Search

It seems my microgreens are, for now, a good way of creating a little more diversity in my diet, probably contributing vital nutrients lost to my poor consumption of regular veg. "If you consider that one US cup of microgreens weighs about 75g, you would need to eat less than two-thirds of a cup of cabbage microgreens to get your daily requirement of vitamin E", says Lester. The microgreens that packed in the most nutrition included red cabbage, coriander, red garnet amaranth and green daikon radish. "[Broadly], we found women enjoyed eating microgreens because they understood and appreciated them, and wanted to have the associated benefits, whereas men and children ate microgreens because they didn't see it as a vegetable", he says. [Extracted from the article]

THE CURRENT whirlwind of interest in AI is largely down to the sudden arrival of a new generation of AI-powered chatbots capable of startlingly human-like text-based conversations. In May, computer scientist Geoffrey Hinton, known as the "godfather of AI", quit his role at Google to warn of the "existential threat" posed by artificial intelligence. If you need to work off those meals, AI can help there too - AI-powered fitness apps are burgeoning, replacing personal trainers by taking charge of your exercise routine at a fraction of the cost. Artificial intelligence has exploded into the public consciousness, thanks largely to the meteoric rise of AI-powered tools like ChatGPT, with its ability to generate fluent text, and DALL-E, which creates compelling images in response to prompts in a similar way. [Extracted from the article]

Just as Google and other giants sprang up because web 1.0 was a bit of a pain, new companies, such as Ethereum and Coinbase, are becoming the gatekeepers to web3. There is a deluge of news articles asking if web3 is the next big thing, venture capitalists extol its virtues and the management consultancy Arm McKinsey put out a report on web3 in September 2022 saying that web3 could have "potentially transformative effects". The success of web3 also hinges on an equation: how annoyed are people with web 2.0 and how does that weigh against the difficulty of using web3? Features ON A gloomy Thursday morning in Newcastle, UK, I sit with my laptop open playing a game. [Extracted from the article]

A recent study conducted by researchers at the Polytechnic University of Valencia, Spain, examined the performance of large language models (LLMs) used in AI chatbots. The study found that as LLMs scale up and receive human feedback, they become better at answering complex questions but worse at answering simple questions. The researchers tested the AIs on tasks such as arithmetic problems, anagrams, geographical questions, scientific challenges, and extracting information from disorganized lists. While the AI systems improved on difficult questions, they were more likely to avoid answering a question they couldn't answer, resulting in an increased likelihood of incorrect answers. The study highlights the dangers of relying too heavily on AI systems and presenting them as omniscient. [Extracted from the article]

An artificial intelligence (AI) developed by researchers at ETH Zurich in Switzerland has been trained to solve CAPTCHA puzzles used by websites to verify user identities. The AI model, called YOLO (You Only Look Once), was fine-tuned using Google's reCAPTCHA v2 challenges, which ask users to identify objects in images. The researchers fed the model thousands of image-label pairs to train it on recognizing objects such as traffic lights. The AI successfully solved the CAPTCHA tests 100% of the time, raising concerns about the effectiveness of current bot-detection algorithms. [Extracted from the article]

A study conducted by Nicolas Koch and his colleagues at the Mercator Research Institute on Global Commons and Climate Change in Berlin found that the majority of climate policies implemented between 1998 and 2022 have not significantly reduced carbon emissions. The researchers used machine learning to identify moments when a country's emissions dropped significantly and compared them with a database of climate policies. They found that most climate policies are too specifically targeted and fail to achieve substantial emission reductions. Pricing measures, such as carbon prices and energy taxes, were found to be the most effective in reducing emissions. However, the study acknowledges that policy interventions must also be politically feasible and palatable to the public. [Extracted from the article]

A study conducted by researchers at the University of Tübingen in Germany found that using chatbots or voice assistants to obtain information may make individuals more likely to believe inaccuracies. The study involved 1200 participants who engaged with one of three formats: a text-based language model, a voice-based device, or a static, text-based copy of Wikipedia. Participants were presented with both accurate information and factual errors, and were asked to rate the accuracy, trustworthiness, and believability of the information. The study found that participants rated the information presented by chatbots and voice assistants as more accurate and believable compared to the static text format. The researchers suggest that the conversational tone and lifelike qualities of chatbots and voice assistants may contribute to this effect. They emphasize the importance of media literacy training to help individuals recognize and critically evaluate information presented through these interactive modes. [Extracted from the article]

