Category: Blog

**Meta Reports AI Adoption by Major Tech and Banking Giants**

Meta has announced that its AI models have been downloaded nearly 350 million times since last year, representing a remarkable increase of over tenfold compared to this time in 2023, with more than 20 million downloads occurring just last month. The company attributes the success of Llama to the power of open source, stating that its open availability has fostered a vibrant and diverse AI ecosystem, offering developers unprecedented choice and capability.

The adoption of Llama spans across startups and major enterprises, with significant use cases on-premises and through cloud services. Companies like Goldman Sachs, Zoom, AT&T, DoorDash, and Shopify have integrated Llama into their operations, with cloud usage more than doubling in a matter of months. Experts see the potential for Llama 3.1 to transform customer interactions, enabling models to handle simple queries and provide instant assistance, thereby reducing the need for human intervention.

Industry leaders foresee a shift in customer service dynamics, predicting that as the cost of AI-driven intelligence declines, traditional call centers may become obsolete. In a recent earnings call, CEO Mark Zuckerberg discussed Meta’s strategy to monetize AI, envisioning a future where personalized advertisements are created through AI, allowing advertisers to simply set objectives and budgets, while Meta’s systems manage execution. For ongoing updates on AI developments, consider subscribing to daily AI newsletters.

A new cybersecurity tool for hydropower plants is being developed at the National Renewable Energy Laboratory (NREL) aimed at enhancing protection against increasing cyber threats in the energy sector. Named the Cybersecurity Situational Awareness Tool for Hydropower (CYSAT-Hydro), this innovative tool addresses the vulnerabilities posed by the rising integration of internet-connected distributed energy resources and recent high-profile cyberattacks, such as the 2021 Colonial Pipeline ransomware incident.

With a focus on critical infrastructure, CYSAT-Hydro employs artificial intelligence to detect unusual activities within operational technology networks, offering real-time alerts and analytics to ensure quick recovery from potential attacks. It combats various threats, including denial-of-service attacks and unauthorized access to hydropower-integrated systems.

Designed with a user-friendly interface, CYSAT-Hydro is open-source and compatible across different operating systems, facilitating its adoption in various sectors beyond hydropower, such as water and gas pipelines. As NREL advances this tool, plans for field demonstrations and case studies are underway, emphasizing a commitment to fostering a resilient power infrastructure in the face of evolving cybersecurity challenges.

The Delhi High Court expressed the urgent necessity for action against deepfakes, highlighting the significant risks they pose to society. Acting Chief Justice Manmohan and Justice Tushar Rao Gedela remarked that deepfake technology could become a “serious menace,” urging the Centre to prioritize this issue. Deepfakes, which employ artificial intelligence to superimpose one person’s likeness onto another, raise serious concerns regarding misinformation and fabricated narratives. The court reflected on the troubling reality of being unable to trust one’s own sensory perceptions amid the proliferation of such technology.

Amid the discussion, petitions filed by journalist Rajat Sharma and lawyer Chaitanya Rohilla called attention to the lack of regulations around deepfakes and the need to restrict access to related applications. The court instructed the petitioners to submit further recommendations and scheduled additional hearings for October 24. The Additional Solicitor General acknowledged the urgency of the matter, proposing counter AI technologies for detection and prevention, along with support mechanisms and awareness initiatives. The bench concluded that solutions to AI-related challenges must be rooted in technological advancements.

TEXAS LEGISLATORS NAVIGATE THE COMPLEXITIES OF AI REGULATION

A Senate committee recently explored the intricacies of artificial intelligence (AI) during a nearly four-hour hearing, shedding light on Texas’s approach to regulating this rapidly advancing technology.

Concerns About AI

The Texas Senate Business and Commerce Committee highlighted concerns regarding the potential risks associated with AI, such as the dissemination of misinformation, biased decision-making, and infringements on consumer privacy. By the end of the hearing, some committee members concluded that Texas may need to implement regulations governing how private companies utilize AI technology. Sen. Lois Kolkhorst, R-Brenham, emphasized the importance of safeguarding the public, stating, “If you really think about it, it’s a dystopian world we could live in. Our challenge is figuring out how to implement those safeguards.”

