Discover more from Last Week in AI

Weekly text and audio summaries of the most interesting AI news, as well as editorials commenting on recent events

Over 20,000 subscribers

AI for proposing math conjectures, building chatbots of the deceased, and more!

Last Week in AI #102

Last Week in AI

Feb 8, 2021

Black and white photo of Srinivasa Ramanujan. — Image Credits: Historic Collection/Alamy via Nature

Mini Briefs

AI maths whiz creates tough new problems for humans to solve

Researchers have developed an algoroithm that automatically proposes conjectures regarding how certain important mathmatical constants can be computed. Named after the famous Indian mathematician, the Ramanujan machine specifically proposes continued fractions, "infinite sequences of fractions nested in each other's denominators." Some of the proposed conjectures have latter been proven by researchers, showing the system can "make contributions to really hard problems." However, given the limited scope and capabilities of the algorithm, "calling this the ramanujan machine is over the top." Looking toward the future of algorithms in math, on mathematician comments:

Until I can detect a well-developed ‘sense of mathematical taste’ in AI, I expect its role to be that of an important auxiliary tool, not that of independent discoverer.

AI chat bots can bring you back from the dead, sort of

Microsoft recently submitted a patent about an AI chat bot system that is trained to replicate the communication styles of a specific person, dead or alive. More realistic virtual avatars are possible now due to the latest developments in language modeling, deep fakes, and voice synthesis. However, such a system would raise numerous ethical concerns, from privacy and identity-theft to interrupting people's grief process. In addition, of course such a chat bot, trained on text messages and social media data, is at best a crude approximation to how the actual person would respond, and it would say things that the person would have never said.

Big tech companies often submit "defensive" patents that they have no intention of implementing but would prevent lawsuits from patent trolls. This patent may be one of those, as Microsoft's general manager of AI saying "there's no plan for this," and calling the patent "disturbing." However, there's not much stopping these firms from pursuing such products, and Microsoft is not the only company who has a patent relevant to building conversational AI's in the style of a specific person.

Podcast

Check out our weekly podcast covering these stories!
Website | RSS | iTunes | Spotify | YouTube

News

Advances & Business

Microsoft launches Custom Neural Voice in limited access - "Microsoft today announced the general availability of Custom Neural Voice, an Azure Cognitive Services product that lets developers create synthetic voices with neural text-to-speech technology."
Human brain cells on microchips aim to "push boundaries of AI" - "An international team led by scientists at Aston University hope that combining the power and adaptability of the human brain with traditional electronics will bring about a revolution in computing and produce supercharged AI. "
As Robots Fill the Workplace, They Must Learn to Get Along - "So many robots work at Changi General Hospital in Singapore that until recently it wasn't uncommon to find two delivery bots sitting in a hallway or outside an elevator in a standoff."
AI in health care creates unique data challenges - "The health care industry produces an enormous amount of data. An IDC study estimates the volume of health data created annually, which hit over 2,000 exabytes in 2020, will continue to grow at a 48% rate year over year. "
Baltimore startup Melalogic wins grant from Mozilla fund exploring the effects of AI on racial justice - "Melalogic, a Baltimore startup building an app for Black skin health, won a grant from a fund created by software community Mozilla to support Black artists examining AI. The app that seeks to create a database of Black skin issues to combat racial bias in AI and medicine."

Concerns & Hype

This is how we lost control of our faces - "The largest ever study of facial-recognition data shows how much the rise of deep learning has fueled a loss of privacy."
Clearview AI's Facial Recognition App Called Illegal in Canada - "Canadian authorities declared that the company needed citizens' consent to use their biometric information, and told the firm to delete facial images from its database."
Here's a Way to Learn if Facial Recognition Systems Used Your Photos - "When tech companies created the facial recognition systems that are rapidly remaking government surveillance and chipping away at personal privacy, they may have received help from an unexpected source: your face."
Predictive policing is still racist'whatever data it uses - "It's no secret that predictive policing tools are racially biased. A number of studies have shown that racist feedback loops can arise if algorithms are trained on police data, such as arrests. But new research shows that training predictive tools in a way claimed to lessen bias has little effect."
Amazon faces spying claims over AI cameras in vans - "Amazon has begun using AI-powered cameras in delivery vans that constantly record footage of drivers and upload any mistakes they make. The firm says it is an investment in safety but privacy campaigners described it as "surveillance"."
How Censorship Can Influence Artificial Intelligence - "Artificial intelligence is hardly confined by international borders, as businesses, universities, and governments tap a global pool of ideas, algorithms, and talent. Yet the AI programs that result from this global gold rush can still reflect deep cultural divides."
Two Google engineers resign over firing of AI ethics researcher Timnit Gebru - "An engineering director and a software developer have quit Alphabet Inc's Google over the dismissal of AI researcher Timnit Gebru, a sign of the ongoing conflicts at the search giant over diversity and ethics."

Analysis & Policy

Pentagon No. 2 pick calls for metrics to judge AI progress - "President Joe Biden's deputy secretary of defense nominee would establish measurements to judge how the Pentagon adopts artificial intelligence as it seeks to modernize for future battles. "
AI and the List of Dirty, Naughty, Obscene, and Otherwise Bad Words - "It started as a way to restrict autocompletes on Shutterstock. Now it grooms search suggestions on Slack and influences Google's artificial intelligence research."

Expert Opinions & Discussion within the field

Reinforcement Learning At Facebook with Jason Gauci - "Interview with a Facebook engineer on how the company uses Reinforcement Learning"

That's all for this week! If you are not subscribed and liked this, feel free to subscribe below!