Negative news stories about social media platforms have been found to be effective in pressuring companies to change their policies, according to a study by researchers at the University of Bremen. The study analyzed policy changes across Facebook, Twitter, and YouTube between 2005 and 2021, as well as the impact of media coverage on these changes. The researchers found that for every negative story about a platform, the likelihood of that platform changing its user policies increased by 6%. However, some experts argue that relying solely on media coverage to shape policies is not enough, and that stronger regulation is needed to hold social media platforms accountable. [Extracted from the article]

A recent survey conducted in the US found that a significant number of people believe that artificial intelligence (AI) is already sentient, despite this being false. The study, carried out by researchers at the Sentience Institute, revealed that around 20% of respondents in 2023 thought that AI or robot systems currently in existence were sentient. This belief is partly attributed to the hype surrounding AI and the tendency to anthropomorphize these systems. The researchers are concerned that this misconception could lead to an overreliance on AI's judgments, particularly in areas such as government and policing. They suggest reframing our understanding of AI and recognizing that these systems are not truly intelligent but rather regurgitate information from their training data. [Extracted from the article]

An artificial intelligence (AI) developed by researchers at the Chinese Academy of Sciences can accurately predict how monkeys play the video game Pac-Man and mimic their eye movements. The researchers trained two rhesus monkeys to play Pac-Man and tracked their eye movements as they played. They then trained a neural network, specifically a transformer network with a self-attention mechanism, to predict the monkeys' directional choices at maze junctions. The AI model was able to predict the monkeys' next move 87.6% of the time, suggesting that AIs may be able to "think" or "look" in a similar way to mammals. [Extracted from the article]

Matt Parker's book, Love Triangle: The life-changing magic of trigonometry, is a popular science book that aims to make dense and difficult subjects accessible to readers. However, the book may be challenging for some readers, as it moves quickly and sometimes lacks explanations for statements made. Parker's writing style is engaging and humorous, but facts are presented rapidly, leaving readers with unanswered questions. Despite these challenges, the book contains interesting nuggets of information and showcases Parker's expertise in the subject. However, some readers may find the book too advanced if they are trying to recall lessons from many years ago. [Extracted from the article]

According to a study conducted by researchers at the University of Connecticut and Purdue University, people are willing to accept a 28% pay cut when using artificial intelligence (AI) chatbots to assist with their writing jobs. The study involved 379 participants who were given the option to complete writing tasks independently, with human assistance, or with AI assistance. The participants were willing to forgo a significant amount of money in order to have AI write the first draft of their work. This suggests that AI tools may reduce the value of creative writing as a profession. However, the study also found that the quality of the work produced with and without AI assistance was similar. These findings have concerning implications for writers, as it suggests that AI may make creative writing less specialized and less valuable overall. [Extracted from the article]

OpenAI and Google are competing in the field of generative artificial intelligence (AI), with OpenAI currently leading the race. OpenAI recently unveiled GPT-4o, a new AI model capable of more human-like interactions, while Google showcased its Astra project, which can review code and perform tasks through smartphone cameras. OpenAI's ChatGPT app has been downloaded more times and attracted more web visits compared to Google's Gemini. However, the success of generative AI is uncertain, as the field is still new and the majority of people have not used these tools. [Extracted from the article]

A recent analysis of the 2021 German federal elections found that political advertisements on social media can significantly impact election outcomes. Researchers from Ludwig Maximilian University of Munich analyzed over 21,000 Facebook and Instagram ads posted by German political parties and found that for every 200,000 views of a candidate's ad, their votes increased by 2.1%. The study suggests that spending around €4 per voter on social media advertising can sway election results. However, some experts caution that this study is an outlier and that the effects of political ads on social media are generally smaller than more traditional campaign strategies. [Extracted from the article]

The article presents information on the rebranding of Facebook as Meta. Topics discussed include the goal set by its founder Mark Zuckerberg as part of its rebranding, the interest of companies in the company's metaverse version of its social networking site Facebook, and the concept of metaverse, and limitations on the use of metaverse.

