?? Newsletter IA-ismo: October Towards New Energy Sources and Liquid Architectures

Don’t forget to like ??, share , and subscribe to keep up with the latest in AI and tech!

---------------------------------------------------------------------------------

• Audio ???: Don’t miss this version where two AIs comment on the news through Google Podcasters!??

• Spanish version at the bottom of this article.
• Audio ???: No te pierdas los comentarios de las AIs Google Podcasters comentando estas noticias!??

Welcome to IA-ismo! ??

We’re thrilled to bring you the latest developments and insights from the fast-paced world of artificial intelligence. In this edition, we highlight some of the most exciting breakthroughs and collaborations from October—covering cutting-edge AI models, major industry partnerships, and innovative technologies that are reshaping the landscape.

?? Liquid Foundation Models (LFM)

A new generation of foundational AI models is making waves as a strong alternative to Transformers. These Liquid Foundation Models (LFMs) are built on innovative principles from dynamic systems, signal processing, and linear algebra. What sets them apart? They use computational units known as "Liquids" to condense and retain information more effectively, enhancing reasoning and knowledge retention. LFMs are particularly adept at processing sequential data like video, audio, and text, all while being optimized for specific platforms—delivering both performance boosts and greater efficiency. Curious to see them in action? Try them out here: Liquid Foundation Models

?? LightRAG

Meet LightRAG, an innovative framework that integrates graph structures into language generation, optimizing text retrieval and indexing for large language models (LLMs). It’s designed as a lightweight and efficient alternative to GraphRAG, making it ideal for platforms with limited resources. LightRAG streamlines Retrieval-Augmented Generation (RAG), making it easier to deploy cutting-edge AI solutions where they’re needed most. they’re needed most. Explore it here: LightRAG on GitHub

Major Partnerships and Investments in AI

The future of AI is being shaped by powerful alliances between industry giants like Nvidia, Accenture, Microsoft, BlackRock, and Amazon. These high-stakes collaborations, with multi-million-dollar investments, are fueling the growth of AI infrastructure and driving groundbreaking innovation across sectors.

• Nvidia and Accenture: This dynamic duo is helping businesses embrace AI by training 30,000 Accenture employees on Nvidia technologies, enabling them to support clients in implementing next-gen AI solutions.
• Microsoft and BlackRock: Together, they’ve launched a $30 billion investment fund focused on AI infrastructure in the U.S., with plans to extend into emerging sectors like nuclear power.

Models Derived from Llama 3.1

The AI landscape continues to evolve rapidly, with new models and tools pushing the boundaries of what's possible. This month, several groundbreaking models built on the Llama 3.1 architecture are making waves.

• Nemotron 70B: Built by Nvidia and leveraging SterrLM regression reward modeling, this model could potentially outperform even GPT-4.0. It’s creating a lot of buzz and is now available for experimental use on Hugging Face.Curious to see how it performs? Try it out here: Hugging Face .
• Ichigo: Another impressive Llama 3.1-based model, Ichigo is transforming human-AI interaction with its real-time responses in local languages, all under 150ms latency. Thanks to FishSpeech integration, Ichigo excels at understanding human speech and delivering lightning-fast verbal responses. It’s also incredibly efficient, running locally with just 24GB of VRAM.Explore it here: GitHub | Ichigo .
• Llama 3 Tuners: Llama 3 is also powering innovative tuners like Zamba2 (with 3B parameters) and Ministral (with 8B parameters), both designed to push the limits of fine-tuning capabilities.
• Play 3.0 Mini: If you’re looking for a lightweight, multilingual text-to-speech (TTS) model, Play 3.0 Mini has you covered. It supports over 30 languages, making it a powerful tool for global communication—reliable, cost-effective, and fluent.Learn more here: Play 3.0 mini .

Innovative AI Applications and Tools

AI innovation is accelerating the development of new tools and applications, transforming the way we work, communicate, and create.

