YouTube has implemented new privacy guidelines allowing individuals to request the removal of AI-generated videos that imitate them. Initially promised in November 2023, these rules are now officially in effect, as confirmed by a recent update to YouTube’s privacy policies.
According to the updated guidelines, users can request the removal of content that realistically depicts a synthetic version of themselves, created or altered using AI. YouTube will evaluate such requests based on several criteria, including whether the content is changed, disclosed as synthetic, identifiable, realistic, and whether it serves public interest like parody or satire. Human moderators will handle complaints, and if validated, the uploader must either delete the video within 48 hours or edit out the problematic parts.
These guidelines aim to protect individuals from potentially harmful content like deepfakes, which can easily mislead viewers. They are particularly relevant in upcoming elections in countries such as France, the UK, and the US, where misusing AI-generated videos could impact political discourse.
Anthropic is launching a new program to fund the creation of new benchmarks for better assessing AI model performance and its impact. In its blog post, Anthropic stated that it will offer grants to third-party organisations developing improved methods for evaluating advanced AI model capabilities.
Urging the AI research community to develop more rigorous benchmarks that address societal and security implications, Anthropic advocated for revising existing methodologies through new tools, infrastructure, and methods. Highlighting how they aim to develop an early warning system to identify and assess risks, it specifically called for tests to evaluate a model’s ability to conduct cyberattacks, enhance weapons of mass destruction, and manipulate or deceive individuals.
Moreover, Anthropic also aims for its new program to support research into benchmarks and tasks that explore AI’s potential in scientific study, multilingual communication, bias mitigation, and self-censorship of toxicity. In addition to grants, researchers will have the chance to consult with the company’s domain experts. The company also expressed interest in potentially investing in or acquiring the most promising projects, offering various ‘funding options tailored to the needs and stage of each project’.
Why does this matter?
Benchmarks are the process of evaluating the quality of an AI system. The evaluation is typically a fixed process of assessing the capability of an AI model, usually in one area, while AI models like Anthropic’s Claude and Open AI’s ChatGPT are designed to perform a host of tasks. Thus, developing robust and reliable model evaluations is complex and is riddled with challenges. Anthropic’s initiative to support new AI benchmarks is commendable, with their stated objective of the program serving as a catalyst for progress towards a future where comprehensive AI evaluation is an industry-standard. However, given their own commercial interests, the initiative may raise trust concerns.
Google announced that it will require advertisers to disclose election ads that use digitally altered content depicting real or realistic-looking people or events to combat misinformation during elections. This latest update to Google’s political content policy mandates advertisers to select a checkbox for ‘altered or synthetic content’ within their campaign settings.
The proliferation of generative AI, capable of rapidly creating text, images, and video, has sparked concerns over potential misuse. Deepfakes, which convincingly manipulate content to misrepresent individuals, have further blurred the distinction between fact and fiction in digital media.
To implement these changes, Google will automatically generate an in-ad disclosure for feeds and shorts on mobile devices and in-stream ads on computers and television. Advertisers must provide a prominently displayed disclosure for other ad formats that is clearly visible to users. According to Google, the exact wording of these disclosures will vary based on the context of each advertisement.
Why does it matter?
Earlier this year, during India’s general election, fake videos featuring Bollywood actors surfaced online, criticising Prime Minister Narendra Modi and urging support for the opposition Congress party. The incident highlighted the growing challenge of combating deceptive content amplified by AI-generated media.
In a related effort, OpenAI, led by Sam Altman, reported disrupting five covert influence operations in May that aimed to manipulate public opinion using AI models across various online platforms. Meta Platforms had previously committed to similar transparency measures, requiring advertisers on Facebook and Instagram to disclose the use of AI or digital tools in creating political, social, or election-related ads.
The UN General Assembly has adopted a resolution on AI capacity building, led by China. This non-binding resolution seeks to enhance developing countries’ AI capabilities through international cooperation and capacity-building initiatives. It also urges international organisations and financial institutions to support these efforts.