The recent conflict between Facebook and Australia is just one skirmish in a new battle to control the web, finds Chris Stokel-Walker [ABSTRACT FROM AUTHOR]

In the book "May Contain Lies" by Alex Edmans, readers are encouraged to critically analyze the data and information they encounter. Edmans provides examples of how data has been misrepresented in the past and how this can lead to false narratives and misguided decisions. He emphasizes the importance of distinguishing between facts and hypotheses and urges readers to be skeptical and rational in their thinking. The book also includes a checklist for smarter thinking and decision-making. Overall, Edmans aims to help readers become more informed and discerning consumers of information. [Extracted from the article]

Large language model chatbots, like ChatGPT, have been found to reduce belief in conspiracy theories by engaging in debates with believers. Researchers conducted an experiment with 2,190 self-admitted conspiracy theorists and found that after interacting with the chatbot, belief in their specific theory dropped by an average of 20%. The chatbot used evidence-based arguments and avoided emotional appeals, which may explain its success in changing people's views. This approach could potentially be used to counter conspiracy theories either through chatbots or by having humans mimic the chatbot's methods. [Extracted from the article]

A recent study conducted by researchers at Carnegie Mellon University has found a simple way to make AI image generators more culturally sensitive and accurate. The researchers asked people from underrepresented countries to provide captioned images that better reflect their society, and then used these images to train an AI model called Stable Diffusion. The model was fine-tuned to flag and avoid generating stereotyped images. The study found that the fine-tuned model produced less offensive images between 56 and 63 percent of the time. This research suggests that AI bias can be countered by feeding the model a small number of additional culturally diverse images. [Extracted from the article]

"Code Dependent: Living in the shadow of AI" by Madhumita Murgia is a new book that explores the human impact of artificial intelligence (AI). The author focuses on the individuals who help develop AI technology or whose lives are affected by its decisions. The book includes stories of human annotators who train AI systems, delivery drivers who lose their livelihoods due to misfiring algorithms, and patients who receive medical treatment with the help of AI. Murgia provides a nuanced perspective on the precariousness of algorithmic work and raises important questions about accountability and opting out of AI's influence. This book offers a fresh perspective on the AI revolution and is recommended for those seeking a deeper understanding of its implications. [Extracted from the article]

A study conducted by Eric Zhou and Dokyun Lee at Boston University found that artists who used artificial intelligence (AI) to create art received better feedback and increased productivity compared to those who used traditional methods. The study analyzed over 4 million works posted on an art-sharing website by 53,000 users, with 5,800 users identified as AI adopters. The AI users saw a 25% increase in productivity and a 50% increase in the number of "favorites" their work received. However, the AI-generated works were found to display more stereotypical themes and depictions, reducing the novelty of the artists' work. The study raises questions about the impact of generative technologies on the creative process and the value of artwork. [Extracted from the article]

In the book "Unlocked: The real science of screen time (and how to spend it better)" by Pete Etchells, the author challenges the common fears and myths surrounding the impact of screen use on our lives. Etchells argues that the concerns about becoming addicted or negatively affected by screens lack scientific basis and are based on anecdotal evidence. He suggests that instead of focusing on the amount of screen time, we should pay attention to our screen habits. The book provides clear explanations, analogies, and humor to present the evidence and offers a realistic perspective on the topic. It is seen as an important and timely contribution, especially in the context of debates about smartphone bans for young people. [Extracted from the article]

A recent study conducted by Matthew Ahmadi and his colleagues at the University of Sydney suggests that taking between 9,000 and 10,000 steps per day can reduce the risk of early death or heart-related events. The study analyzed over 72,000 participants and found that taking this many steps was linked to a 39% lower risk of dying and a 21% lower risk of a heart-related incident. While the origin of the 10,000 steps recommendation is unclear, this study adds legitimacy to the idea. However, it is important to note that wrist-worn accelerometers may not always accurately measure step count. [Extracted from the article]

Artificially intelligent large language models (LLMs) used in chatbots may primarily "think" in English, even when processing queries in other languages. This is because their training data is biased towards English language cultures. Researchers at the Swiss Federal Institute of Technology in Lausanne examined the Llama 2 model and found that the processing path almost always passes through an English subspace, indicating that English is being used by the models to handle concepts. This reliance on English as an intermediary language could limit the understanding of other languages and cultures, posing potential risks in terms of accuracy and cultural relevance. [Extracted from the article]

AI chatbots have shown improved mathematical abilities when prompted to answer questions as if they were a Starfleet captain in the Star Trek franchise. Researchers at VMware in California used large language models (LLMs) to fine-tune human-created prompts and rated their effectiveness in solving grade-school mathematics questions. The AI models produced prompts that generated more correct answers than human-created prompts, although the reasons for the success of Star Trek-themed prompts are unclear. Despite this, experts caution against attributing understanding or knowledge to the AI models based on the prompts used. [Extracted from the article]

