Artificial intelligence Archives | Page 665 of 702

Examiners fooled as AI students outperform real students in the UK

In a groundbreaking study published in PLOS One, the University of Reading has unveiled startling findings from a real-world Turing test involving AI in university exams, raising profound implications for education.

The study, led by the university’s tech team, involved 33 fictitious student profiles using OpenAI’s GPT-4 to complete psychology assignments and exams online. Astonishingly, 94% of AI-generated submissions went undetected by examiners, outperforming their human counterparts by achieving higher grades on average.

Associate Professor Peter Scarfe, a co-author of the study, emphasised the urgent need for educational institutions to address the impact of AI on academic integrity. He highlighted a recent UNESCO survey revealing minimal global preparation for the use of generative AI in education, calling for a reassessment of assessment practices worldwide.

Professor Etienne Roesch, another co-author, underscored the importance of establishing clear guidelines on AI usage to maintain trust in educational assessments and beyond. She stressed the responsibility of both creators and consumers of information to uphold academic integrity amid AI advancements.

The study also pointed to ongoing challenges for educators in combating AI-driven academic misconduct, even as tools like Turnitin adapt to detect AI-authored work. Despite these challenges, educators like Professor Elizabeth McCrum, the University of Reading’s pro-vice chancellor of education, advocate for embracing AI as a tool for enhancing student learning and employability skills.

Looking ahead, Professor McCrum expressed confidence in the university’s proactive stance in integrating AI responsibly into educational practices, preparing students for a future shaped by rapid technological change.

Why does it matter?

The IATSE’s tentative agreement represents a significant step forward in securing fair wages and job protections for Hollywood’s behind-the-scenes workers, ensuring that the rapid advancements in technology do not come at the expense of human employment.

Instagram tests AI for creator interactions

Instagram is trialling a new feature called ‘AI Studio’, allowing creators to develop AI versions of themselves. Meta CEO Mark Zuckerberg recently revealed on his broadcast channel that the feature is undergoing an initial test phase with selected creators and users in the United States.

Zuckerberg highlighted that AI avatars from popular creators and interest-based AI models will soon appear in Instagram messaging. These AI entities are initially designed to interact within messaging threads and will be clearly marked as AI-generated.

During the broadcast, Zuckerberg demonstrated early examples featuring AI-powered chatbots developed in collaboration with creators such as the team behind the meme account ‘Wasted’ and Don Allen Stevenson III. These chatbots aim to assist creators by engaging with their followers and responding to messages on their behalf.

Creators on Instagram can initiate interactions by tapping the ‘Message’ button, prompting users to acknowledge that the responses may be AI-generated and potentially not entirely accurate or appropriate. Each AI-generated message will be prefaced with ‘AI’ and marked with a ‘beta’ tag, indicating ongoing development and testing.

Meta’s launch of AI Studio last year enabled businesses to create AI chatbots for platforms like Messenger, Facebook, and Instagram. The initiative reflects Meta’s ongoing efforts to integrate advanced AI technologies into its social media platforms, enhancing user engagement and interaction capabilities.

Why does it matter?

The rise of AI in financial services: balancing opportunities and challenges

According to industry executives, AI is increasingly seen as a game-changer in the financial services sector, offering significant opportunities for efficiency and client service enhancements. At a recent investor conference hosted by Morningstar in Chicago, discussions highlighted AI’s potential to improve tasks such as explaining portfolio allocations or lending decisions, often influenced by subconscious biases that are difficult for humans to articulate. Zack Kass, formerly with OpenAI, emphasised the potential of AI to enhance these processes, cautioning that careful implementation is crucial to avoid exacerbating existing challenges.

Investors and technology experts believe AI will streamline routine financial tasks, such as compliance form filling and portfolio development, freeing financial professionals to focus more on personalised client interactions and complex problem-solving. Morningstar’s senior research analyst Karen Zaya noted the growing prevalence of AI-powered chatbots in simpler tasks like customer service inquiries but highlighted the complexity of utilising AI for more nuanced financial planning decisions, such as retirement investments.

While AI adoption holds promise, US regulators are actively soliciting public input on its implications for financial inclusion and equity. Treasury Secretary Janet Yellen has cautioned that while AI could reduce transaction costs, it also introduces significant risks that need careful management by financial firms. Despite concerns about job displacement, experts like Margaret Vitrano from ClearBridge Investments argue that AI will likely complement rather than replace human expertise, particularly in software development and customer service roles requiring nuanced decision-making.

