In the News
Training a single AI model can emit as much carbon as five cars in their lifetimes
“Neither I nor other researchers I’ve discussed them with thought the environmental impact was that substantial.” The paper on carbon footprint of natural-language processing specifically examines the model training process for natural-language processing (NLP). They found that the process can emit more than 626,000 pounds of carbon dioxide equivalent—nearly five times the lifetime emissions of the average American car
AI Ethics at War – When AI Governance Shifts from Cooperation to Competition
They share conferences, include each other in thought leadership, and develop white papers together – in order to discover frameworks for governing AI and handling issues around privacy, security, bias, and individual rights.
Robust.AI launches to build an industrial-grade cognitive platform for robots
New Silicon Valley robotics startup Robust.AI aims, firstly, to build the world’s first industrial-grade cognitive platform for robots.
Sponsor
We’ve done the heavy lifting so your product will scale by reaching the right buyers from the top brands in tech.
Gain invaluable access to our extensive email lists of leading members of the tech world. Our engaged audiences are made of 100K subscribers - of which 40%+ are executives from 10K companies in the US/UK, including most of the Forbes 500 companies.
Applied use cases
AI mavericks want to build a better brain for industrial robots
Some big names in AI and robotics are teaming up to develop a robot operating system that will aim to address the shortcomings of today’s smartest machines.
Does object recognition work for everyone? A new method to assess bias in CV systems
Using a publicly available third-party data set of photos of household items in 50 countries, we found accuracy for all these systems was indeed significantly lower for images from certain regions and from households with lower income levels.
1000x Faster Data Augmentation
In this blog post we introduce Population Based Augmentation (PBA), an algorithm that quickly and efficiently learns a state-of-the-art approach to augmenting data for neural network training.
Ethics
£18.5 million to boost diversity in AI tech roles and innovation in online training for adults
Up to £13.5 million new funding for an extra 2500 artificial intelligence and data science conversion degrees, with 1000 scholarships for people from underrepresented groups £5 million government fund to drive innovation in adult online learning.
OII London Lecture: A.I. in society – opportunities and risks
The Oxford Internet Institute is excited to present Professor Luciano Floridi, for the lecture "A.I.
11. Closure: on ethics, code and law
The study of ethics concerns a reflection on the justification (whether utilitarian or deontological) of decision-making that affects human agents and human societies, and/or the development of practical wisdom (virtue ethics).
Robotics
This Robot Ostrich Can Ride Around on Hovershoes
The Hybrid Robotics Lab has been working on robots that walk over challenging terrain—how do wheeled platforms like hovershoes fit in with that?
Do you have a robotics research lab that isn't on the map yet?
The Robot Report reports 26 funding deals closed in May 2019 for a total of at least $1.5 billion. 50% of these meetings led to pitches to individual partners.
Jaco Is a Low-Power Robot Arm That Hooks to Your Wheelchair
Can you tell us about the Jaco arm, and how the process of designing an assistive robot arm is different from the process of designing a conventional robot arm?
Research
CAVIA: Fast Context Adaptation via Meta-Learning
After meta-training, the model is evaluated on a new task: given a small set of labelled data points (in supervised learning) or trajectories (in reinforcement learning), the learned initial parameters are adapted using just a few gradient steps.
Weight Agnostic Neural Networks
As a proof of concept, we also apply our search method on a supervised learning domain, and find it can discover networks that, even without explicit weight training, can achieve a much higher than chance test accuracy of ∼ 92% on MNIST.
UCL NLP Inaugural Meeting
We would like to welcome you to an afternoon of invited keynote talks, presentations by members of the UCL NLP group, a poster session, and an evening social.