The resolution comes in the context of the ongoing technology rivalry between Beijing and Washington, as both nations strive to influence AI governance and portray each other as destabilising forces. Earlier this year, the US promoted a UN resolution advocating for ‘safe, secure, and trustworthy’ AI systems, gaining the support of over 110 countries, including China.
China’s resolution acknowledges the UN’s role in AI capacity-building and calls on Secretary-General Antonio Guterres to report on the unique challenges developing countries face and provide recommendations to address them.
Connecticut is spearheading efforts by developing what could be the nation’s first Citizens AI Academy. The free online resource aims to offer classes for learning basic AI skills and obtaining employment certificates.
Democratic Senator James Maroney of Connecticut emphasised the need for continuous learning in this rapidly evolving field. Determining the essential skills for an AI-driven world presents challenges due to the technology’s swift progression and varied expert opinions. Gregory LaBlanc from Berkeley Law School suggested that workers should focus on managing and utilising AI rather than understanding its technical intricacies to complement the capabilities of AI.
Several states, including Connecticut, California, Mississippi, and Maryland, have proposed legislation addressing AI in education. For instance, California is considering incorporating AI literacy into school curricula to ensure students understand AI principles, recognise its use, and appreciate its ethical implications. Connecticut’s AI Academy plans to offer certificates for career-related skills and provide foundational knowledge, from digital literacy to interacting with chatbots.
Despite the push for AI education, concerns about the digital divide persist. Senator Maroney highlighted the potential disadvantage for those needing more basic digital skills or access to technology. Marvin Venay of Bring Tech Home and Tesha Tramontano-Kelly of CfAL for Digital Inclusion stress the importance of affordable internet and devices as prerequisites for effective AI education. Ensuring these fundamentals is crucial for equipping individuals with the necessary tools to thrive in an AI-driven future.
Grammy Award-winning musician and producer Chance the Rapper is known for his innovative approach to music and fashion. Recently, he has teamed up with Meta for their Super Fan event, showcasing his interest in cutting-edge technology, particularly Meta AI. The collaboration highlights how AI transforms various aspects of his work, from engaging fans to creating music and fashion.
Chance has long been a pioneer in using digital platforms to connect with fans and distribute his music. With the advent of AI, he is now pushing the boundaries of creativity even further. He likens AI to the patchwork denim look he sported at the Meta event, describing it as an amalgamation of different design patterns. The comparison underscores his view of AI as a tool for combining diverse elements to create something unique.
The Meta AI suite, integrated into platforms like Instagram and Facebook, allows Chance to explore new artistic directions. He uses these tools to experiment with music production, generate unique soundscapes, and refine his musical style. Chance also finds inspiration on Instagram, drawing from various topics and incorporating these influences into his work.
Additionally, Chance sees potential in Meta’s new Ray-Ban Meta smart glasses, which offer responsive tech for human engagement and photography. By leveraging AI tools, he enhances his artistic process, engages more effectively with fans, and supports initiatives like the growth of women’s sports. As he prepares to release his new project, ‘Star Line Gallery,’ Chance the Rapper continues to inspire and innovate in the realms of music and fashion.
The City of London Corporation, London and Partners and Microsoft have launched an AI Innovation Challenge, where participants will vie to spot and stop cybercriminals using fake identities and audio and visual deepfakes to commit fraud. With the increase of such events and the ubiquity of GenAI models, Nvidia, the multinational AI chip-maker, is increasingly becoming the modern-day Standard Oil. Nvidia’s chips can be found in just about all areas of economic activity, from education to medicine and in nearly all financial and professional services.
With its growing usage, its potential for fighting cybercrime increases, given its ability to analyse vast amounts of data rapidly, decipher patterns, and ultimately lead to higher fraud detection rates and greater trust in and securitise customer services. Banks in the United Kingdom lead the way in AI adoption, particularly as some 90 percent of them have already onboarded generative AI models to their asset portfolios.
