MA Research

- Large Language Models (LLMs) are vulnerable to "jailbreak" attacks. - Existing defenses of LLM chatbots can be bypassed using a time-based approach. - Automated jailbreak prompts can be generated with a high success rate. - Findings have been responsibly disclosed to service providers.

- GPT-4 outperforms GPT-3 and GPT-3.5 in sentiment and theme classification. - Prompt engineering improves GPT-4's performance in sentiment classification.

- Almost a third of companies are using generative AI in at least one business function. - High performers attribute at least percent of their EBIT to AI adoption. - Companies need to tackle various issues to ensure a strong return on investment in generative AI. - These issues include identifying opportunities, governance, managing third parties, and understanding risks. - Finding the balance between near-term gains and long-term scalability is crucial.

- Role of prompt engineering in improving GPT models - GPT-4 outperforms GPT-3 and other LLMs

- AI is a tool of power politics and state diplomacy. - EU approaches AI from an economic, social, and regulatory angle. - Discusses the impact of AI on EU's geopolitical power and relationships. - Presents scenarios for how AI may change international balance of power. - Recommends ways for EU and Member States to respond.

- Almost a third of companies are using generative AI in at least one business function. - High performers attribute at least percent of their EBIT to AI adoption. - Organizations plan to increase investment in AI in the years ahead. - Majority of respondents report AI-related revenue increases in each business function.

- AI is a tool of power politics and state diplomacy. - The EU approaches AI from an economic, social, and regulatory angle. - The paper discusses the impact of AI on the EU's geopolitical power. - It presents scenarios for how AI may change the international balance of power. - Recommendations for the EU and its Member States to respond.

- Researchers are excited about the use of AI tools in science. - AI provides faster data processing and saves time and money. - Concerns include reliance on pattern recognition, bias in data, fraud, and irreproducible research. - AI tools for medical diagnostics can perpetuate bias and false information. - Commercial firms dominate computing resources and ownership of AI tools. - Collaboration with scientists at such firms is considered important. - Researchers using LLMs are still in the minority.

- AI will improve healthcare, farming, climate change mitigation, and more. - AI poses risks such as opaque decision-making and discrimination. - EU needs a coordinated approach to promote AI development and address risks. - The purpose of the White Paper is to set policy options. - Europe can become a global leader in innovation in the data economy. - AI should be grounded in European values and fundamental rights. - AI can support the Sustainable Development Goals and the democratic process. - Environmental impact of AI systems should be considered. - A common European approach to AI is necessary to avoid fragmentation. - The White Paper presents policy options for trustworthy and secure AI development.

- AI will improve healthcare, farming, climate change mitigation, and more. - AI poses risks such as opaque decision-making and discrimination. - EU needs a coordinated approach to promote AI development and address risks. - The purpose of the White Paper is to set policy options. - Europe can become a global leader in innovation in the data economy. - AI should be grounded in European values and fundamental rights. - AI can support the Sustainable Development Goals and the democratic process. - Environmental impact of AI systems should be considered. - A common European approach to AI is necessary to avoid fragmentation. - The White Paper presents policy options for trustworthy and secure AI development.

- SciBERT is a pretrained language model based on BERT. - It addresses the lack of high-quality, large-scale labeled scientific data. - SciBERT improves performance on downstream scientific NLP tasks. - It achieves new state-of-the-art results on several tasks. - The code and pretrained models are available at https://github.com/allenai/scibert/.