• AI KATANA
  • Posts
  • AI giants to unveil pact to fight political deepfakes

AI giants to unveil pact to fight political deepfakes

Also: ‘Better than a real man’: Young Chinese women turn to AI boyfriends

Author

AI KATANA
February 13, 2024

Hi!

In a convergence of technology and societal trends, recent developments highlight the dynamic interaction between AI and various aspects of daily life and global issues. Tech giants such as Meta, Microsoft, Google, and OpenAI have announced a collaborative effort to combat the misuse of AI in creating political deepfakes, aiming to safeguard elections through methods that identify, label, and control misleading AI-generated content. Concurrently, in China, AI boyfriends are gaining popularity among young women, offering companionship and emotional support through customizable chatbots, despite raising questions about privacy and the nature of these digital interactions. Meta's AI Chief, Yann LeCun, discusses the pursuit of AGI, criticizing current approaches and advocating for open-source development to mitigate AI risks. Meanwhile, the US Patent Office reaffirms that AI cannot hold patents, emphasizing human contribution in the invention process. In the realm of scientific research, a new AI tool, BioCLIP, significantly advances the classification of biological images, utilizing a vast database to enhance our understanding of biodiversity. These stories encapsulate the multifaceted impact of AI on innovation, intellectual property, companionship, and the integrity of democratic processes, underlining the technology's potential to shape future societal and scientific landscapes.

Sliced:

  • 🗳️ AI giants to unveil pact to fight political deepfakes

  • 💟 ‘Better than a real man’: Young Chinese women turn to AI boyfriends

  • ⚠️ Meta’s AI Chief Yann LeCun on AGI, Open-Source, and AI Risk

  • 📂 US patent office confirms AI can’t hold patents

  • 🧑‍🔬 New AI tool helps leverage database of 10 million biology images

Tech giants including Meta, Microsoft, Google, and OpenAI are forming a pact to address AI-generated content misleading voters. Set to be announced at the Munich Security Conference, this collaboration aims to develop methods for identifying, labeling, and controlling deceptive AI images, videos, and audio. The initiative, driven by concerns over AI misuse in elections, includes adopting a common watermarking standard to mark AI-generated content. This response is to growing apprehensions about deepfakes, highlighted by recent incidents in the US and Pakistan, where AI was used to create misleading political content.

In China, young women are increasingly turning to AI boyfriends for companionship, with apps like Glow and Wantalk offering virtual partners who provide emotional support, conversation, and companionship. These AI chatbots, customizable in personality and appearance, are popular among those seeking to alleviate loneliness amidst China’s fast-paced lifestyle and societal pressures. Despite concerns over privacy and the authenticity of such interactions, these AI companions offer a unique solution to emotional and social needs, blending technology with personal life in contemporary society.

Yann LeCun, Meta's chief AI scientist, discusses the challenges in achieving AGI, criticizing the effectiveness of LLMs as a pathway to human-level intelligence. He emphasizes the need for AI systems to learn from the physical world, similar to early human development, to reach even basic levels of intelligence. LeCun also advocates for open-source AI development to ensure diversity and democracy in future AI systems, and he addresses concerns about AI risks, considering fears about AI posing existential threats to humanity as exaggerated.

The US Patent and Trademark Office (USPTO) has reaffirmed that AI systems cannot be recognized as inventors on patent applications. However, it clarified that humans utilizing AI tools in the invention process are required to disclose their use of such technology. This policy stems from public consultations and aims to ensure that patents are granted only when there is a significant human contribution to the invention's conception, highlighting the importance of human intellect in the creative process.

Researchers at Ohio State University have developed BioCLIP, a groundbreaking AI tool that utilizes the largest dataset of biological images to date, comprising over 10 million images across more than 454,000 taxa. BioCLIP, which outperforms existing models by 17% to 20% in classifying images across the tree of life, even for rare and unseen species, represents a significant advancement in using AI for biological research. The tool's capacity to discern subtle differences between similar-looking organisms enhances its utility for a wide range of scientific inquiries, promising to accelerate the discovery and understanding of biological diversity.

🛠️ AI tools updates

NVIDIA has released "Chat with RTX," a free tech demo allowing users with NVIDIA RTX GPUs to create personalized chatbots on Windows PCs. This tool uses advanced technologies like retrieval-augmented generation and NVIDIA RTX acceleration to integrate local files or YouTube content into a dataset for generating contextually relevant answers quickly. It supports various file formats and runs locally, ensuring fast responses without sharing data with third parties or requiring an internet connection. Additionally, "Chat with RTX" highlights the potential for developers to build and deploy their own applications using NVIDIA's technologies.

💵 Venture Capital updates

Marqo, an AI startup founded by former Amazon employees, secured $12.5 million in Series A funding led by Lightspeed Venture Partners, with participation from Blackbird Ventures and January Capital. The funds are earmarked for developing technology that enables companies to search through large volumes of unstructured data, such as emails and web server logs. This technology aims to unlock the potential of data that makes up about 90% of corporate information but has been challenging to utilize effectively. Marqo's platform offers solutions for e-commerce personalization and sentiment analysis, leveraging machine learning for vector search in industries like healthcare and e-commerce. The company, which has raised a total of $17.8 million, is relocating its headquarters to San Francisco from London and Melbourne.

🫡 Meme of the day

⭐️ Generative AI image of the day