYMFONISK is the second speaker collaboration for the Swedish furniture maker after its first project, the Eneby series of speakers, hit stores earlier this year.

Via The Verge

http://www.techradar.com/news/ikeas-smart-speaker-prototype-doubles-as-a-shelf

Ever since the internet turned Microsoft’s Tay racist and genocidal, forcing Microsoft to shut down the chatbot within 24 hours, we’ve known how susceptible artificial intelligence (AI) can be to turning evil.

To study how AI can become corrupted by biased data, the Massachusetts Institute of Technology (MIT) decided to intentionally turn its AI into a psychopath named Norman—a reference to the villain in Alfred Hitchcock’s Psycho. 

The MIT team pulled images and captions from an infamous subreddit dedicated to photos of death and violence, then incorporated the captions into its data of how to describe objects. 

When the team subsequently gave Norman some Rorschach inkblots to analyze, Norman’s responses—compared to a “standard image captioning neural network”—were incredibly alarming. 

Image 1 of 3

Image 2 of 3

Image 3 of 3

Bright flowers became the splatter from a gunshot victim. Where the control AI interpreted an open umbrella, Norman saw a wife screaming in grief as her husband died. 

The morbid nature of the AI’s assumptions blinded it from considering any other possibilities besides murder and pain. 

Protecting our AI children’s young minds

Of course, with the darkest corner of Reddit as its only data set for interpreting the inkblots, Norman was destined to become monstrous, especially compared to an AI exposed to a controlled image set. 

Facebook's image-recognition AI uses Instagram captions to interpret objects

Facebook’s image-recognition AI uses Instagram captions to interpret objects

But MIT’s extreme example could hint at how any AI could have its interpretive criteria corrupted, depending on where it pulls data from. 

AI for image or facial recognition are being relied upon more and more by tech companies and various organizations. Facebook, for example, uses Instagram captions to teach its AI how to interpret images. 

If a subset of Instagram users began using racial slurs or sexist language to describe their photos, then Facebook’s AI could internalize these biases. 

Norman’s creators invited people to provide their own interpretations of the inkblots in a Google Doc, and “help Norman fix itself”. 

Their research could help future AI creators determine how to balance out against potentially pernicious data sources, and ensure that their AI creations remain as untainted and impartial as possible. 

http://www.techradar.com/news/mit-creates-psychopath-ai-using-the-dark-side-of-reddit