Data Science Insider: May 27th, 2022

In This Week’s SuperDataScience Newsletter: Google's New Text-to-Image AI Promises Photorealism. Clearview AI Fined for Illegally Storing Facial Images. Sociologists Claim that AI is Inventing Its Own Culture. Neuromorphic Chips and DL. Lessons from Minecraft’s Code-Writing AI. 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. --------------------------------------------------------------- [Google's New Text-to-Image AI Promises Photorealism]( brief: Google has unveiled a new AI text-to-image system that can create images based on text input, called Imagen. The diffusion model, created by the Brain Team at Google Research, offers "an unprecedented degree of photorealism and a deep level of language understanding." It is not yet being made available to the public as the AI is encoded with social biases. Google’s researchers summarise this problem in their paper, stating: “[T]he large scale data requirements of text-to-image models [...] have led researchers to rely heavily on large, mostly uncurated, web-scraped dataset [...] Dataset audits have revealed these datasets tend to reflect social stereotypes, oppressive viewpoints, and derogatory, or otherwise harmful, associations to marginalized identity groups.” The AI has the potential to be used to create deepfakes and hoaxes, alongside a tendency to create images of people with lighter skin tones and place them in certain stereotypical gender roles. Why this is important: As readers of this SuperDataScience newsletter will know, text-to-image programs are nothing new. Up until now the leader in the field has been DALL-E, a program created by commercial AI lab OpenAI (and updated just back in April). However, Google’s announcement of Imagen is notable due to the quality of images it produces. However, until issues of bias and nefarious applications are addressed, this will remain untested by the public. [Click here to sign up!]( [Clearview AI Fined for Illegally Storing Facial Images]( brief: The Information Commissioner’s Office (ICO) has fined Clearview AI more than £7.5 million ($9.4 million) and told them to delete the data of UK residents. The UK's privacy watchdog acted after the facial recognition company failed to follow the UK’s data protection laws. The company has fallen foul of the regulator by gathering images from the internet in order to create a global facial recognition database. The ICO claims that globally, more than 20 billion facial images have been obtained and stored. These images are taken by Clearview AI from pictures posted publicly on social media sites such as Facebook and Instagram, generally without the knowledge or permission of the platform. John Edwards, the UK’s information commissioner, released a statement, which claims: “The company not only enables identification of those people, but effectively monitors their behaviour and offers it as a commercial service. That is unacceptable.” Why this is important: Controversy surrounding Clearview AI and their failure to use data in a secure and legal way is an ongoing saga. This is the fourth time Clearview has been ordered to delete national data in this way, following similar orders and fines issued in Australia, France, and Italy. [Click here to read on!]( [AI is Inventing Its Own Culture]( In brief: A new study by researchers at the Center for Human and Machines at the Max Planck Institute for Human Development has found that humans have the ability to learn new things from AI systems and pass them to other humans. Furthermore, they found that these new ideas have the potential to spread in ways that could influence wider human culture. The study found that algorithms can teach humans how to best resolve problems but successful adoption of these solutions is curtailed by human biases, which means that people preferred suggestions from other humans. This issue prevents performance improvements from being as long-lasting as researchers would have expected. The study sought to explore the question: If social learning, or the ability of humans to learn from one another, forms the basis of how humans transmit culture or solve problems collectively, what would social learning look like between humans and algorithms? Why this is important: The suggestion that ML could influence human learning and culture is both incredibly interesting and incredibly scary. Particularly when you consider the fact that scientists are often unclear about how their algorithms actually work! [Click here to discover more!]( [Neuromorphic Chips and DL]( In brief: Semiconductors are known as the brains of electronics, but they have their limitations. As time has gone on, the power of these tiny silicon chips has grown exponentially, whilst their circuitry has shrunk to sizes so small that they’re almost impossible to comprehend. In fact, semiconductors have now reached the limits of miniaturisation and capacity, forcing a new approach to semiconductor design. A promising alternative that has already garnered much interest is neuromorphic computing, and a new study by the Institute of Theoretical Computer Science at the Graz University of Technology (TU Graz) in Austria has found that neuromorphic chips are significantly more energy-efficient at operating large DL networks than non-neuromorphic hardware. Neuromorphic computing works by mimicking the physics of the human brain and nervous system by establishing what is known as spiking neural networks and has the capacity to be rolled out in AI projects around the world. Why this is important: As power-hungry autonomous vehicles, robots, drones, and other self-reliant machines requiring small, energy-efficient chips become more prevalent, the need for hardware made specifically for DL tasks increases. While it may appear to be a no-brainer that specialised hardware would be more efficient at performing tasks, TU Graz claims this is the first time that this has been scientifically proved to be the case. [Click here to see the full picture!]( [Lessons from Minecraft’s Code-Writing AI]( In brief: Microsoft’s Build developer conference was held virtually this week, which offered company updates on upcoming releases on Microsoft Office, Windows, and more. A notable moment came when the company’s chief technology officer, Kevin Scott, demonstrated an AI helper for the game Minecraft. Although this may have initially appeared to be a fun demonstration of how characters in a computer game can be controlled by AI, this article by Wired does a great job of explaining how it in fact demonstrated the ML technology that Microsoft has been testing for auto-generating software code and potentially the future of computers. The Minecraft experiment was designed to showcase the recent advances in AI which have the potential to change the shape of personal computing. This would be most apparent through the replacing of interfaces that you tap, type, and click to navigate into with interfaces that you converse with. Why this is important: Scott used the conference to claim that in the coming years, there’s likely to be a slew of “AI-first” products. He claimed that code-writing AI “lets you think about doing software development in a different way—so you can express an intention for something that you want to accomplish.” [Click here to find out more!]( [Super Data Science podcast]( this week's [Super Data Science Podcast](, the co-founder behind Onfido, an AI-based ID verification, joins Jon Krohn to discuss his path to start-up success. Tune in to hear valuable information from Husayn Kassai. --------------------------------------------------------------- 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