AI by PRC # 36
Gemini, the new AI twins. Illustration Credit: Perez-Rivas Consulting LTD

AI by PRC # 36


Hello esteemed readers and LinkedIn connections! ??

Google unveils Gemini, its most advanced AI model yet developed by Google DeepMind. Gemini excels at understanding text, images, audio and more across modalities. Its models including Ultra, Pro and Nano cater to different needs from complex tasks to on-device applications. Gemini promises to revolutionize various fields like science and finance by enhancing knowledge and creativity. ?? #AI #innovation

Other topics covered include:

  • A chatbot to aid scientists with research by summarizing publications, identifying relevant information, and suggesting research directions. It leverages text and image embeddings for semantic search. ?? #research
  • An approach combining computational predictions with high-speed testing to efficiently discover powerful immunomodulators that reshape vaccine and immunotherapy effectiveness. ?? #healthcare
  • Perspectives on aligning cybersecurity initiatives with business dynamics through skilled CISO leadership and executive support. ??? #cybersecurity
  • Emerging AI-related job roles like AI Trainer, Prompt Engineer, Data Detective and Cybersecurity Analyst. ???? #AIjobs
  • A survey on Deep Active Learning strategies and challenges for large foundation models to enhance data annotation and training efficiency. ?? #machinelearning
  • An efficient privacy-preserving AI approach called PrivateLoRA that reduces communication overhead for edge devices by over 95%. ?? #privacy
  • DeepMind's Lyria model for high-quality AI-generated music and vocals, being trialed on YouTube to enhance creative processes. ?? #music


Exciting times ahead! What are your thoughts? Let me know in the comments!


AI Newsletter: Google Introduces Gemini, A Breakthrough AI Model

Google unveils Gemini, its most advanced AI model yet, as announced by CEO Sundar Pichai. Developed by Google DeepMind, Gemini represents a significant leap in AI capabilities, being multimodal and excelling in understanding text, images, audio, and more. It outperforms existing models in various benchmarks, including natural language and coding tasks. Gemini's models, including Ultra, Pro, and Nano, cater to diverse needs, from complex tasks to on-device applications. This groundbreaking model promises to revolutionize various fields, from science to finance, enhancing knowledge and creativity.

Brainstorming with a chat bot

a chatbot system designed to help scientists with research tasks like brainstorming and developing new ideas. This chatbot is trained on scientific literature to provide relevant information. Key capabilities include summarizing publications, identifying relevant information, and suggesting research directions. The system ingests scientific documents and uses text embedding to retrieve contextual passages, helping formulate helpful responses while avoiding fabricated "nonsense" answers. It also leverages image embeddings for semantic search over figures and diagrams from papers. Overall, this shows potential to accelerate science by embedding domain knowledge into AI to aid discovery.


Efficient Discovery of Potent Immunomodulators: Combining Computational Predictions with Rapid Testing

Scientists made a computer model to predict molecules that could change innate immune responses. They tested 139,998 molecules to find ones altering signaling of NF-kB and IRF, which manage immunity. By training the model and experimenting in cycles, they explored the library, testing only 2% of molecules. They found very strong immunomodulators, including molecules suppressing NF-kB 15 times, boosting it 5 times, and boosting IRF 6 times. One molecule caused 3 times more IFN-β, an antiviral protein. This shows efficient discovery of powerful immunomodulators by combining computer predictions with high-speed testing to reshape vaccine and immunotherapy effectiveness.


?? Forbes Insights: Navigating Cybersecurity Transformation in Corporations

In a compelling piece, JC Gaillard sheds light on the complex cybersecurity landscape faced by large organizations. Gaillard points out that many cybersecurity projects falter due to insufficient accountability and a short-term outlook. He emphasizes the necessity of aligning cybersecurity with business dynamics, warning against deprioritizing security for quick gains. The key, he suggests, is a transformative approach that values process and people over technology. Effective cybersecurity hinges on skilled CISOs backed by supportive executives who champion these initiatives. Gaillard also underscores the importance of management expertise, personal influence, and political savvy for CISOs to lead successful security transformations. ???