• Mapify: This intuitive tool makes creating mind maps a breeze, helping you visualize and organize your ideas more effectively. Perfect for brainstorming and project planning, it’s a must-try for anyone looking to boost productivity.
• Speechify: A handy Chrome extension that converts text into audio, Speechify enhances accessibility and allows you to listen to content on the go, making it perfect for multitasking and increasing productivity.
• FacePoke and Expression Editor: These fun and creative tools, available on Hugging Face, let you easily modify facial expressions in images. Whether you're looking to experiment for fun or need a professional tool, these apps offer an exciting way to manipulate images.Give them a try: FacePoke | Expression Editor .

• Stable ProjectorZ

Calling all 3D artists! Stable ProjectorZ is a powerful, free tool designed to work seamlessly with Flux and Stable Diffusion models, allowing you to generate and manipulate high-quality textures for 3D models. With the use of Automatic1111 Stable Diffusion technology, it preserves the original texture coordinates (UV) and offers a range of features like brush blending, multi-view projection, 3D inpainting, and color and contrast adjustments. Whether you're a professional or a hobbyist, Stable ProjectorZ is a must-have for your toolkit. Explore it here: Stable ProjectorZ .

• Copilot in WhatsApp: Big news—Copilot is now integrated into WhatsApp! With a simple bidi code, you can add it to your chat like any other contact, making it incredibly easy to access Copilot’s powerful AI assistance right from your messaging app. Now, your AI assistant is just a message away!

AI in Science and Medicine

Artificial intelligence is making significant strides in revolutionizing science and medicine, with AI-driven breakthroughs being recognized on the world stage.

• Nobel Prizes in Physics and Chemistry: This year’s Nobel Prizes highlight the intersection of AI and scientific discovery. Demis Hassabis and John J. Jumper from Google DeepMind, along with David Baker from the University of Washington, were honored for their pioneering work with AlphaFold, which predicts protein structures with remarkable accuracy. In Physics, John J. Hopfield and Geoffrey E. Hinton were recognized for their trailblazing contributions, with Hinton’s foundational work on neural networks and machine learning taking center stage. Interested in the latest AI-driven medical advances? We recommend reading this recent article on Surprising AI Advances in Cancer Diagnosis .

AI and Energy: The Alliance with Nuclear Power

AI is teaming up with nuclear energy to forge new paths toward a cleaner, more sustainable future. Tech giants are leading the charge in this promising collaboration.

• Amazon’s Investment in Nuclear Power: Amazon has made a bold move by investing $500 million in X-energy, which is working to develop small modular nuclear reactors (SMRs). The goal? To meet the growing energy demands of Amazon’s data centers by 2039. This strategic shift reflects broader efforts by Microsoft, Google, and others to explore nuclear power as a sustainable solution for future energy needs.

Robotics and AI: Towards a More Automated Future

The rapid progress in robotics and AI is bringing us closer to an increasingly automated world, with innovations ranging from humanoid robots to autonomous vehicles.

• Humanoid Robots: Meet Ameca and Aze, the lifelike robots developed by Engineered Arts. These humanoids are turning heads with their human-like interactions and even a sense of humor, often showcased through witty, scripted performances that delight and captivate audiences.

• Autonomous Vehicles: At Tesla’s recent “We, Robot” event, the spotlight was on their cutting-edge creations, including the Robotaxi, Robovan, and the Optimus robot. These innovations are pushing the boundaries of automation in transportation and robotics. However, despite the impressive display, the response from shareholders has been somewhat lukewarm, showing that the road to full automation still faces some bumps.

Limitations of Generative Models

While the advancements in AI have been impressive, generative models still face significant challenges—particularly in areas like mathematical reasoning.

• Mathematical Reasoning: Large language models (LLMs) from companies like OpenAI and Meta have made great strides, but they continue to struggle with mathematical and logical tasks. According to a study by Apple, these models perform well on simple problems but encounter difficulties when questions are slightly rephrased or become more complex. As the number of clauses in a question increases, the models' performance drops sharply.

For a deeper dive into this research, check out the full study here: GSM-Symbolic Study .

Artificial Intelligence and Creativity

AI is opening up new ways of interacting with traditional media, revolutionizing how we consume and engage with content.

