AI and Archaeology

Welcome to the 21st edition of the AI and Global Grand Challenges newsletter, where we explore how AI is tackling the largest problems facing the world.

The aim: To inspire AI action by builders, regulators, leaders, researchers and those interested in the field.

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

----

Packed inside

• Archaeologists trained a neural network to sort pottery fragments for them
• The creation of fake realistic fingerprints using GANs
• OpenAI investing $100m into AI companies looking to create large disruption
• and more...

This week's newsletter was written on my birthday! Either take that as a sign of my commitment to our AI learning or poor planning :D

Make sure to subscribe to our site as well: www.nural.cc

Marcel

__________________________________

Source: DeepLearning.AI

Key recent developments

---

Archaeologists train a neural network to sort pottery fragments for them

What: Two researchers have trained a convolutional network (CNN) to correctly recognise and classify pottery fragments from different eras. The network was compared to four experts and managed to outperform two of the experts and tied with the other two.

Key Takeaway: The ability to correctly classify pieces of pottery to the correct time period using machine learning will save countless hours for archaeologists. It is great example of using AI in towards a new application area and bring increased efficiency while maintaining performance levels.

---

Google announces health tool to identify skin conditions

What: Google have built a web tool "that uses artificial intelligence to help people identify skin, hair, or nail conditions". In testing, the tool identified the correct condition in the top three suggestions 84 percent of the time. It included the correct condition as one of the possible issues 97 percent of the time.

Key Takeaway: The tool is not designed to diagnose the skin conditions on its own but could prove the be a valuable tool for remote doctor consultations. It will be interesting to see how they continue to develop as the tool received a Class I medical device mark in the European Union, designating it as a low-risk medical device.

---

High Fidelity Fingerprint Generation: Quality, Uniqueness, and Privacy

What: Researchers have used GANs, the networks behind deepfakes, to create 50k synthetic (fake) fingerprints which are "unique, diverse, and resemble the training dataset in terms of minutiae configuration and quality, while not revealing the underlying identities of the training data."

Key Takeaway: The generation of synthetic data based on real source data has great implications to preserve privacy while allowing greater work on the data that has then been generated. What implications do you think we should be considering when creating unique fingerprints?

Public Fingerprint Dataset

---

AI and Climate Change: The Promise, the Perils and Pillars for Action

__________________________________

AI Ethics

?? Amazon extends moratorium [block] on police use of facial recognition software

?? NYC’s School Algorithms Cement Segregation. This Data Shows How

?? Report finds startling disinterest in ethical, responsible use of AI among business leaders

Other interesting reads

?? Security and privacy in machine learning wrap-up

?? Airbnb to offer AI tools to hosts to tackle property shortfall

?? These Ex-Journalists Are Using AI to Catch Online Defamation

?? This Is How A.I. Will Transform Medicine: The Same Way It Has Transformed Chess

Investment

?? OpenAI investing $100m in startups with big ideas about AI

__________________________________

Cool companies I have come across this week

Data Privacy

Tessian - Tessian’s Human Layer Security platform automatically stops data breaches caused by employees on email.

Climate

EarthNet2021 - A machine learning challenge and dataset for Earth surface and localized impact forecasting.

__________________________________

AI/ML must knows

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.

Tensorflow/keras/pytorch - Widely used machine learning frameworks

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.

Thanks for reading and I'll see you next week!

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

___________________________________

Marcel Hedman

Nural Research Founder

www.nural.cc

Marcel Hedman, is the current Choate Memorial Fellow at Harvard University focusing his efforts in data science. He also serves on the senior leadership team of the Future Leaders Network.

This newsletter is an extension of the work done by Nural Research, a group which explores AI use to inspire collaboration between those researching AI/ ML algorithms and those implementing them. Check out the website for more information www.nural.cc

Feel free to send comments, feedback and most importantly things you would like to see as part of this newsletter by getting in touch here.