The Scope of Artificial Intelligence

Artificial intelligence encompasses various technologies, from chatbots that assist users through language processing to generative AI that creates original content. It can also automate decision-making processes, such as determining home insurance rates or evaluating job applicants. Additionally, AI is utilized to create digital replicas of artistic works. According to Amanda Crawford, chief information officer for the Texas Department of Information Resources, over 100 of the 145 state agencies currently employ AI in some capacity. Crawford is part of the newly established AI Council, formed by Gov. Greg Abbott, Lt. Gov. Dan Patrick, and House Speaker Dade Phelan, to assess the use of AI in state agencies and evaluate the necessity of an ethical framework for AI applications. The council’s report is anticipated by the year’s end.

Impact on State Agencies

Leaders from various state agencies testified on the advantages of AI, noting significant time and cost savings. For instance, Edward Serna, executive director of the Texas Workforce Commission, shared that a chatbot implemented in 2020 has addressed 23 million inquiries. Likewise, Tina McLeod, information officer in the Attorney General’s Office, mentioned that their AI tool helps employees save at least an hour each week when managing lengthy child support cases. However, some stakeholders raised concerns about potential negative impacts on Texans due to AI. Country singer Josh Abbott expressed worry that AI could replicate his voice for unauthorized song creations on platforms like Spotify, stating, “AI fakes don’t care if you’re famous. AI frauds and deep fakes affect everyone.”

The Call for Responsible AI Usage

Policy analyst Grace Gedye from Consumer Reports indicated that private companies have previously used biased AI models for crucial housing and employment decisions detrimental to consumers. She urged lawmakers to mandate audits for companies that rely on AI for decision-making processes, similar to New York City’s law requiring audits for automated employment tools, although the compliance rate has been low. When crafting future legislation, officials must approach the matter delicately to avoid prohibiting beneficial AI applications while addressing associated risks, according to Renzo Soto, executive director of TechNet, which advocates for technology CEOs. “You almost have to look at it industry by industry.”

Texas has already enacted laws like the one in 2019 criminalizing misleading videos intended to influence elections and a 2023 law banning deep fake videos for pornographic purposes. As lawmakers contemplate future regulations, particular attention must be paid to First Amendment rights, as noted by attorney Ben Sheffner from the Motion Picture Association. Throughout the hearing, legislators sought examples from other states and countries to inform their AI policy development. Currently, a patchwork of regulations at both state and federal levels attempts to manage AI usage but with limited success. California has proposed legislation that requires AI developers to mitigate “catastrophic harm” risks from their technologies, while Colorado has already passed regulations concerning AI usage in high-stakes areas like education, employment, and healthcare.

Upcoming Challenges and Developments

As discussions about AI regulation progress, it remains critical for Texas lawmakers to ensure they encourage innovation while protecting citizens from potential abuses of technology.

**AI Alignment as Responsible Innovation**
By Anulekha Nandi
Published on: August 27, 2024

AI alignment research aims to embed human values within AI development, necessitating ongoing vigilance due to diverse contexts and value systems. In recent years, significant shifts in AI alignment research have gained attention, especially following notable exits from OpenAI by co-founder John Schulman and researcher Jan Leike. Both joined Anthropic to deepen their focus on AI alignment, highlighting the rising concerns surrounding AI safety—issues that cannot be solely managed through legal frameworks.

The concept of AI alignment fundamentally refers to encoding human values to minimize potential harm from AI systems. This approach mandates an understanding of the broader social and contextual nuances in which these systems operate. The urgency for effective AI alignment is evident in various sectors like healthcare, where AI systems have erroneously suggested unsafe cancer treatments. Other areas facing similar challenges include AI-driven content moderation, the financial sector’s algorithm-driven trading, and criminal justice systems that rely on biased recidivism algorithms. As a result, it’s clear that AI alignment must ensure sustainable and ethical development across multiple domains.