• Media and Entertainment: A new intelligent radio, powered by Symphoni AI, has just been launched. This innovative device doesn’t just play music; it holds interactive dialogues with users, much like a virtual assistant, while adding dynamic podcasting activities. It curates content by analyzing current trends and news, ensuring that you get relevant, up-to-the-minute information. For a fully immersive experience, it also comes with TWS Headphones, allowing for high-quality audio in every interaction.

Cybersecurity and Threats Associated with AI

As AI continues to advance, it also brings new cybersecurity challenges. Emerging threats, like deepfake fraud and “digital arrests,” are raising significant concerns across industries.

• Vulnerabilities in Services: AI-driven attacks are increasingly targeting widely-used services like Gmail. One common method involves sending fake recovery attempt notifications, followed by phone calls that seem to come from official Google numbers. These AI-generated calls are incredibly convincing, making it easier for victims to fall for the scam.

Featured AI Events and Launches

• Adobe Photoshop 2025: Adobe is preparing to launch the 2025 version of Photoshop, now equipped with AI integration through Firefly to enhance image editing capabilities, offering new levels of creativity and efficiency.
• Adobe Illustrator: Major updates to Illustrator now allow designers to create new perspectives with just a single click—saving valuable time and boosting productivity.
• Research Labs Expansion: Companies like Apple are expanding their research and development efforts, with new labs opening in Shenzhen, China, to strengthen AI research in the world's largest smartphone market.

State of AI 2024: Key Takeaways

The State of AI 2024 report offers valuable insights into the biggest trends and developments in artificial intelligence over the past year. Here are some key highlights:

• Data Center Investment: There’s been a dramatic increase in data center investments, sparking concerns about energy consumption. To address this, companies are adopting renewable or nuclear energy sources.
• Security and Ethics: As AI models become more sophisticated, so do concerns about security. The need for ethical AI development is more urgent than ever.
• New Possibilities: AI is breaking new ground in fields like medicine, quantum mechanics, and scientific exploration. Its influence is spreading across industries, from art to biology, opening doors to countless new possibilities.
• Dominance of Foundational Models: Proprietary models are starting to take a back seat as foundational models demonstrate their versatility across multiple disciplines, including mathematics, biology, and neuroscience.
• Global AI Market: The U.S. and China continue to dominate the AI market, with the combined value of AI companies now reaching an astounding $9 trillion, despite sanctions aimed at slowing China’s progress.
• Generative AI: Companies like Pica Art are showcasing the immense commercial potential of generative AI, generating significant revenue by producing content like videos.
• Regulations and Benchmarking: The European Union is stepping up its regulatory efforts, while concerns grow over the reliability of current AI benchmarks, many of which rely heavily on internet-trained data.
• AI Architectures: While Transformers remain the leading AI architecture, new specialized models like Brain LM are gaining attention for their ability to predict the future of AI.
• Research Paper Production: The U.S. and China dominate the production of AI research papers, accounting for 55% of global output and further reinforcing their leadership in the field.

Thought Leadership: Dario Amodei

Dario Amodei, CEO of Anthropic, has stirred up debate with his recent essay, where he explores the future of AI. His bold ideas—ranging from universal basic income to forming coalitions for democracy—have sparked widespread discussion. Interested in reading the full essay? You can dive into his thoughts here: Machines of Loving Grace .

• SPANISH VERSION ??

?Bienvenidos a IA-ismo! ??

Destacamos de este Octubre 2024 noticias como:

• Liquid Foundation Models (LFM) una nueva generación de modelos fundacionales de inteligencia artificial que se consolidan como otra alternativa a los Transformers. Los modelos líquidos usan una arquitectura innovadora basada en principios de sistemas dinámicos, procesamiento de se?ales y álgebra lineal. Los LFM cuentan con unidades computacionales llamadas "Liquid" que condensan información, permitiendo una mayor capacidad de conocimiento y razonamiento. Estos modelos son especialmente efectivos en el manejo de datos secuenciales como video, audio y texto, ofreciendo mejoras en rendimiento y eficiencia al ser optimizados para plataformas específicas. Puedes probarlos desde la plataforma https://www.liquid.ai/liquid-foundation-models
• LightRAG es un marco innovador que mejora la generación de lenguaje mediante la integración de estructuras de grafos en los procesos de indexación y recuperación de texto. Dise?ados para optimizar los procesos de Retrieval-Augmented Generation (RAG) para modelos de lenguaje grande (LLM) esta optimización es la alternativa ideal a Graphrag para despliegue en recursos limitados. https://github.com/HKUDS/LightRAG?tab=readme-ov-file

