Data Science Insider: April 15th, 2022

In This Week’s SuperDataScience Newsletter: The Future of AI Protein-Folding. DALL-E Raises New Questions About Bias. A New Use for Facial Recognition Software. Understanding Transformers. Razer-Designed Linux Laptop targets ML Developers. 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. --------------------------------------------------------------- [The Future of AI Protein-Folding]( brief: As we here at SuperDataScience covered at the time, in July 2021 the British artificial intelligence subsidiary of Alphabet Inc - DeepMind - made an AI tool called AlphaFold2 publicly available. The software was able to predict the 3D shape of proteins from their genetic sequence. This enabled us to have 3-D structures for 98.5% of the human proteome, whereas we’d previously known the 3-D structures for only about 17% of the roughly 20,000 proteins in the human body. At the time Forbes claimed “AlphaFold is the most important achievement in AI—ever”, but now that the dust has settled, what does the future hold for AlphaFold and the AI protein-folding revolution? This article by Nature examines that very question and looks at how ‘AlphaFold mania’ has transformed the field and resulted in rocketing numbers of research papers being published. Why this is important: Often new discoveries result in a flurry of press coverage full of hyperbole about how the world will be changed. This article goes beyond the initial excitement and examines how the scientific community has actually been using AlphaFold2, what the future is likely to be, and examines what the real-world applications for the AI are. [Click here to sign up!]( [DALL-E Raises New Questions About Bias]( brief: This week saw the unveiling of a new AI system in pre-release from OpenAI, known as DALL-E. The technology has been lauded as having cracked the code on true AI-generated art as it has the ability to convert simple text prompts into digital illustrations in an array of styles. You may well have seen images created by DALL-E circulating the web, from a sea otter inspired by Johannes Vermeer’s ‘Girl with a Pearl Earring’ to teddy bears shopping for groceries in the style of Japanese Ukiyo-e prints. The AI has the ability to transform the art world but also reinforces stereotypes and bias. As the AI is trained by scraping off millions of images and their corresponding captions from the internet, it brings with it the biases that created the initial data. Examples of this include solely creating images of middle-aged white males when tasked with illustrating a lawyer. Why this is important: Researchers have attempted to address these issues. For example, they attempted to filter out sexual content from the training data as they were aware of the likelihood of it creating disproportionate harm to women. However, there is no easy solution; in this scenario, they found that when they tried to filter sexual content out, DALL-E 2 generated fewer images of women in general, leading to erasure. As data scientists, we must weigh up these issues when looking to produce technologies that will have inherent biases. [Click here to read on!]( [A New Use for Facial Recognition Software]( In brief: Facial Recognition software and particularly that of controversial facial recognition company, Clearview AI, has been a regular source of stories for these SuperDataScience newsletters. Usually, this has focused on its employment by law enforcement, data breaches, and associations with the far-right. However, since it gave access to its technology to the Ukrainian government last month it has been put to good, if sad, use: Identifying the dead. This rather harrowing article by the BBC highlights the way that facial recognition technology has been used to offer closure to families who have lost loved ones to the war with Russia. Clearview AI founder, Hoan Ton-That, has said that following Putin's invasion of Ukraine: "We saw images of people who were prisoners of war and fleeing situations, and you know, it got us thinking that this could potentially be a technology that could be useful for identification, and also verification." Why this is important: Using facial recognition to identify the dead is not new, and Clearview isn't the only platform being used in Ukraine. However, it is on a scale much larger than ever seen in active conflict before. [Click here to discover more!]( [Understanding Transformers]( In brief: Over the past two years, AI-based language processing programs have come on leaps and bounds, reaching a surprising level of linguistic fluency. This is primarily a result of architecture called the transformer, which was invented in 2017 and served as a blueprint for the programs which followed, in the form of a list of equations. Until now, this bare mathematical outline was the sum total of our understanding of how transformers work and what they are doing with the words they process. The prevailing theory has been that they somehow have the ability to pay attention to multiple words at once, allowing for a big picture analysis, but the understanding of exactly how this works has proved elusive. Recently, the company Anthropic published two research papers showing that they are starting to have an understanding of what transformers are doing when they process and generate text. Why this is important: In order to understand how transformers work, the researchers from Anthropic simplified the architecture. By stripping out all the neuron layers and all but one or two layers of attention heads, they were able to identify a link between transformers and simpler models that they already had a clear understanding of. This process shows us the value of using small versions in order to gain an understanding of more complex processes. [Click here to see the full picture!]( [Razer-Designed Linux Laptop targets ML Developers]( In brief: Razer has primarily been known as a PC maker with a fine line of sleek gaming laptops, but it has now joined forces with Lambda – a deep learning-focused company – in order to launch the new Lambda TensorBook, a laptop designed for DL developers and intended to supercharge ML work. The new laptop combines hardware from Razer with Lambda’s DL software, and it boasts some impressive figures when it comes to ML performance. The launch of the TensorBook has excited the industry as it stands apart from other ML laptops due to the fact that it comes with powerful hardware. This includes a dedicated Nvidia GeForce RTX 3080 Max-Q graphics processor. It also comes with Ubuntu Linux 20.04 LTS pre-installed, including the Lambda Stack deep learning suite and tools including PyTorch, Tensorflow, CUDA, and cuDNN, providing you with the tools you need for ML workloads. Why this is important: We don’t usually cover product launches but the launch of the TensorBook has much to offer the data scientist or anyone interested in ML work. The pre-loaded hardware and software specs make it a valuable tool for DL developers and equally as powerful for gaming. [Click here to find out more!]( [Super Data Science podcast]( this week's [Super Data Science Podcast](, Artificial General Intelligence (AGI) takes center stage as Jeremie Harris joins Jon Krohn to discuss AI safety, the development of AGI, and the existential risks it presents to humans. --------------------------------------------------------------- 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