AI alignment serves a crucial purpose: providing scalable oversight to address issues of misalignment, promote unbiased responses, ensure robustness in unpredictable situations, enhance interpretability, and uphold human control over AI systems.

Despite the straightforwardness of its principles, achieving alignment is notoriously complicated. The intricacies arise from diverse contexts, preferences, and societal sensitivities, coupled with the opaque nature of AI algorithms. This presents a persistent trade-off between transparency and interpretability, as well as computational performance.

Moreover, as AI capabilities evolve, they pose significant risks such as becoming overconfident or prone to hallucinations—emphasizing the necessity of continuous oversight. Methods like Reinforcement Learning from Human Feedback (RLHF) are still susceptible to biases, highlighting a growing need for vigilance in model training and evaluation.

Responsible innovation in the context of AI entails a techno-institutional approach that emphasizes the continuous adjustment of technology to align with both organizational strategies and societal expectations. This foundation stresses collective responsibility for the outcomes generated from scientific and technological advancements.

To operationalize responsible innovation, systemic transformations are essential, which includes the establishment of comprehensive policies and processes integrated with technological safeguards. Research indicates that understanding responsibility is more transparent when uncertainty about action impacts is low. For example, recognizing bias in hiring algorithms is a clear instance requiring remediation. Conversely, scenarios with high uncertainty—such as determining the limits of freedom of expression—demand more nuanced interventions to balance various concerns.

The complexity of AI development within multi-stakeholder ecosystems necessitates a clear understanding of accountability. Distributing responsibilities based on stakeholder roles supports both external congruency in norms and values and internal alignment with organizational capabilities—critical for sustainable initiatives.

AI alignment as responsible innovation emphasizes proactive engagement rather than reactive accountability. It encompasses anticipating, responding, and iteratively refining design choices. Such reactive measures help ensure AI systems reflect shared societal values while accommodating diverse stakeholder interests. In conclusion, as AI systems continue to evolve, it is imperative to cultivate an ecosystem that embraces responsibility, reflective design, and inclusive practices to guide the development and deployment of AI technologies wisely.

Anulekha Nandi is a Fellow at the Observer Research Foundation, specializing in technology policy and digital innovation.

**Pivotal Moment for EU AI Regulation: Who Will Lead the GPAI Code of Practice?**

Published on 26/08/2024 – 15:38 GMT+2 by Kai Zenner, Head of Office, MEP Axel Voss; Cornelia Kutterer, Managing Director, Considerati.

The AI Office is set to appoint a key group of external individuals within the next three weeks to shape the implementation of a critical component of the EU AI Act: the chairs and vice-chairs of the Code of Practices for General-Purpose AI (GPAI) models.

Recent advancements in generative AI, including applications like OpenAI’s ChatGPT, have introduced both economic disruption and political challenges, prompting concerns from member states about potential regulatory overreach that could hinder EU startups. In contrast, the European Parliament has pushed for a robust legal framework to address market concentration and protect fundamental rights.

In light of these differing perspectives, a co-regulatory approach has been adopted, outlining obligations for GPAI model providers via codes and technical standards. While this strategy promises flexibility, critics warn it may lead to minimal compliance from companies.

The AI Office plans to leverage past experiences in developing GPAI, aiming for a strong governance structure. Public consultations and involvement from multiple stakeholders, including GPAI companies, are integral to drafting the Codes of Practice.

The upcoming selection of chairs for the GPAI Code is of paramount importance. Their expertise will significantly influence the regulatory landscape, addressing complex socio-technical challenges and ensuring alignment with EU values. The effectiveness and legitimacy of the GPAI Code hinge on these crucial appointments.

About the Authors:
Kai Zenner is the Head of Office and Digital Policy Adviser for MEP Axel Voss (Germany, EPP) and played a key role in the AI Act negotiations at the technical level. Cornelia Kutterer is Managing Director of Considerati, an adviser to SaferAI, and a researcher in AI at the University of Global Administrative.

