CatAIstrophe

19

October

2018

5/5 (1)

Out of the many trends emerging from recent technological developments, few have spread as quickly and exponentially as social media. As of October 2018, Facebook -the titan of this industry- totaled 2.2 billion users worldwide, with its little brother Instagram rising to 1 billion (Statista, 2018). Also worth mentioning are WhatsApp and its Chinese equivalent WeChat, with respectively  1.5 and 1.1 billion users. These platforms have changed the way we interact, with users creating and broadcasting content from their own lives. The amount of data produced on Facebook alone is tremendous. As example, every 60 seconds, 510,000 comments are posted, 293,000 statuses are updated, and 136,000 photos are uploaded (Zephoria.com, 2018).

But what if this data could be used productively? Advancements in the field of AI, and especially machine learning, led to the emergence of deep learning. Thanks to its processing power, it allows the analysis of much bigger data samples than before (Brynjolfsson, 2017). Applications are endless, but a particularly interesting sector is public safety.

In spite of monitoring, surveillance, and predictions, all countries are subject one way or another to catastrophes, natural -as the recent earthquake in Indonesia- or man-made -like terrorist attack in western Europe. It takes time for necessaries parties -the police, firemen, ambulances etc- to be informed of the situation, choose the necessary response, and arrive on site to provide help.

On the other hand, information about such events is spread on social media within minutes. What some companies are suggesting is that we make use of this global sensor network of eyewitnesses (Dataminr.com, 2018). The idea is to use a system, which analyzes and cross-checks real-time social media data. It could revolutionize the way we use emergency services, by detecting the earliest indications of high-impact events and critical information, and by alerting both civilians and support services of hazards.

Of course, this system comes with questions regarding its use. Which data can be used for analysis -especially with GDPR in place? If the use is for public safety, does that mean getting a social media account should be mandatory? How can you be certain of the veracity of the data used?

 

 

References:

Brynjolfsson, E., and Mcafee, A. 2017. The business of artificial intelligence: what it can and cannot do for your organization. Harvard Business Review

Most popular social networks worldwide as of October 2018, r. (2018). Global social media ranking 2018 | Statistic. [online] Statista. Available at: https://www.statista.com/statistics/272014/global-social-networks-ranked-by-number-of-users/ [Accessed 19 Oct. 2018].

Dataminr.com. (2018). Dataminr for Corporate Security | Dataminr. [online] Available at: https://www.dataminr.com/corporate-security [Accessed 19 Oct. 2018].

Zephoria.com. (2018). [online] Available at: https://zephoria.com/top-15-valuable-facebook-statistics/ [Accessed 19 Oct. 2018].

Please rate this

Deepfakes give new meaning to fake news

12

September

2018

No ratings yet.

 Not everyone may be familiar with “Deepfake” technology, but it might be a game changer for so-called fake news. Using the technology, one is be able to make anyone say or do anything, or at least make it appear so. Hence, the technology might usher in a new era where establishing credibility will become even harder. 

The technology was once only available to researchers and academics but has been leaked by a reddit user around december 2017. Essentially the tech uses a form of artificial intelligence called deep learning to study footage of a real person. The algorithm can then allow a user to alter or add on to existing footage. The more footage that is fed to the learning algorithm, the more convincing the deep fake will be.

In a recent PSA, comedian Jordan Peele released a video showing Barack Obama insulting current US president Donald Trump in a rather childlike manner. However, the video (also shown above) is a complete fabrication, with Peele voicing the entire clip. The Peele PSA only took about 56 hours of Obama footage in order to convincingly produce the clip.

It is not surprising then that the deepfake tech is causing concern, with some NYU researchers calling it a “menace on the horizon”. Since the technology was leaked on reddit, it has mostly been used to create fake pornography images and clips with celebrities. The user has since been banned from reddit, however the tech is still available through downloadable apps.

Most importantly, the worry stems around the implications for the political landscape and discourse; it will be easy to falsify videos showing e.g. politicians saying things they have never said in real life or being implicated in corruption scandals. Likewise, it will be easy to discredit real footage as simply being the result of a deepfake.

Plenty of examples of deepfakes can already be found with a quick youtube search.

References

https://www.bloombergquint.com/topic/fake-videos-using-voice-clones#gs.RjVsecE

https://www.buzzfeednews.com/article/davidmack/obama-fake-news-jordan-peele-psa-video-buzzfeed#.ugOXGqvAn3

 

Please rate this