Grandes asociaciones e Inversiones en IA

Las asociaciones entre Nvidia, Accenture, Microsoft, BlackRock y Amazon están impulsando el futuro de la inteligencia artificial. Colaboraciones multimillonarias en infraestructura y desarrollo de IA, respaldando proyectos innovadores y potencian el crecimiento del sector.

• Nvidia Corp y Accenture: se asocian para ayudar a empresas clientes a adoptar herramientas de IA. Accenture ha capacitando a 30,000 empleados en el uso de estas tecnologías de Nvidia para ayudar a los clientes a implementarla.
• Microsoft y BlackRock: Crearon un fondo de inversión de $30,000 millones enfocado en la infraestructura de IA en Estados Unidos, incluyendo posibles inversiones en energía nuclear.

Modelos derivados de Lama 3.1

El panorama de la IA está evolucionando rápidamente con lanzamientos y mejoras de modelos y herramientas revolucionarias.

• Nemotron 70B: Podría superar a GPT-4o en rendimiento. Este modelo desarrollado por Nvidia sobre la arquitectura de Lama 3.1 usa un modelado de recompensa de regresión SterrLM. Liberado en Hugging Face para uso experimental. https://huggingface.co/chat/
• Ichigo: Otro modelo de IA basado en Llama 3.1 que aprende a escuchar en tiempo real y en lenguaje local, revolucionando la interacción humana-IA. Ichigo es capaz de entender el habla humana y responder verbalmente con baja latencia (menos de 150ms), gracias a la integración de FishSpeech. Se puede correr localmente con 24 VRAM. https://github.com/homebrewltd/ichigo https://ichigo.homebrew.ltd/
• En la misma línea se han desarrollado afinados de Llama 3 como Zamba2 y Ministral con 3 y 8B.
• Se lanza Play 3.0 mini, un modelo TTS multilingüe ligero, confiable y rentable que puede hablar en más de 30 idiomas. https://play.ht/news/introducing-play-3-0-mini/

Aplicaciones Innovadoras y Herramientas de IA

La innovación en IA está impulsando el desarrollo de nuevas aplicaciones y herramientas que están transformando la forma en que trabajamos.

• Mapify: Herramienta para crear mapas mentales fácilmente.
• Speechify: Extensión para Chrome que convierte texto en audio, mejorando tu accesibilidad y la productividad.
• FacePoke y Expression Editor en Huggingface te ayudan a cambiar las expresiones de rostros en imágenes https://huggingface.co/spaces/jbilcke-hf/FacePoke https://huggingface.co/spaces/fffiloni/expression-editor

• Stable ProjectorZ gratuito , trabaja con modelos flux y stable diffusion. Stable Projectorz es una herramienta gratuita dise?ada para generar y manipular texturas de alta calidad para modelos 3D utilizando la tecnología de Automatic1111 Stable Diffusion. Preservando las coordenadas de textura (UV) originales, con características como la mezcla de capas mediante pincel, proyección multivista, inpainting 3D, y ajustes de color y contraste. https://stableprojectorz.com/
• Copilot ya está integrado en Whatsapp, con este bidi puedes a?adirlo para chatear como si fuera un nuevo contacto.

IA en Ciencia y Medicina

La inteligencia artificial está comenzando a transformar ciencia y medicina

• Premios Nobel en Física y Química: Otorgados a Demis Hassabis, John J. Jumper de Google DeepMind, y David Baker de la Universidad de Washington, por avances en IA aplicada a la predicción de estructuras proteicas con AlphaFold. Y en Física se ha concedió a John J. Hopfield y Geoffrey E. Hinton por sus contribuciones fundamentales al desarrollo de las redes neuronales artificiales y el aprendizaje automático.
• Recomendamos este artículo reciente con avances sorprendentes en inteligencia artificial y medicina https://bluehealthcare.es/innovaciones-medicas-verano-2024/#Inteligencia_Artificial_en_el_diagnostico_del_cancer

IA y Energía: La Alianza con la Energía Nuclear

La colaboración entre IA y energía nuclear está abriendo nuevas oportunidades para un futuro energético más limpio y sostenible.