Actress Jenna Ortega recently deleted her account on X, formerly known as Twitter, after a distressing incident involving AI-generated explicit images of herself as a minor. In an interview with The New York Times, the 21-year-old actress expressed her deep disdain for artificial intelligence, stating, “I hate A.I.” Although she recognized the technology’s potential for positive use, she raised concerns about its alarming misuse.

Reflecting on her experience, Ortega recalled how, at just 14, she was confronted with inappropriate, edited content depicting her as a child. Describing the experience as “terrifying, corrupt, and wrong,” she remembered opening a direct message at age 12 that contained an unsolicited explicit image. The situation escalated after the release of her popular series “Wednesday,” inundating her with a flood of unwanted explicit content.

“It was disgusting, and it made me feel bad,” Ortega commented. Ultimately, she decided to leave the platform, thinking, “Oh, I don’t need this anymore.”

Her experience highlights a growing problem of nonconsensual AI-generated deepfakes, which have surged online. Research revealed that in 2023, more explicit deepfake videos were uploaded than in all prior years combined, with Ortega being one of the top targeted celebrities on deepfake sites. This trend extends beyond celebrities; many teenage girls in the U.S. have also fallen victim to AI-generated fake nude content, underscoring the urgent need for stronger protections against the misuse of AI technology, especially for vulnerable individuals.

Chinese artificial intelligence (AI) app developers are increasingly targeting international markets as domestic competition intensifies, with many firms exploring opportunities abroad to seek growth. As industry research shows, out of 1,500 active AI companies worldwide, 103 are from China and are now looking to expand overseas.

Alibaba Group has initiated this trend with the launch of SeaLLMs, tailored for Southeast Asia, enhancing its e-commerce and cloud operations in the region. Similarly, ByteDance has introduced a range of consumer-focused applications, including the AI homework helper Gauth and AnyDoor, an interactive character app, demonstrating the potential of AI-driven platforms in international markets.

Industry leaders emphasize that foreign users are generally more willing to invest in software, providing a lucrative avenue for Chinese firms struggling to attract local users. Companies like Motiff and Kunlun Tech are adapting their offerings for these new audiences, focusing on regions such as North America, Europe, and Japan, where the revenue potential is higher. However, navigating political challenges resulting from global tensions complicates this expansion, prompting some firms to distance themselves from their Chinese origins.

Despite these hurdles, Chinese developers are leveraging their strengths in consumer-focused applications and innovative business models to make significant inroads into international markets.

**NEWS Excerpt:**

Today’s headlines cover a range of topics including the latest updates in US politics, technology reviews, and global climate change concerns. The 2024 election is heating up, with discussions around key candidates and their platforms. In tech, there’s a focus on the latest gadgets and gaming innovations, while health news highlights developments related to COVID-19 and mental health.

**LIFE Excerpt:**

In the realm of lifestyle, health remains a priority with articles on managing fall allergies and mental well-being. Parenting tips are gaining traction, alongside insights on style and beauty. For food enthusiasts, there are new recipes and travel ideas that inspire exploration and culinary adventures.

**ENTERTAINMENT Excerpt:**

The entertainment section is buzzing with celebrity news, recent TV shows, and movie releases. Music lovers can find interviews and exclusive content highlighting their favorite artists, while shopping guides make it easier to discover new trends in fashion and entertainment.

**FINANCE Excerpt:**

In finance, the stock market shows dynamic trends, with guides on crypto investments and the Biden economy. Personal finance features include advice on mortgages, credit cards, and banking options to help readers manage their financial health effectively.

**SPORTS Excerpt:**

Sports coverage includes updates on NFL, NBA, MLB, and soccer, with a focus on scores, stats, and player performances. Fantasy sports enthusiasts can find the latest news and strategies for optimizing their teams, while fans of other sports enjoy extensive coverage and video content.

**NEW ON YAHOO Excerpt:**

Yahoo introduces fresh gaming options and tech articles to keep users updated on the latest trends.