1. Robustness and reliability: Ensuring that AI systems perform as intended under various conditions and do not exhibit unexpected or harmful behavior.
2. Ethical considerations: Addressing issues related to fairness, accountability, transparency, and privacy in AI systems to prevent harm or discrimination against individuals or groups.
3. Value alignment: Aligning the goals and values of AI systems with those of human users and society at large to prevent conflicts or unintended consequences.
4. Risk mitigation: Developing strategies and mechanisms to identify, assess, and mitigate potential risks associated with the development and deployment of AI technologies, such as unintended side effects, misuse, or unintended consequences.
5. Long-term impacts: Anticipating and planning for the long-term societal, economic, and existential impacts of AI technologies, including issues related to employment, inequality, and the potential for AI to surpass human capabilities.

Overall, the goal of AI safety is to ensure that AI technologies are developed and deployed in a way that maximizes their benefits while minimizing their risks and potential for harm to individuals, society, and the broader environment.

AI safety research is still an emerging field, and new questions and concerns about safety can arise almost daily. However, with increasing use of AI in different aspects of academia, it’s important for stakeholders to understand key issues in AI safety.

Data Poisoning

Data poisoning is a type of cyber attack or manipulation aimed at corrupting the training data used to develop or fine-tune machine learning models. In data poisoning attacks, adversaries strategically inject malicious or misleading data into the training dataset with the goal of undermining the performance or integrity of the machine learning model.

Data poisoning attacks can take various forms, including

1. Label Flipping: Adversaries manipulate the labels or annotations associated with data points to mislead the model during training. For example, they may change the label of a cat image to “dog” to confuse the model.
2. Feature Tampering: Attackers modify certain features or attributes of the data to introduce biases or distortions that can mislead the model’s learning process. This can involve altering pixel values in images or modifying text to include misleading information.
3. Data Injection: Adversaries inject entirely fabricated or malicious data points into the training dataset to skew the model’s decision boundaries or induce specific behaviors. These injected data points can be carefully crafted to exploit vulnerabilities in the model’s learning algorithms.
4. Data Manipulation: Attackers may manipulate the distribution of the training data by selectively adding or removing samples to bias the model’s predictions in favor of certain outcomes or classes.

Deepfakes

Deepfakes are synthetic media, typically videos, that are created using deep learning techniques, particularly generative adversarial networks (GANs) and deep neural networks (DNNs). These technologies allow for the manipulation of visual and audio content to produce highly realistic forgeries that are often difficult to distinguish from genuine footage.

Deepfakes have garnered significant attention due to their potential for misuse, including

1. Misinformation and Fake News: Deepfakes can be used to create convincing but entirely fabricated videos of public figures, politicians, or celebrities saying or doing things they never actually did. This poses a significant risk for spreading misinformation and undermining trust in media and public figures.
2. Privacy Concerns: Deepfake technology can be used to create non-consensual pornography or to fabricate compromising videos of individuals without their consent, leading to privacy violations and potential harm to victims.
3. Fraud and Social Engineering: Deepfakes could be employed for fraudulent purposes, such as impersonating individuals in video calls or creating fake audio messages to deceive people into believing they are communicating with someone they trust.

Transparency and Bias

Many AI systems, particularly those based on complex deep learning models, operate as “black boxes,” making it difficult to understand how they arrive at their decisions. Ensuring transparency and explainability in AI is essential for building trust, enabling accountability, and facilitating human oversight in critical applications where the consequences of errors or failures can be significant.

Moreover, bias in AI systems can lead to unfair or discriminatory outcomes, particularly when these systems are used in high-stakes decision-making processes such as hiring, lending, and criminal justice. Addressing bias and promoting fairness in AI requires careful consideration of the data used to train models, as well as the design and evaluation of algorithms to mitigate biased outcomes.

The post AI Safety Aspects That Universities, Publishers, and Journals Need to Understand appeared first on CACTUS.

The recent Barcelona Declaration, on the other hand, advocates for such research information to be open, that is, freely available to all users and without restrictions on how such data can be used.

Below, we’ve compiled some of the main questions universities, publishers, and researchers have around the Barcelona Declaration.