Participants of the AI Innovation Challenge have until 26 July 2024 to register for the competition, which is scheduled for six weeks between September and November. The final event promises to be a display of fraud detection and other cybersecurity innovations developed during the course of the competition.
The surge in US stock prices, driven by enthusiasm for AI, draws comparisons to the dot-com bubble two decades ago, sparking concerns over inflated valuations. The S&P 500 has reached new records, climbing more than 50% from its October 2022 low, while the Nasdaq Composite has surged over 70% since the end of 2022. A few massive tech stocks, including Nvidia, are leading this rally, reminiscent of the ‘Four Horsemen’ tech stocks of the late 1990s.
Despite the impressive gains, some analysts caution that today’s tech stocks are more financially robust than their dotcom-era counterparts. However, fears persist that the AI-driven surge might end in a crash similar to the dotcom bust, which saw the Nasdaq Composite plummet nearly 80% from its March 2000 peak, and while some companies like Amazon thrived post-bubble, many did not recover.
Current tech stock valuations, while high, are more grounded in solid earnings prospects rather than speculative growth, a key difference from the dot-com era. For instance, Nvidia trades at 40 times forward earnings estimates, far lower than Cisco’s 131 times in 2000. Although the S&P 500’s price-to-earnings ratio of 21 is above its historical average, it remains below the peak levels of the late 1990s. Nonetheless, investors remain cautious, wary of metrics becoming overly stretched if economic growth continues and tech stocks keep climbing.
In the evolving landscape of marketing and advertising, the integration of generative AI presents both promise and challenges, as highlighted in a recent Forrester report. Key obstacles include a lack of AI expertise among agency employees and concerns over job obsolescence. Also, the human factor poses a significant hurdle that the industry must address urgently to fully harness the potential of genAI.
The potential economic impact of genAI on agencies is profound. Seen as a transformative force akin to the advent of smartphones, genAI promises to redefine creativity in marketing by combining data intelligence with human intuition. Agency leaders overwhelmingly recognise it as a disruptive technology, with 77% acknowledging its potential to fundamentally alter business operations. However, the fear of job displacement among employees remains palpable, exacerbated by recent industry disruptions and the rapid automation of white-collar roles.
To mitigate these concerns and fully embrace genAI, there is a pressing need for comprehensive AI literacy and training within agencies. While existing educational programmes and certifications provide a foundation, they are insufficient to meet the demands of integrating AI into everyday creative processes. Investment in reskilling and upskilling initiatives is crucial to empower agency employees to confidently navigate the AI-driven future of marketing and advertising.
Industry stakeholders, including agencies, technology partners, universities, and trade groups, must collaborate to establish robust training frameworks. In addition, a concerted effort will not only bolster agency capabilities in AI adoption but also ensure that creative workforce remains agile and competitive in an increasingly AI-centric landscape. By prioritising AI literacy and supporting continuous learning initiatives, agencies can position themselves at the forefront of innovation, delivering enhanced value to clients through AI-powered creativity.
A recent study has revealed that ChatGPT and similar large language models (LLMs) are highly effective in launching cyberattacks, raising significant concerns in the cybersecurity field.
Researchers Richard Fang, Rohan Bindu, Akul Gupta, and Daniel Kang tested ChatGPT-4 against 15 real-life ‘one-day’ vulnerabilities, finding that it could exploit these vulnerabilities 87% of the time. These vulnerabilities included websites issues, container management software, and Python packages, all sourced from the CVE database.
The study utilised a detailed prompt with 1,056 tokens and 91 lines of code, including debugging and logging statements. The research team noted that ChatGPT-4’s success stemmed from its ability to handle complex, multi-step vulnerabilities and execute various attack methods. However, without the CVE code, ChatGPT-4’s success rate plummeted to just 7%, highlighting a significant limitation.
The researchers concluded that while ChatGPT-4 currently stands out in its ability to exploit one-day vulnerabilities, the potential for LLMs to become more powerful and destructive is a major concern. They emphasised the importance of the cybersecurity community and LLM providers collaborating to integrate these technologies into defensive measures and carefully consider their deployment.
