From McDonald’s automating its drive through-service to Github’s AI Copilot writing 30% of code on the platform – these are the 10 things happening in AI you need to know right now.
Facebook backing down on facial recognition
In the coming weeks, Facebooks parent company Meta will shut down the Face Recognition system on Facebook, the company writes on its blog. Meta wants to limit the use of facial recognition in its products. People who choose to opt-out for Face Recognition will no longer be automatically recognized in photos and videos. The facial recognition template used to identify them will be deleted.
According to Meta, the decision represents ‘one of the largest shifts in facial recognition usage in the technology’s history. Its removal will result in the deletion of more than a billion people’s individual facial recognition templates.’
Autonomous delivery startup Nuro raises $600 mln
Autonomous delivery vehicle maker Nuro raised 600 million dollars in its latest funding round this week, lifting its valuation to 8.6 billion dollars – from 5 billion dollar in the previous round. Investors include Google, Toyota and SoftBank. Nuro’s delivery unit R2 lacks seats and steering wheel as its sole purpose is delivering packages. Pizza company Domino’s was one of the first to adopt R2 as a pilot project to deliver their food autonomously to customers. The companies founders previously worked on Google’s self-driving car project, now known as Waymo.
McDonald’s partners with IBM to automate drive-through service
McDonald’s has partnered up with IBM to implement artificial intelligence at its drive-thru lanes. The fast food chain has previously bought the startup Apprente (and renamed it McD Tech Labs) to automate drive-through orders through voice recognition. McDonald’s claimed the system could already handle about 80 percent of the orders. In the latest deal, IBM will acquire McD Tech Labs. McDonald’s CEO Chris Kempczinski stated: “In my mind, IBM is the ideal partner for McDonald’s given their expertise in building AI-powered customer care solutions and voice recognition.”
Yuval Harari warns for massive data collection
Yuval Harari talked on the American television show 60 Minutes about the possible dangers of AI. The world-famous historian and author warned for a dystopian future where a handful of large corporations and governments can ‘hack humans’. Harari: “The world is increasingly kind of cut up into spheres of data collection, of data harvesting. In the Cold War, you had the Iron Curtain. Now we have the Silicon Curtain, that the world is increasingly divided between the USA and China. The question is: does your data go to California or does it go to Shenzhen and to Shanghai and to Beijing?”
Google introduces a new concept of AI called Pathways
Researchers at Google are building a new kind of AI framework, called Pathways, to handle many tasks at once and learn new tasks quickly. Today, machine learning models excel at a certain task, but often fail to scale or adapt to new situations. On its blog the company says that pathways will enable users to train a single model to do thousands or millions of things.
Google makes the comparison to a human learning to jump a rope: “Imagine if, every time you learned a new skill, you forgot everything you’d learned before – how to balance, how to leap. That’s more or less how we train most machine learning models today.”
Pathways combines its existing skills to learn new tasks faster and more effective, whether that is through text, images or audio. Take the word ‘leopard’ for example. Pathways will understand the concept of the animal as a written word, when someone says is as input or even when analysing a video. Impressive.
NATO releases its strategy for Artificial Intelligence
In the past weeks, defence ministers from various countries, agreed to NATO’s first-ever strategy for artificial intelligence. The organization pledges a billion dollars in funding for AI research and applications. The published manifesto serves as a framework for the ‘responsible use of AI technologies, in accordance with international law and NATO’s values’. It also addresses the threats posed by the use of AI by adversaries. NATO also wants to establish a trusted cooperation with the innovation community on AI.
Autonomous boat seen in Amsterdam Canals
After years of closed research, the atonomous boats are now being tested in the water of Amsterdam. The so-called RoBoats can sail without a driver and are intended to transport waste or even people. While the commercial operation of autonomous boats in the canals isn’t allowed at this point, developer AMS Institute thinks that will change in a few years time. The boats even have sensors to avoid recreational swimmers, engineer Rens Doornbusch tells local news network AT5.
AI maps all solar panels worldwide as seen from space
Researchers from the University of Oxford used computer vision technology and satellite images to map all the solar parks worldwide. They published their map and dataset of more than 68,000 existing solar parks in Nature on Wednesday. The researchers used photos from various satellites as a starting point. They trained a machine learning algorithm to see per pixel whether solar panels are present and how much energy they generate. The panels are recognizable by their color and reflection.The researchers had 550 terabytes of satellite images. It took the computer two months to process it, reports the Dutch newspaper NRC.
GitHub’s AI Copilot is helping write 30% of new code
The code hosting platform GitHub recently published its AI assistant for programmers called Copilot. After a few months in use, the service is being used a lot. According to GitHub 30% of all new code on its platform is written with the help of Copilot. “We hear a lot from our users that their coding practices have changed using Copilot,” he said. “Overall, they’re able to become much more productive in their coding.”
This AI tool creates a deepfake based on a single image
A new tool called Malivar.io can generate synthetic media – better known as deepfakes – based on a single image. You simply upload a photo and a video of another person, and the software will swap faces. The mission of Malivar.io is to democratize video content and synthetic human creation. The software has a free trial version. For longer video duration and 4k video the fees vary from 140 to 2100 dollars a month, which is quite steep. It is not yet clear who the target audience is for the deepfake solution – but a fun tool to experiment with for sure.