Welcome to Nural's newsletter where you will find a compilation of articles, news and cool companies, all focusing on how AI is being used to tackle global grand challenges.

Our aim is to make sure that you are always up to date with the most important developments in this fast-moving field.

Packed inside we have

  • DeepMind's model improves short-term forecasting of heavy rain
  • AI helps Greece screening travellers for asymptomatic Covid
  • plus, will Amazon Astro be your new Artificial Friend?

If you would like to support our continued work from £1 then click here!

Graham Lane & Marcel Hedman

Key Recent Developments

Nowcasting the next hour of rain

Our latest research and state-of-the-art model advances the science of Precipitation Nowcasting.

What: Deep Mind has developed a precipitation “nowcasting” model to predict the amount, timing and location of rainfall up to 90 minutes ahead. Twenty minutes of radar precipitation data are used to predict the next 90 mins. The model was assessed by 50 experts from the UK Met Office and rated their first choice in 89% of cases compared to other methods.

Key Takeaways: The research focuses on integrating AI into real-world operational decision-making for critical events, safety and planning guidance. Meteorologists rated the model highly due to the trade-off between accuracy, location, extent, motion and rainfall intensity. The model was described as “representing the risk best” while having “much higher detail” than alternatives.

Greece's machine-learning algorithm identifying COVID cases among travellers

A machine-learning algorithm to target COVID testing of travellers
Automated system allocates COVID tests to people arriving in Greece.

What: In 2020, Greece implemented a AI-based system to identify incoming travellers most likely to test positive for Covid-19. At that stage, there were only sufficient tests for 18% of all passengers. The model used information, such as age and gender, derived from the passenger locator form completed by all travellers and was continually updated with new test results. The system identified between 1.25 and 1.45 more infected travellers - many of them asymptomatic -  compared to testing based on country of origin.

Key Takeaway: The system was designed with careful regard to privacy and data protection concerns. It was implemented across all 40 entry points into the country. It provides an example of an AI system integrated into existing bureaucratic procedures and adding value.

Amazon announces Astro the home robot

Amazon announces Astro the home robot
Astro can patrol the home when you’re not there, or be remotely controlled.

What: Amazon has announced a home robot on wheels and with a screen. Astro can carry small items, recognise people and patrol the house monitoring for anything unusual. The response has been unenthusiastic for a gadget that appears rather pointless but gives rise to a range of privacy concerns.

Key Takeaways: Astro is impressive technology. The ability to map the home and move around while avoiding unexpected obstacles is impressive, as is the capacity to store and process data locally. These are areas of ongoing research and Amazon will doubtless collect valuable technical data along with information about unexpected applications and what users find useful. Amazon has the resources for a long game. It’s a bit like selling books on the web ...

AI Ethics

🚀 Are AI ethics teams doomed to be a facade? Women who pioneered them weigh in

Should ethics teams in large enterprises be independent and siloed, or closely integrated with other parts of the organization?

🚀 The ethical norms for the new generation artificial intelligence, China

The Chinese Government has published a set of ethical norms and fundamental requirements albeit human rights concerns persist.

🚀 Minority voices ‘filtered’ out of Google natural language processing models

Black, Hispanic and LGBTQ+ source material disproportionately filtered out of data used to train Google language models.

🚀 3 lessons from IBM on designing responsible, ethical AI

Key elements are: governance structure; open source toolkits to deliver ethical commitments; and multi-stakeholder partnerships.

Other interesting reads

🚀 Deep learning’s diminishing returns

“To halve the error rate, you can expect to need more than 500 times the computational resources.

🚀 October 19, 2021: Machine learning for climate science and Earth observation

Climate Change AI monthly webinar

🚀 Aging in an era of fake news

The paper cites evidence that older adults are more likely to circulate fake news, and investigates the nuanced reasons behind this.

Cool companies found this week

Data Labelling

Data labelling is typically a difficult, unglamourous, resource-intensive - yet crucial - step in many AI projects. There are crowd sourced solutions available, such as Amazon Mechanical Turk, but these can be of variable quality. Here are 3 startups working in this space.

Snorkel - specialises in AI-based, programmatic generation of data labels. Recently raised $85 million in round C funding

LabelBox - provides an integrated platform for large-scale labelling projects. Received $40 million in funding in early 2021

task.ai - offers distributed data labelling at scale. Has recently raised $4 million in seed funding

The future of art? Eight AI generated immersive experiences have been converted into non-fungible tokens and auctioned at Sotheby's

Sotheby’s immersive NFT installation auction kicks off | Reuters Video
Sothebys online auction of media artist Refik Anadols artificial intelligence art collection kicked off. This edit contains flash images.

AI/ML must knows

Foundation Models - any model trained on broad data at scale that can be fine-tuned to a wide range of downstream tasks. Examples include BERT and GPT-3. (See also Transfer Learning)
Few shot learning - Supervised learning using only a small dataset to master the task.
Transfer Learning - Reusing parts or all of a model designed for one task on a new task with the aim of reducing training time and improving performance.
Generative adversarial network - Generative models that create new data instances that resemble your training data. They can be used to generate fake images.
Deep Learning - Deep learning is a form of machine learning based on artificial neural networks.


Marcel Hedman
Nural Research Founder

If this has been interesting, share it with a friend who will find it equally valuable. If you are not already a subscriber, then subscribe here.

If you are enjoying this content and would like to support the work financially then you can amend your plan here from £1/month!