A Blog by Jonathan Low


Mar 21, 2018

An Emotion-Tracking Artificial Intelligence For Drivers In Autonomous Vehicles

It's when there are no longer drivers, but only passengers, in autonomous vehicles that the data capture gets really interesting. JL

Khari Johnson reports in Venture Beat:

An MIT Media Lab spinoff, (this) is Affectiva’s third service for tracking the emotional response of product users. It’s a  strategy to build emotional profiles drawn from smart speakers, autonomous vehicles, and other platforms with a video camera or conversational interface. In semi-autonomous vehicles, Affectiva will monitor drivers to increase safety and facilitate the handoff between human and machine. Once fully autonomous vehicles hit the road, Affectiva could switch its focus to passengers, using voice and facial recognition to personalize things inside a vehicle.

AI System Recreates Image You're Thinking About By Decoding Brain Signals

To the AI algorithm, it's all just data. JL

Tim Collins reports in The Mail:

Experts used a neural network to create images based on information taken from fMRI scans, which detect changes in blood flow to analyse electrical activity. The machine was able to reconstruct owls, aircraft, stained-glass windows and red postboxes after three volunteers stared at the pictures. 'While our model was trained with natural images, our method successfully generalized the reconstruction to artificial shapes, indicating that our model 'reconstructs' images from brain activity, not simply matches to exemplars.'

European Union Proposes Tough Online Revenue Tax For Big Tech Firms

The bill is unlikely to pass in its current form because smaller countries like Ireland, Estonia, Latvia and Lithuania stand to lose the advantage they have gained as tech havens.

But the intent is clear and it is prudent to assume that both higher taxes and greater regulation are coming, the question being not if, but when and how. JL

Philip Blenkinsop reports in Reuters:

Companies with significant digital revenues in Europe will pay a 3% tax on their turnover on various online services in the European Union, bringing in an estimated 5 billion euros ($6.1 billion). Google, Facebook and Amazon foot a large chunk of any bill. The tax, designed as a short-term measure before the EU finds a more comprehensive way to tax profits based on where they do business, could also encompass other high-profile U.S. firms such as Airbnb and Uber

How Mapping Apps Can Actually Make Traffic Worse

It's similar to what happened when investment firms began using algorithmic models: there is an early user advantage, but as more people apply them, widespread usage cancels out individual benefits. JL

Alexis Madrigal reports in The Atlantic, photo in Reuters by Bret Hartman:

A few people using route-planning makes things better, but a lot of people using them might force deterioration of driving conditions. As more users become informed through apps, benefits disappear. In some scenarios, traffic-beating apps might work for an individual, but make congestion worse overall. The flow of a freeway changes in response to an accident when there are more app-using drivers: congestion builds up at off-ramps, creating more traffic. Autonomous vehicles could deepen the problem.

An Artificial Intelligence Tool Makes Job Descriptions More Inclusive

In a war for talent, every advantage helps. JL

Fast Company reports:

A writing-enhancement service powered by a data set of 350 million job posts uses AI to analyze job descriptions, highlighting jargon and words that come across as particularly masculine or feminine. It predicts how different people will respond to content and suggests alternatives. Johnson & Johnson reported an additional 90,000 female applicants (a 9% increase) after using (it) to refine its job postings. Nvidia fills jobs twice as quickly, and Evernote’s apply-through rate on online job boards tripled.

Why Facebook's Real Problem Is It's Business Model: Will Data Regulation Result?

The real issue is not Cambridge Analytica's sleazy use of Facebook-generated personal data (to say nothing of the fact that its now-deposed chief thought it was ok to brag about it and Facebook's leaders have offered only defensively tone-deaf excuses).

The problem for a data-centric economy is that the real and perceived breaches for which Facebook has been routinely accused and for which it has paid fines over time may - finally - have created a wave of consumer/voter/government opposition to the core of the business model: trading free but eminently monetizable personal information for web service access. An entire economic system has been built around this.

The looming question is to what degree that trust-based model - and the extraordinary margins it provided - may now be at risk. JL

Will Oremus reports in Slate, Roger McNamee and Sandy Parakilas report in The Guardian:

It isn’t just that Facebook was careless with its users’data or that its policy was cavalier and misguided. It’s that Facebook is the chief architect of the socio-commercial arrangement by which people offer their personal information in exchange for free online services. Facebook isn’t just the source of the data that Cambridge Analytica used. It’s the reason this sort of data exists in the first place. What Cambridge Analytica did was what Facebook was optimized for.

Mar 20, 2018

Lyft and Uber Apps Generated 65% More Rides Than Taxis In New York City

Behavioral economics. JL

Johana Bhuiyan reports in Re/code:

By December 2017, drivers using ride-hail apps — Uber, Lyft, Gett, Via and Juno — performed 65% more rides per month than taxi drivers in New York City. Ride-hail apps are filling some of the gaps in the outer boroughs of New York City such as Queens and Brooklyn, 10 times more rides in the outer boroughs than taxis. Even in Manhattan, where yellow cabs have reigned, ride-hail apps were close to surpassing taxis.