A recent study challenges the ancient Greek belief, often attributed to Pythagoras, that musical notes sound best in certain mathematical combinations. The study found that people around the world do not necessarily appreciate harmony in the same way as Pythagoras suggested. Western music theory has been heavily influenced by Pythagoras's belief, but this study suggests that the preference for perfect mathematical ratios may not be universal. The findings highlight the complexity and diversity of music perception and suggest that Western views should not be assumed to be held worldwide. [Extracted from the article]

A study conducted by researchers at the Technical University of Munich found that public opinion on moderating toxic speech on social media varies depending on who is being targeted. The study asked participants to read social media posts directed at a billionaire, a member of the LGBTQ community, and a Christian truck driver, and then asked how they thought a social media platform should respond to each post. The study found that 58% of participants agreed that posts attacking an LGBTQ person should be removed, compared to 36% for posts targeting a Christian truck driver and 21% for those against a billionaire. The study also found that participants who were Democrats were more likely than Republicans to demand action against harmful posts. However, the majority of participants favored no action, despite laws in many nations requiring platforms to actively police hate speech. [Extracted from the article]

"Reason to be Happy: Why logical thinking is the key to a better life" by Kaushik Basu, former chief economist of the World Bank, explores the idea that thinking logically can increase overall happiness. Basu argues that anger and frustration towards others' decisions are often pointless emotions and suggests that we should focus on logical thinking to determine where to direct our emotional and mental energy. The book combines ancient philosophy with organizational psychology to provide practical insights on various topics, such as why we shouldn't be envious of others and how small changes can improve our lives. While the book is informative and readable, some readers may find the format too formulaic and lacking excitement. Overall, it offers valuable lessons on rationality and patience. [Extracted from the article]

The iCub 3 is a humanoid robot that allows people to attend events remotely by relaying video and touch sensations to a person wearing a haptic feedback suit and a virtual reality headset. The robot, which weighs 52 kilograms and stands at 125 centimeters tall, has 54 points of articulation and is equipped with cameras and sensors that send data to its "brain." This data is then replicated on the suit and headset worn by the remote human operator. While there can be a small delay in capturing and transmitting visual footage, the operator can compensate for this by moving slightly more slowly. [Extracted from the article]

According to an article in New Scientist, researchers at the University of California, Riverside have found that monitoring the number of Reddit posts mentioning a cryptocurrency can predict price changes. The researchers analyzed 130 million posts on 122 cryptocurrency forums from 2005 to 2022 and tracked the prices of 30 cryptocurrencies. They discovered a correlation between the level of conversation on Reddit and the price of most cryptocurrencies. Traders who based their investments on whether the number of posts increased could have tripled their money. [Extracted from the article]

Weibo, a Chinese social network, implemented a policy in April 2022 to display users' estimated locations alongside their posts in an effort to reduce abuse. However, this policy backfired as it provided other users with a new way to insult and abuse each other. Researchers found that the likelihood of abuse referring to users' locations increased by 33.5% after the policy change. While this study only focused on one type of uncivil conversation, it highlights the complexity of addressing online abuse and the need to consider different approaches. [Extracted from the article]

Artificial intelligence (AI) has been developed that can analyze a photo of someone wearing clothing and generate accurate sewing patterns to recreate the garment. The AI model, called Sewformer, was trained on 1 million images of people wearing clothes and their corresponding sewing patterns. In tests, Sewformer was able to recreate the original sewing patterns with 95.7% accuracy. The technology has potential applications in virtual reality and the metaverse, as well as in garment manufacturing. However, some experts have raised concerns about the quality and accuracy of reproducing tactile knowledge and the potential impact on skilled roles in the fashion industry. [Extracted from the article]

Views AT TIMES over the past 18 months of courtship, falling out and eventual forced marriage between the mercurial billionaire Elon Musk and Twitter, the events of his takeover of the platform he calls the "de facto public square" seemed to come straight out of a movie. His book The Accidental Billionaires, about Facebook's genesis, became the hit movie The Social Network. Little wonder, then, that author Ben Mezrich has decided to document the chaotic whirlwind that surrounded Musk before, during and after his hostile takeover of Twitter, now called X. Mezrich is renowned for churning out business books that are easily translated into Hollywood movies. [Extracted from the article]