Financial advisers like Brenda Ingram in Chicago are optimistic about AI’s potential to streamline mundane tasks such as compliance reporting, anticipating that AI could enhance efficiency and reduce operational costs in the industry. As financial firms navigate the adoption of AI, the emphasis remains on thoughtful implementation to maximise benefits while mitigating risks, ensuring that AI enhances rather than disrupts the client-adviser relationship and operational workflows.

Baidu unveils Ernie 4.0 Turbo to compete in AI market

Chinese search engine giant Baidu has introduced an upgraded version of its AI model, Ernie 4.0 Turbo, to stay competitive in China’s AI market. The new feature follows the October 2023 release of Ernie 4, which Baidu claimed had capabilities on par with OpenAI’s GPT-4. The new model will be available to the public through web and mobile interfaces, with developers able to integrate it via Baidu’s Qianfan AI platform. Since its launch, Ernie has amassed 300 million users.

At the same event, Baidu also announced enhancements to its PaddlePaddle AI ecosystem, which now supports 14.65 million developers and serves 370,000 businesses and institutions. The introduction of Ernie 4.0 Turbo comes as OpenAI plans to block access to its API from China and other countries starting 9 July, impacting many Chinese startups relying on OpenAI’s technology. In response, Baidu and other domestic firms like Alibaba have launched initiatives to attract these users by offering free migration services and incentives.

NBC using AI to recreate Al Michaels’ voice for Olympics recaps

NBC is set to bring sportscaster Al Michaels back to the Olympics with a twist this summer: his voice will be powered by AI. The network announced on Wednesday that AI software will recreate Michaels’ voice to deliver daily recaps of the Summer Games for subscribers of its Peacock streaming platform. That marks a significant milestone for the use of AI by a major media company.

The AI-driven recaps will be part of a new feature called ‘Your Daily Olympic Recap on Peacock,’ offering 10-minute highlight packages. These packages will include event updates, athlete backstories, and other content personalised to subscriber preferences. NBC claims the highlights can be packaged in about 7 million different ways, drawn from 5,000 hours of live coverage from Paris, showcasing the efficiency of AI in delivering tailored content.

Al Michaels expressed initial scepticism about the project but became intrigued after seeing a demonstration. He is being compensated for his involvement. Michaels, known for his long broadcasting career, including the iconic Miracle on Ice Game at the 1980 Winter Olympics, lent his past NBC broadcast audio to train the AI system. NBC assures that all content will be reviewed by a team of editors for factual accuracy and proper pronunciation. The highlights tool will be available on Peacock via web browsers and iOS and iPadOS apps starting 27 July.

Time magazine partners with OpenAI for content access

Time magazine has entered a multi-year agreement with OpenAI, granting the AI firm access to its news archives. The deal allows OpenAI’s ChatGPT to cite and link back to Time.com in user queries, although financial details were not disclosed. OpenAI, led by Sam Altman, has forged similar partnerships with prominent media outlets such as the Financial Times, Axel Springer, Le Monde, and Prisa Media.

These collaborations help train and enhance OpenAI’s products while providing media companies access to AI technology for developing new products. Despite some media companies suing OpenAI over content usage, such partnerships are crucial for training AI models and offer a potential revenue stream for news publishers. Such a trend comes amid broader industry tensions, highlighted by Meta’s decision to block news sharing in Canada following new legislation requiring payment for news content.

Why does it matter?

The OpenAI-Time deal is part of a larger movement where publishers seek fair compensation for their content amid the rise of generative AI, which has prompted discussions on ethical content usage and compliance with web standards.

AI-generated Elon Musk hijacks Channel Seven’s YouTube

Channel Seven is currently investigating a significant breach on its YouTube channel, where unauthorised content featuring an AI-generated deepfake version of Elon Musk was streamed repeatedly. The incident on Thursday involved the channel being altered to mimic Tesla’s official presence. Viewers were exposed to a fabricated live stream where the AI-generated Musk promoted cryptocurrency investments via a QR code, claiming a potential doubling of assets.

During the stream, the fake Musk engaged with an audience, urging them to take advantage of the purported investment opportunity. The footage also featured a chat box from the fake Tesla page, displaying comments and links that further promoted the fraudulent scheme. The incident affected several other channels under Channel Seven’s umbrella, including 7 News and Spotlight, with all content subsequently deleted from these platforms.