What is the Barcelona Declaration on Open Research Information?

The Barcelona Declaration is a call for the scientific community to make the following 4 commitments:

1. To make openness the default for the research information they use and produce
2. To use services and systems that support and enable open research information
3. To support the sustainability of infrastructures for open research information
4. support collective action to accelerate the transition to openness of research information

Who prepared it?

Over 25 research information experts, from organizations involved in conducting, funding, and evaluating research, as well as those providing research information infrastructures, collaborated to prepare the Barcelona Declaration on Open Research Information. They convened in Barcelona in November 2023 for a workshop hosted by the SIRIS Foundation.

When was it released?

April 16, 2024.

Who supports it?

The Barcelona Declaration already has an impressive list of signatories, ranging from prominent funders like the Bill & Melinda Gates Foundation to organizations like the UK Reproducibility Network and Science Europe. cOAlitionS, Crossref, DOAJ (Directory of Open Access Journals), and Europe PMC are also in support of this declaration.

How can I learn more about it?

A webinar in English will be held on April 23 1.00-2.30pm CEST, to officially launch the Barcelona Declaration in English. Attendees can learn about the text of the Barcelona Declaration and its aims. They can also interact with members of signatory organizations and find out the next steps for community coordination. 

Organizations that would like to know more about the Declaration or that would like to sign the Declaration can email contact@barcelona-declaration.org.

The post A Step Towards Open Research Information: FAQs about the Barcelona Declaration appeared first on CACTUS.

AI holds immense promise in augmenting clinical decision-making processes, analyzing vast datasets, and uncovering patterns that may evade human perception. Its ability to sift through complex data sets with speed and precision has led to breakthroughs in disease detection, personalized medicine, and treatment optimization.

However, with great potential comes inherent challenges. The complexity of AI algorithms, the potential for bias, and the lack of standardized reporting pose significant hurdles in ensuring the reliability and reproducibility of AI-driven research. Therefore, it is imperative for medical journal editors to provide their authors with clear guidelines that address the methodology, validation, and interpretation of AI algorithms in research articles.

The below infographic aims to elucidate different AI-related guidelines tailored for medical journal publications. By fostering transparency, accountability, and adherence to best practices, these guidelines not only enhance the quality of AI-driven research but also safeguard the integrity of medical literature, ultimately advancing patient care and outcomes.

The post Reporting Guidelines Essential for Biomedical Journals Publishing AI-Related Research appeared first on CACTUS.

The academic world has made strides in recognizing gender disparities, yet women continue to face challenges that impede their engagement with conferences. Issues ranging from caregiving responsibilities to subtle biases persist, creating a disparity in representation and limiting the diverse perspectives that conferences thrive on. It is incumbent upon research societies to champion inclusivity, acknowledging the critical need for diverse voices in shaping the future of academia.

This infographic explores concrete strategies to dismantle barriers and enhance the participation of women in academic conferences. Embracing these measures not only amplifies the contributions of women in academic spheres but also fosters a more robust and equitable scholarly community.

The post Fostering Inclusivity: 7 Ways Research Societies Can Empower Women in Academic Conferences appeared first on CACTUS.

See also: The Digital Development of China’s Academic Publishing Industry

Educational Emphasis and Research Commitment

China’s unwavering commitment to education is evident in its prioritization of research and development. The government’s consistent focus on building a knowledge-based society has led to a proliferation of universities and research institutions across the nation. This emphasis creates a robust demand for cutting-edge academic materials, providing Western publishers with an eager and intellectually curious audience.

See also: What Is PubScholar, China’s New Academic Database? An Introduction for Western Research Publishers, Societies, and Universities

Growth of Higher Education Institutions

The number of universities in China has been steadily increasing, reflecting the country’s dedication to fostering a highly educated workforce. This surge in higher education institutions translates into a substantial market for Western academic publishers. Collaborations with Chinese universities can facilitate the dissemination of knowledge and ideas, creating a symbiotic relationship that benefits both parties.

Global Knowledge Exchange

China’s integration into the global academic community has been accelerating. International collaborations and partnerships with Western institutions have become increasingly common. Western publishers entering the Chinese market not only tap into a vast consumer base but also contribute to the cross-cultural exchange of knowledge, fostering a more interconnected global academic landscape.

