Microsoft for the People
In recent events we’ve seen controversy in the Artificial Intelligence space, specifically regarding the topic of human rights. According to Reuters, Microsoft Corp rejected the request of an unnamed California law enforcement agency to install AI-powered facial recognition technology in the vehicles and body cameras of police officers. The agency wanted to run face scans during every traffic stop.
In a statement made by Microsoft President Brad Smith, the artificial intelligence behind the technology has been trained on mostly white and male photographs. This would create an impact that would result in minorities and females being frequently held for questioning, due to higher error rates within the facial-recognition technology, according to The Verge. Smith’s comments revealed that he is additionally concerned about the possibility for the future to look like a George Orwell 1984-esque society, as reported by CNet.
Last December, the company pledged to be open about the issues with their facial recognition software and warned of improper uses of it. Smith has called for further governance over the development and use of AI in order to avoid issues of human right violations and negative bias, according to CNet. Although Microsoft has nixed several deals for companies and cities to access this product, Microsoft sold the software to a United States prison based on conclusions that the technology would improve safety.
One Giant Leap
If you haven’t seen it yet, here it is: the first ever image of a black hole.
According to the BBC, a black hole is a region of space from which nothing, not even light, can escape. This is the reason for their lack of visibility and, consequently, their name.
Since 2012, 200 scientists have been working on the Event Horizon Telescope project, consisting of a network of eight telescopes spanning from Antarctica to Chile trying to directly observe the environment of a black hole, according to Reuters.
29-year old computer scientist Dr. Katie Bouman began working on this project as a grad student at MIT in the United States, reported by the BBC. She led the development of the algorithm that converted the raw data that was captured by the telescopes into a photograph that could be seen by the naked eye. Astronomers, engineers, and mathematicians were among the 200 professionals working with Dr. Bouman to do this.
This black hole is located at the center of the Messier 87, according to Reuters. It is over three million times the size of the Earth—larger than our entire Solar System— according to the proposer of the experiment, Professor Heino Falcke, from Radboud University in the Netherlands. In 1993 when Falcke was a PhD student, he realized that a certain type of radio emission that was likely to be located near a black hole was also powerful enough to be detected by telescopes on Earth.
Dr. Ziri Younsi, from the University College London and also part of the EHT team, told the BBC, “Although they are relatively simple objects, black holes raise some of the most complex questions about the nature of space and time, and ultimately of our existence.”
Are You There, Alexa?
Amazon has revealed some background regarding their “Alexa voice review process,” in which teams of people listen to and transcribe voice recordings captured by the Amazon Echo speaker in order to improve the understanding of human speech by Amazon’s digital assistant, Alexa. Thousands of people make up these teams, working in offices that span the globe.
According to Bloomberg, these employees are made up of contractors and full-time Amazon employees. They have signed nondisclosure agreements that prohibit them from speaking about their work, which can entail examining nearly a thousand audio clips per day. Internal chat rooms are used between employees to clear up confusion about a mumbled phrase or a funny audio clip that’s worth sharing. Sometimes criminal or abusive exchanges are overheard by these teams. Amazon has stated it has appropriate procedures in place to support to employees who are exposed to upsetting recordings.
Although Amazon provides its customers the option to disable the use of voice recordings for new feature development, the customer might still have their recordings heard and manually reviewed due to Alexa’s standard review process. Although the Echo speaker is meant to store audio only when the speaker’s activation button is pressed or a wake word is detected, according to Bloomberg, auditors each transcribe as many as 100 recordings a day that were not triggered by any sort of wake command.
An Amazon spokesperson said in an email to Bloomberg, “We only annotate an extremely small sample of Alexa voice recordings in order [to] improve the customer experience. For example, this information helps us train our speech recognition and natural language understanding systems, so Alexa can better understand your requests, and ensure the service works well for everyone…Employees do not have direct access to information that can identify the person or account as part of this workflow. All information is treated with high confidentiality and we use multi-factor authentication to restrict access, service encryption and audits of our control environment to protect it.”
While we knew much of this already, it’s important to not forget that there is still a large manual, human element at play in these “machine learning” devices.
If you enjoyed this post, check out more of our This Week in Tech News articles:
- This Week in Tech News: Our World in Data
- This Week in Tech News: Why Am I Seeing This?
- This Week in Tech News: A New Gadget Era
Body image credits:
- Event Horizon Telescope Collaboration [CC BY 4.0]
- Photo by Rahul Chakraborty on Unsplash
- Photo credit: Microsoft Corporation [Public domain]