A spokesperson from Channel Seven acknowledged the issue, confirming they are investigating alongside YouTube to resolve the situation swiftly. The network’s main YouTube page appeared inaccessible following the breach, prompting the investigation into how the security lapse occurred. The incident comes amidst broader challenges for Seven West Media, which recently announced significant job cuts as part of a cost-saving initiative led by its new CEO.

Why does it matter?

The breach underscores growing concerns over cybersecurity on social media platforms, particularly as unauthorised access to high-profile channels can disseminate misleading or harmful information. Channel Seven’s efforts to address the issue highlight the importance of robust digital security measures in safeguarding against such incidents in the future.

AI and the UK election: Can ChatGPT influence the outcome?

With the UK heading to the polls, the role of AI in guiding voter decisions is under scrutiny. ChatGPT, a generative AI tool, has been tested on its ability to provide insights into the upcoming general election. Despite its powerful pattern-matching capabilities, experts emphasise its limitations and potential biases, given that AI tools rely on their training data and accessible online content.

ChatGPT suggested a strong chance of a Labour victory in the UK based on current polling when prompted about the likely outcomes of the election. However, AI’s predictions can be flawed, as demonstrated when a glitch led ChatGPT to declare Labour as the election winner prematurely incorrectly. This incident prompted OpenAI to refine ChatGPT’s responses, ensuring more cautious and accurate outputs.

ChatGPT can help voters navigate party manifestos, outlining the priorities of major parties like Labour and the Conservatives. By summarising key points from multiple sources, the AI aims to provide balanced insights. Nevertheless, the psychological impact of AI-generated single answers remains a concern, as it could influence voter behaviour and election outcomes.

Why does it matter?

The use of AI for election guidance has sparked debates about its appropriateness and reliability. While AI can offer valuable information, the importance of critical thinking and informed decision-making must be balanced. As the election date approaches, voters are reminded that their choices hold significant weight, and participation in the democratic process is crucial.

OpenAI acquires Rockset to boost data processing and AI capabilities

OpenAI has made a pivotal advancement in the AI sector by acquiring Rockset, known for its search and analytical database technology. The acquisition is a strategic move to elevate OpenAI’s data processing and utilisation capabilities.

By integrating Rockset’s advanced infrastructure into its product suite, OpenAI aims to significantly enhance its AI tools, making data retrieval more precise and actionable. Founded in 2016 by former Meta engineers, Rockset excels in cloud-based data management, including real-time streaming data ingestion and sophisticated querying for time series, geospatial, and vector data.

Notably, its recent Approximate Nearest Neighbour (ANN) search upgrade boosts AI efficiency through rapid similarity searches across large datasets. With an acquisition valued at several hundred million dollars, OpenAI seeks to address business challenges in data integration with AI systems, thereby improving the accuracy and utility of its offerings, such as ChatGPT.

Brad Lightcap, OpenAI’s COO, highlighted customer benefits, emphasising Rockset’s ability to turn data into actionable insights. He reassured existing Rockset clients of a smooth transition, ensuring service continuity during the integration.

A strategic move like this underscores OpenAI’s dedication to enhancing AI accessibility and functionality, promising significant improvements in data-driven AI applications.

IMF report highlights Nigeria’s AI infrastructure challenges

A new International Monetary Fund (IMF) report reveals that Nigeria and several other developing nations need more digital infrastructure to effectively deploy AI. The shortfall persists despite Nigeria’s recent launch of its first Multilingual Large Language Model (LLM) and unveiling of an AI strategy in April. The IMF’s ‘AI Preparedness Index,’ which evaluates 174 economies, highlights that many developing countries like Nigeria are trailing in AI readiness due to inadequate digital infrastructure.

According to the Index’s interactive map, most African nations, with exceptions like Namibia, Botswana, and South Africa, exhibit low preparedness for AI. Wealthier economies are generally better equipped for AI adoption, and the IMF warns that the disparity could exacerbate existing global inequalities. The report suggests that while AI has the potential to enhance productivity and expand opportunities in countries like Nigeria, it may also widen the gap between those who can leverage the technology and those who cannot.

To address Nigeria’s challenges, the IMF recommends that emerging markets and developing economies invest heavily in digital infrastructure and worker training. For advanced economies, the priority should be on expanding social safety nets and fostering AI innovation and integration. International coordination is also essential to establish regulations that protect against AI risks and abuses while building public trust in the technology.