See also: How Culture Shapes Academia in China: A Guide for Western Publishers and Journals

Digital Transformation and E-Learning Trends

The rapid digitalization of education in China aligns with global trends, opening up avenues for Western publishers to provide e-learning solutions and digital resources. Embracing technology allows publishers to meet the evolving needs of Chinese researchers, educators, and students, offering dynamic and interactive content that enhances the learning experience.

See also: Five Key Digital Marketing Trends in China That Western Academic Publishers Need to Know

Conclusion

In 2024, China’s role as a pivotal market for Western academic publishers is undeniable. The confluence of a commitment to education, a growing number of higher education institutions, an expanding middle class, and the embrace of digital learning creates a fertile ground for publishers to thrive. By understanding and tapping into the unique opportunities presented by China, Western academic publishers can not only expand their market share but also contribute to the global dissemination of knowledge in this dynamic era of research.

The post Unlocking Opportunities: China’s Significance for Academic Publishers in 2024 appeared first on CACTUS.

See also: China National Knowledge Infrastructure (CNKI): A Vital Player in China’s Research Landscape

PubScholar, currently in its initial phase, aims to redefine the landscape by providing comprehensive services, including academic resource searching, content acquisition, exchange, and sharing. Positioned as both a domestic hub and a global index of high-quality public welfare academic resources, the platform integrates the wealth of sci-tech achievements, sci-tech publishing, and academic exchanges hosted by the CAS.

See also: The Chinese Academy of Sciences: Its Importance in China’s Research Ecosystem

Following the global movement toward Open Science, PubScholar is committed to openness, allowing academic resources to integrate services under open access conditions and agreements. This inclusive approach has already resulted in the collection of about 170 million resources on the platform. Notably, approximately 80 million of these resources, encompassing journals, dissertations, patent literature, and scientific data, are available for free, challenging the conventional subscription models.

See also: The Digital Development of China’s Academic Publishing Industry

The platform’s emergence has garnered widespread attention and interest since its launch on November 1. It aims to uphold intellectual property rights and international norms while providing an alternative to established academic databases. The National Science Library at the CAS, overseeing PubScholar, is actively responding to the surge in public interest by upgrading the infrastructure to ensure optimal service response times and accommodate the growing user base. As PubScholar continues to evolve, it holds the potential to reshape the academic landscape, offering researchers and scholars expansive, free access to a wealth of scholarly materials. The infographic below provides a basic introduction to PubScholar and its key characteristics.

The post What is PubScholar, China’s New Academic Database? An Introduction for Western Research Publishers, Societies, and Universities appeared first on CACTUS.

• Sharing information about the latest research developments to a wider audience
• Helping the general public to understand the significance of research and to develop scientific literacy
• And most importantly, fostering a robust and continuous stream of talent in STEM fields.

Therefore, for research societies, it is crucial to engage with and provide support to science educators as essential steps in promoting their respective fields and fulfilling their missions.

Why Should Research Societies Support Science Educators?

In many countries, science education is a mandatory part of curricula for not just high school education and above, but also kindergartens, elementary schools, and institutions serving students with special needs. Science educators come from diverse backgrounds, ranging from individuals with advanced scientific degrees to those with basic knowledge. Often, these educators are not highly specialized and frequently seek additional resources to enhance their knowledge of science and teaching effectiveness. Research societies possess the expertise and knowledge to support these educators in achieving their objectives.

By supporting science education and educators, research societies stand to gain from larger, more diverse, and highly motivated future memberships, along with an improved public image and acknowledgment of their valuable work.

The infographic below outlines practical ways by which research societies can engage with and support science educators, in order to advance their own mission.

The post 5 Ways Research Societies Can Support Science Educators appeared first on CACTUS.

The post How Publishers and Journals Can Encourage OA Submissions from Underserved Groups appeared first on CACTUS.

The post 12 Steps Towards Transparent and Credible Peer Review: A Guide for Journals and Publishers appeared first on CACTUS.

The post What Academic Publishers Need to Know About STM’s SDG Roadmap: A Handy Infographic appeared first on CACTUS.