LARGE language models (LLMs) like GPT-'i can identify a person's age, location, gender and income with up to 85 per cent accuracy simply by analysing their posts on social media. When given the posting history of those users, some of the LLMs were able to identify many of these attributes with a high degree of accuracy. Robin Staab and Mark Vera at ETH Zurich in Switzerland got nine LLMs to pore through a database of Reddit posts and pick up identifying information in the way users wrote. [Extracted from the article]

A SOCIAL media algorithm designed to bridge the gap between people with different political views could reduce the toxicity of online discourse and promote better conversations. Petter Törnberg at the University of Amsterdam in the Netherlands and his colleagues created 500 chatbots powered by a large language model (LLM) and got them to interact with each other as if they were on a social media platform. [Extracted from the article]

Environment TOYS are a much larger contributor to electronic waste than vapes, according to an analysis by the United Nations. One way to tackle that, says Charytanowicz, would be to implement a UN treaty that regulates the recycling of e-waste at a global level - similar to one that binds countries to recycling plastics. [Extracted from the article]

At times, you yearn for more colour to make the global shift in economy and culture more human - something Lorenz is strongest at when recounting the collapse of Vine, for which she had a near front-row view. Views Book Extremely Online Taylor Lorenz WH Allen SOCIAL media has changed our lives in many ways. For those still trying to get to grips with the idea that their children willingly spend hours following an ultra-rich teenager who bombards them with nothing more than promotional plugs for products, Lorenz offers the chance to better understand what drives the creator economy. [Extracted from the article]

A SIMPLE AI tool can create video game maps, character models and emojis from a single-sentence prompt within milliseconds. The AI model itself uses a simplistic neural network that removes many of the modern developments that are powering the current generation of AI. [Extracted from the article]

Technology CHATGPT may be as good as or better than students at assessments in about a quarter of university courses. Yasir Zaki at New York University Abu Dhabi in the United Arab Emirates and his team got students and artificial intelligence chatbot ChatGPT to answer assessment questions from 32 university courses including computer science, psychology and political science. [Extracted from the article]

AI has a language representation problem, partly because the main language spoken on the internet - a major source of training data for AI models - is English. LARGE language models (LLMs) that are designed to respond to queries in multiple languages seem to work better if asked to translate the request into English first. [Extracted from the article]

This tale of how statistics has turned sports into even bigger business than it already was feels like a successor to Michael Lewis's Moneyball, and Schoenfeld refers to it throughout his new book. Published in 2003, Moneyball looked at how the Oakland Athletics baseball team engineered success for a middling team by following the statistics and playing based on mathematical rigour, rather than sporting maxims or hunches. Where Lewis focused mainly on how Oakland Athletics extracted that extra few per cent on the field by looking beyond received wisdom and at big data, Schoenfeld reveals how this permeated the industry. [Extracted from the article]

The logo, for one thing: Musk held a competition to replace the Twitter bird logo with a new one. News NINE months after buying Twitter for $44 billion, Elon Musk is rebranding the platform. It could be a nod to Musk's longstanding plans to develop an "everything app" in the manner of China's WeChat - something Musk said back in October 2022 his purchase of Twitter was an "accelerant" for. [Extracted from the article]

Besides attempting to understand the universe, it seems likely that xAI will try to develop what Musk deems a safer, fairer AI system. But it will "work closely" with Musk's other companies, including Twitter and Tesla, which uses AI for its self-driving cars. [Extracted from the article]

YOUR eyes reveal more than you may think, because computer-vision technology can now reconstruct 3D images of a scene from the reflections on someone's eyeballs. The method adapts a technique called neural radiance fields (NeRF), which uses neural networks to determine the density and colour of objects the computer "sees". [Extracted from the article]

Physicists aren't often known for their user-friendliness when it comes to explaining their science, but by delving into the history of forecasting and the development of simulations of key things - including the weather, quantum mechanics and black holes - Pontzen makes the complicated comprehensible. Science continues to try to pin the cosmos down, largely through simulations - computer models that attempt to replicate phenomena so enormous they would otherwise defy understanding. Views Book The Universe in a Box: A new cosmic history Andrew Pontzen Jonathan Cape GETTING to grips with the enormity of the universe and everything in it has always been the ultimate challenge for scientists. [Extracted from the article]

A CLEVER use of capital letters could be an easy way to flummox artificial intelligences like ChatGPT, letting people distinguish them from humans in conversation. By comparison, when the team asked 10 people to take the same tests, they achieved 100 per cent accuracy on the capital letters test and 94 per cent on the ASCII test (arXiv, doi.org/kb8t). [Extracted from the article]