AI in the Workforce: 10 Emerging Job Roles

The Workable article outlines ten new job roles emerging due to AI advancements. These include AI-specific positions such as AI Trainer, Prompt Engineer, and AI Auditor, focusing on communication, training, and ethical regulation of AI. Roles like Machine Manager and Data Detective are crucial for managing AI systems and analyzing vast data. Cybersecurity Analysts are increasingly important for countering AI-based threats. Additionally, AI Business Strategists, Data Brokers, and AI Explainers play key roles in integrating AI into business strategies and making AI technologies understandable for all.



Revolutionizing Data Annotation with Deep Active Learning


Wang et al 2023

Discover the potential of Deep Active Learning (DeepAL) in the realm of large foundation models through an insightful paper. This comprehensive survey delves into various strategies and challenges of DeepAL, highlighting its pivotal role in enhancing data annotation efficiency and selecting optimal training subsets. The paper explores diverse applications of DeepAL across various domains, offering valuable perspectives for both researchers and practitioners eager to navigate the evolving landscape of these powerful models.


PrivateLoRA: Balancing Privacy and Efficiency in Edge Computing for AI


In a compelling exploration of AI's privacy and efficiency balance, the paper "PRIVATELORA FOR EFFICIENT PRIVACY PRESERVING LLM" presents a novel LLM service paradigm. This approach, PrivateLoRA, maintains data locality on edge devices while leveraging cloud computing for heavy tasks. The key innovation is the significant reduction of communication overhead, achieved by utilizing the low rank of residual activations. PrivateLoRA demonstrates over 95% communication reduction and offers superior performance, making advanced personalization on edge devices feasible while respecting data privacy.




Google DeepMind's blog post announces Lyria, an advanced AI music generation model. Lyria, designed in collaboration with YouTube, excels in creating high-quality music with instrumentals and vocals. It's being trialed in Dream Track, a YouTube Shorts experiment, allowing creators to generate music in the style of various artists. DeepMind is also exploring music AI tools with industry professionals, aiming to enhance the creative process. Additionally, the team is focusing on responsible AI deployment, employing watermarking tools like SynthID for AI-generated audio.


要查看或添加评论,请登录

Dr. Mario Javier Pérez Rivas的更多文章

  • AI by PRC # 49

    AI by PRC # 49

    Get the latest on all things AI in our newest newsletter edition ???? Issue #49 covers: Voyager, an AI in Minecraft…

  • AI by PRC # 48

    AI by PRC # 48

    Get the latest on all things AI in our newest newsletter edition ???? Issue #48 covers: AI in Public Health and…

  • AI by PRC # 47

    AI by PRC # 47

    Get the latest on all things AI in our newest newsletter edition ???? Issue #47 covers: The European Parliament has…

  • AI by PRC # 46

    AI by PRC # 46

    Get the latest on all things AI in our newest newsletter edition ???? Issue #46 covers: In the fashion industry, models…

    2 条评论
  • AI by PRC # 45

    AI by PRC # 45

    Get the latest on all things AI in our newest newsletter edition ?????? Issue #45 covers: Elon Musk has filed a lawsuit…

  • AI by PRC # 44

    AI by PRC # 44

    Get the latest on all things AI in our newest newsletter edition ?????? Issue #44 covers: Google’s AI Image Fail: The…

    4 条评论
  • AI by PRC # 43

    AI by PRC # 43

    Get the latest on all things AI in our newest newsletter edition ?????? Issue #43 covers: Imagine a helicopter that can…

  • AI by PRC # 42

    AI by PRC # 42

    Get the latest on all things AI in our newest newsletter edition ?????? Issue #42 covers: AI will not be the destroyer…

    6 条评论
  • Ah, the classic "urgent package" scam!

    Ah, the classic "urgent package" scam!

    screenshot of the received SMS Encountered an intriguing repetition in my inbox today – the same message appeared…

  • AI by PRC # 41

    AI by PRC # 41

    Hello AI enthusiasts and LinkedIn community! ?? Get the latest on all things AI in our newest newsletter edition…

    14 条评论

社区洞察