THE DECODER - EVERYTHING AI

1/ OpenAI is demanding access to research papers, notes and other internal documents from the New York Times as part of the newspaper's copyright lawsuit against the AI company, arguing that they are crucial to assessing whether the Times' copyright claims are valid. 2/ The New York Times vehemently opposes the release of the requested documents, calling OpenAI's demands "unprecedented," "harassing," and an attempt to undermine established intellectual property rights and intimidate the news organization. 3/ OpenAI's lawyers appear to be focusing on copyright and technical issues rather than directly seeking to clarify fair use, possibly playing for time until fair use becomes less relevant to AI companies as they find ways to train AI models without potentially infringing copyright.

1/ OpenDevin is an open-source platform for developing AI agents that can interact with their environment and other applications via software interfaces. 2/ The platform provides a flexible architecture with a secure sandbox environment where AI agents can execute code, access tools such as Jupyter notebooks, and perform complex software development tasks. 3/ Initial benchmarks show competitive performance of OpenDevin agents, and the platform is intended to drive the development of safe and useful AI agents as a community project.

1/ A survey of 2,500 employees in the US, UK, Australia and Canada by the Upwork Research Institute shows that many employees feel overwhelmed by the AI tools that their managers are demanding. 2/ 96 percent of managers expect AI tools to increase productivity. However, 47% of employees do not know how to meet these expectations. According to 77%, the tools have even reduced productivity and increased workload. 3/ The results indicate that generative AI has not yet delivered on its promise of productivity. There are many reasons for this. Companies should identify specific use cases, optimize AI for them and introduce it step by step instead of simply providing chatbots.

1/ A study by researchers at China's Zhejiang University found that while AI models have made progress in processing text, images, speech, and video together, they struggle with understanding abstract visuals like diagrams and charts. 2/ The researchers created a dataset of 11,193 abstract images with related questions, covering eight scenarios: dashboards, road maps, diagrams, tables, flowcharts, relationship graphs, visual puzzles, and 2D floor plans. 3/ When tested on this dataset, advanced models like GPT-4o and Claude 3.5 Sonnet only achieved average accuracies of 64.7% and 59.9% respectively, falling short of human performance of at least 82.1%.

1/ A three-year study by OpenResearch, a research lab funded by OpenAI founder Sam Altman, examined the effects of an unconditional basic income of $1,000 per month on 3,000 randomly selected people in Texas and Illinois. 2/ Recipients used the money primarily for basic needs such as food, rent, and transportation. The largest relative increase was in financial support for others. Recipients also moved more often and used more medical services. 3/ Employment rates and hours worked increased for both groups, but less so for recipients. They showed more entrepreneurial interest, especially black and female participants. The study illustrates how an unconditional income can increase agency in several areas of life.

1/ Chinese technology companies are developing AI models such as video generators and chatbots that can compete with leading U.S. systems. They are also benefiting from open-source technologies from Western companies. 2/ Regulating open-source AI is a challenge: If the West slows down its own open-source projects, Chinese technologies could become the new standard. Chinese companies are also more willing to share their AI systems with consumers. 3/ In China, AI models must be reviewed for compliance with socialist values before they are released. The Kling video generator illustrates the problem of importing AI from countries with different values: videos of the Tiananmen Square protests are rejected with an error message, but a video of the burning White House is generated.

1/ Elon Musk shared a video on X (formerly Twitter) that uses AI to imitate the voice of U.S. Vice President Kamala Harris, putting fake statements into her mouth without indicating that it was a fake or a parody. 2/ The doctored video appears to violate X's policy, which states that synthetic media must be labeled or deleted if it could cause public confusion or harm. This is especially true for videos doctored by AI. 3/ The original creator of the video said it was a parody. Neither Musk nor Platform X have yet responded to criticism that the video could be misleading.

1/ In an op-ed for the Washington Post, OpenAI CEO Sam Altman warns of an authoritarian AI future and calls for a US-led global coalition for a democratic AI vision. 2/ He suggests four key points: Development of robust security measures, expansion of physical infrastructure, coherent trade diplomacy for AI, and new models for global AI norms. 3/ Altman emphasizes that democratic AI is superior to authoritarian AI because the U.S. system allows for research and innovation. He believes it is the responsibility of leading nations and technologists to act now.

1/ Despite explicit prohibitions in 15 percent of organizations, 99 percent of development teams use AI tools to generate code. 2/ Only 29 percent of companies have established governance for the use of generative AI. In 70 percent of cases, individual departments make ad hoc decisions about the use of AI tools. At the same time, security concerns are growing. 3/ 47 percent of respondents are open to allowing AI to make unsupervised code changes. Security teams are faced with the challenge of dealing with a flood of potentially vulnerable AI-generated code.

1 AI romantic companions, such as myanima.ai, Eva AI, Nomi.AI, and Replika, are becoming increasingly popular, offering personalized interactions and emotional support to users, especially those struggling with loneliness or intimacy issues. 2/ Research suggests that AI chatbots can provide companionship, reduce loneliness, and elicit emotional responses similar to human relationships; however, concerns arise regarding the impact on users' social skills, expectations, and emotional dependency. 3/ While AI companions show promise in enhancing romantic well-being and supporting marginalized communities, privacy issues and ethical concerns highlight the need for a well-regulated approach to their integration into romantic lives.