1. Inside Uber ATG’s data-mining operation
How many ways can a pedestrian cross the street? Turns out, quite a lot. And if you’re designing self-driving vehicles, knowing that is pretty important. Here’s how Uber Advanced Technologies Group’s machine learning teams predict possible real-world outcomes related to a pedestrian’s decision to cross the road.
|
|
2. Walmart employees are out to show its anti-theft AI doesn’t work
A group of Walmart employees said they were “past their breaking point” with Everseen, an anti-theft technology Walmart began using in 2017. The employees claim it misidentifies innocuous behavior as theft and often fails to stop actual instances of stealing. They say they’re dismayed that their employer—one of the largest retailers in the world—was relying on AI they believe is flawed.
|
|
3. Is facial recognition in the US dead (or just on hold)?
Amid the police reform protests, IBM has announced that the company won’t develop, research, or offer facial recognition technology. Amazon announced a one-year moratorium on facial recognition. And Microsoft announced that it will not sell the controversial technology to police departments until there’s a federal law regulating it. But neither Microsoft nor Amazon said whether the new policies would bar other government agencies, such as US Immigration and Customs Enforcement, from deploying their facial recognition technology. And none of these companies are major players in facial recognition for law enforcement. Some of the largest providers—NEC Corp., IDEMIA, and Clearview AI—have not joined in on the voluntary moratoriums.
+ “The Activist Dismantling Racist Police Algorithms”
+ “Black Lives Matter Could Change Facial Recognition Forever—If Big Tech Doesn’t Stand in the Way.”
+ “Facial Recognition Is Accurate if You’re a White Guy.”
+ John Oliver covers facial recognition
|
|
4. Facebook released a dataset of 100,000 deepfakes
Facebook wants to use AI to help fight back against AI-generated fake videos. So the company released a dataset of over 100,000 clips using 3,426 actors and a range of existing face-swapping techniques to help train AI to spot manipulated videos.
|
|
5. The dumb reason your AI project will fail
Harvard Business Review looks at a common reason great proof of concepts fail: the inability to integrate AI models into a company’s overall technology architecture.
|
|
|
Machine learning in production
There’s a scarcity of resources on ML models in production. Hannes Hapke collects some of the best, showing you how to get started with this urgent task in this O’Reilly expert playlist.
|
|
|
6. A new framework for distributed reinforcement learning
This publication explains DeepMind’s Acme—“a framework for building readable, efficient, research-oriented RL algorithms...designed to enable simple descriptions of RL agents that can be run at various scales of execution—including distributed agents. By releasing Acme, [the campany’s] aim is to make the results of various RL algorithms developed in academia and industrial labs easier to reproduce and extend for the machine learning community at large.”
+ The Acme GitHub repository
|
|
7. How to detect unwanted bias in machine learning models
This post looks at detecting and mitigating unwanted bias in machine learning models, guidelines, examples of AI bias arising from both model choice and underlying societal bias, business and technical practices to detect and mitigate biased AI, and legal obligations as they currently exist under the GDPR.
|
|
8. Cortana got a promotion
“Once upon a time, Cortana was going to be Microsoft’s answer to Siri and Alexa. The dad jokes are still there, but Cortana has a new job these days, and it’s all tied to Microsoft 365.”
|
|
|