Data Science Insider: March 11th, 2022

In This Week’s SuperDataScience Newsletter: Transformers and the future of AI. AI could Decipher Gaps in Ancient Greek Texts. First Recall of Autonomous Driving System. Deep Learning with Python. Neural Network Used to Decode Pigs’ Grunts. Cheers, - The SuperDataScience Team P.S. Have friends and colleagues who could benefit from these weekly updates? Send them to [this link]( to subscribe to the Data Science Insider. --------------------------------------------------------------- [Transformers and the future of AI]( brief: A transformer is a DL model and type of neural network architecture that has gained significant popularity since its first appearance in a 2017 paper entitled ‘Attention Is All You Need.’ The model is primarily used in the fields of NLP and computer vision and adopts the mechanism of self-attention, meaning that the transformer runs processes so that every element in the input data unites (or in other words, pays attention) to every other element. This means that as soon as the transformer begins to train it can see traces of the entire data set, distinctively weighting the significance of each part of the input data. This is in contrast to other methods of AI where local patches of input data are the initial focus before being built up to the whole. This article by Quanta examines transformers in depth before questioning whether they are likely to overtake AI. Why this is important: This article offers a fascinating perspective, arguing that transformers began as a simple algorithm and citing their evolution from NLP to more recent advances into CV. As data scientists, we must be keen to explore the possibilities for transformers and how they may help progress the AI of the future. [Click here to sign up!]( [AI could Decipher Gaps in Ancient Greek Texts]( brief: Linguists from Ca’ Foscari University of Venice and Harvard University have published new research which claims that AI can help unlock the secrets of the ancient Greeks by filling in gaps in inscriptions and pinpointing when and where they are from. The research paper entitled ‘Restoring and attributing ancient texts using deep neural networks’ was published in Nature and describes how the team has built an AI system, nicknamed Ithaca, into which they’ve fed more than 63,000 transcribed ancient Greek inscriptions. This allows Ithaca to identify patterns in the order of letters and words. It was then able to accurately suggest when and where another 7,811 inscriptions were from, before going on to suggest a selection of letters and words to fill in artificially created gaps in the inscriptions. These were then ranked by probability. The scientists claim that Ithaca is 62% accurate when restoring letters in damaged text. Why this is important: Historian and ML expert Dr Thea Sommerschield, who co-authored the paper has summed up the significance by stating: “Inscriptions are really important because they are direct sources of evidence ... written directly by ancient people themselves.” [Click here to read on!]( [First Recall of Autonomous Driving System]( In brief: Autonomous vehicle startup Pony.ai has agreed to the world’s first automated driving software recall, following a crash in Fremont in October. The recall affects three vehicles and comes after the company admitted that faults with the car’s AI system led to a collision with a lane divider and street sign. Despite nobody being injured in the crash, in December, California’s Department of Motor Vehicles withdrew the company's permit to operate robo-cars. This was due to the fact that the car was in autonomous mode when the incidence occurred and no other vehicle was involved, raising questions about the product’s safety. This led to The National Highway Traffic Safety Administration (NHTSA) informing pony.ai that it believed their software had a safety defect. NHTSA Deputy Administrator Steven Cliff said: "Whether the vehicle is operated by a human driver or an automated driving system, the need to protect roadway users remains the same." Why this is important: This isn’t the first crash involving an autonomous vehicle but it is unique in the fact that it was in autonomous mode at the time. The recall and seriousness with which the NHTSA has taken the incident highlight the potential life and death nature of this technology. [Click here to discover more!]( [Deep Learning with Python]( In brief: This quick guide by GISuser offers an explanation of what deep learning is and then goes on to explore how Python, as a popular development platform, can be used to develop and train artificial neural networks through deep learning. It begins from the premise that it is generally easy to use the two together as long as you have a clear understanding of data and how to use python libraries. The article gives examples of the steps that you need to follow in order to develop a network model using deep learning with Python. Examples of these steps include: Import the required libraries, load the dataset, check the total number of training and testing samples, visualize the data, build the model by running your data set through the neural network’s existing framework, implement loss and optimization and test the model and improve accuracy. Why this is important: Deep learning is the ML technique behind the most exciting capabilities in diverse areas like robotics, NLP, image recognition, and AI. As the number one programming language python offers fantastic opportunities to expand your skillset and combine the two to impressive ends. [Click here to see the full picture!]( [Neural Network Used to Decode Pigs’ Grunts]( In brief: Scientists from the University of Copenhagen have revealed an AI translator which decodes the oinks, grunts, and squeals of pigs and turns them into recognisable emotions - fueling hopes of improvements to animal welfare. Researchers trained a neural network to record whether pigs were feeling positive emotions (like joy or excitement) based on situations such as huddling with littermates, suckling their mothers, running about and being reunited with the family; or negative emotions (such as fear or distress) based on scenarios such as piglet fights, crushing, castration and waiting in the abattoir - by using thousands of hours of recorded noises and data relating to behaviour. This information was gathered throughout the course of a swine’s life, taking in a variety of experiences and ultimately led to the acquisition of the acoustic signatures of 7,414 pig calls recorded from more than 400 animals. Why this is important: Having a greater understanding of the mental health needs of animals should be able to result in improved treatment for livestock. Previous research and animal welfare legislation has only concentrated on physical health but this kind of research may eventually lead to a more holistic approach. [Click here to find out more!]( [Super Data Science podcast]( this week's [Super Data Science Podcast](, data scientist, YouTuber, and golf lover Ken Jee joins us for a deep dive into the world of sports analytics and shares how he grew his large, online data science community. --------------------------------------------------------------- What is the Data Science Insider? This email is a briefing of the week's most disruptive, interesting, and useful resources curated by the SuperDataScience team for Data Scientists who want to take their careers to the next level. Want to take your data science skills to the next level? Check out the [SuperDataScience platform]( and sign up for membership today! Know someone who would benefit from getting The Data Science Insider? Send them [this link to sign up.]( # # If you wish to stop receiving our emails or change your subscription options, please [Manage Your Subscription]( SuperDataScience Pty Ltd (ABN 91 617 928 131), 15 Macleay Crescent, Pacific Paradise, QLD 4564, Australia