Welcome to Nural's newsletter where we explore how AI is being used to tackle global grand challenges.

As always in the newsletter you will find a compilation of articles, news and cool companies all focusing on using AI to tackle global grand challenges.

Packed inside we have

  • Stanford's state of the art prediction of RNA structures with only 18 training examples
  • Facebook AI announces Neural Databases, queries on unstructured data
  • London joins other cities for ethical AI
  • and ... a flying guide dog

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

Graham Lane & Marcel Hedman

Key Recent Developments

Stanford machine learning algorithm predicts biological structures more accurately than ever

AI algorithm solves structural biology challenges | Stanford News
Stanford researchers develop machine learning methods that accurately predict the 3D shapes of drug targets and other important biological molecules, even when only limited data is available.

What: A new paper describes outstanding results applying novel machine learning techniques in the notoriously difficult area of predicting the 3D structure of RNA molecules. This enables scientists to understand how different molecules work, supporting fundamental research and enabling the design of new molecules and medicines.

Key Takeaways: The most remarkable aspect of this work is that the model was trained with only 18 known RNA structures and without any pre-conceived notions about the structure of the molecules. This contrasts with established models that require huge training sets. The inputs to the neural network are simply the 3D coordinates and chemical element type of each atom. The approach developed in the paper may have broader application in other areas where data is scarce.

Cities worldwide band together to push for ethical AI

Cities worldwide band together to push for ethical AI
The chief digital and technology officers of London and Barcelona speak to Computer Weekly about their joint initiative launched with other cities to promote the ethical deployment of artificial intelligence in urban spaces

What: London, Barcelona and Amsterdam have recently launched a Global Observatory on Urban AI. This is part of the broader Cities Coalition for Digital Rights. London’s Chief Digital Officer is developing an “emerging technology charter for London”.

Key Takeaways: Cities are often at the cutting edge of managing contested issues concerning AI, while lacking formal legislative or regulatory powers. An example is the use in the public realm of facial recognition technology by public bodies and private companies. A concerted approach by large cities could lead to new models of implementing and using AI, particularly with regard to the “equality duty” of the public sector.

Facebook AI Introduces ‘Neural Databases’

Facebook AI Introduces ‘Neural Databases’, A New Approach Which Enables Machines to Search Unstructured Data and Connect The Fields of Databases and NLP
Facebook AI Introduces ‘Neural Databases’, A New Approach Which Enables Machines to Search Unstructured Data and Connect The Fields of Databases and NLP

What: Traditional "data databases are essential components of nearly every computer program and online service. However, they can be rigid structures that constrain how the data could actually be used." Facebook AI have recently announced a new approach to data management called neural databases, bridging the gap between databases and Natural Language Processing. This enables searching vast collections of unstructured data - from text to recordings of songs.

Key Takeaway: The ability to perform searches without data needing to be in a pre-set form opens up the possibility to create extremely complex queries (e,g, searching “what is the third-longest entry about a Russian novelist on Wikipedia?”). However, some have pointed to a high degree of hand-crafting in the paper, so it is uncertain how far we truly are from the widespread use of this technique.

Using a drone as a flying guide dog ...


AI Ethics

🚀 Even experts are too quick to rely on AI explanations, study finds

Participants placed too much trust in numeric explanations and might "find explanatory value where designers never intended”.

🚀 Now that machines can learn, can they unlearn?

What happens if you withdraw consent for your data to be used in a dataset that is used to train a model?

🚀 Increasing the presence of Black people in the field of artificial intelligence

"Black in AI works for presence and inclusion by creating space for sharing ideas, fostering collaborations, mentorship and advocacy."

Other interesting reads

🚀 Climate Change Innovation Grants

Grants up to USD 150K for research projects using AI or ML to address climate change mitigation, adaptation, or climate science

New UK Information Commissioner with brief to balance "protecting rights" against "innovation and economic growth".

🚀 I went to a play written by AI; it was like looking in a circus mirror

“Ironically, for a performance about the incredible capacity of technology, AI drew an unusual amount of attention towards the people [...] behind the scenes.”

Cool companies found this week

A mature AI landscape will require supporting services in areas such as law and insurance ...

bnh.ai - claims to be the only law firm in the world that is jointly run by lawyers and data scientists, focused on helping clients avoid, detect and respond to the liabilities of AI and analytics.

Koop - offers end-to-end, integrated insurance solutions for AV developers, robotics manufacturers, fleet owner-operators, technology vendors, and service providers

Avinew - provides policy discounts for drivers who engage autonomous or semi-autonomous driving features and use them responsibly.

Following the dancing Tesla "robot" last week, we were reminded of the launch of Windows 95 all those years ago ...

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!