• Amazon: Invertió $500 millones en X-energy para desarrollar peque?os reactores nucleares modulares (SMR), buscando satisfacer la creciente demanda energética de los centros de datos en 2039. La apuesta por la energía nuclear está siendo una de las soluciones elegidas por otras empresas del sector como Microsoft y Google.

Robótica e IA: Hacia un Futuro más Automatizado

Los avances en robótica y IA están conduciendo a un futuro cada vez más automatizado, desde robots humanoides hasta vehículos autónomos.

• Robots Humanoides: Ameca y Aze, desarrollados por Engineered Arts, ganan popularidad por sus interacciones humorísticas y similares a las humanas bajo guiones divertidos.

• Vehículos Autónomos: Tesla presentó el Robotaxi, Robovan y el robot Optimus en su evento "We, Robot". Sin embargo los accionistas no han reaccionado como se esperaba.

Limitaciones de los modelos Generativos

Aunque la IA está avanzando a pasos agigantados, todavía enfrenta desafíos y limitaciones significativas.

• Razonamiento Matemático: El razonamiento matemático y lógico sigue siendo un desafío para los modelos de lenguaje grandes (LLMs), según un estudio de Apple. Aunque los modelos de OpenAI y Meta pueden manejar tareas matemáticas simples, enfrentan dificultades cuando se les presentan variaciones menores en la formulación de las preguntas. La performance de los LLMs disminuye drásticamente a medida que aumenta la complejidad de las preguntas, medida en términos de número de cláusulas. Los investigadores de Apple desarrollaron el benchmark GSM-Symbolic, que utiliza plantillas simbólicas para generar preguntas matemáticas. Este enfoque reveló que los LLMs actuales carecen de la capacidad de realizar un razonamiento lógico auténtico y su rendimiento se ve afectado negativamente por peque?os cambios en la entrada.. https://arxiv.org/pdf/2410.05229

Inteligencia Artificial y Creatividad

La IA está impulsando nuevas formas de interacción con los medios tradicionales.

• Medios de Comunicación y Entretenimiento: Se ha lanzado una radio inteligente impulsada por Symphoni AI . Esta radio puede mantener diálogos interactivos con los usuarios, similar a como lo haría un asistente virtual, a?adiendo actividades de podcaster. La IA analiza tendencias y noticias actuales para proporcionar contenido relevante y atractivo. Además el dispositivo cuenta con Auriculares TWS para una experiencia auditiva inmersiva

Ciberseguridad y Amenazas Asociadas a la IA

La ciberseguridad es un desafío creciente en la era de la IA, con amenazas como el fraude mediante deepfakes y los "arrestos digitales".

• Vulnerabilidades en Servicios: Ataques impulsados por IA ponen en riesgo servicios como Gmail. Los atacantes envían notificaciones de intentos de recuperación de cuenta de Gmail, seguidas de llamadas que parecen provenir de números oficiales de Google. Estas llamadas son generadas por IA y suenan muy educadas y profesionales, lo que aumenta la probabilidad de que las víctimas caigan en la trampa.

Eventos y Lanzamientos Destacados en IA

• Adobe Photoshop 2025: Lanzamiento con integración de IA (Firefly) para mejorar la experiencia de edición de imágenes.
• Adobe también ha incluido mejoras significativas en su software de Illustrator, permitiendo crear nuevas perspectivas con un solo click lo que ahorra tiempo a los dise?adores.
• Laboratorios de Investigación: Apertura de laboratorios y centros de investigación, como el de Apple en Shenzhen, China, que marca una significativa expansión de sus capacidades de investigación y desarrollo (I+D) en el mercado de smartphones más grande del mundo.

Informes y blogs

State of AI 2024 Resumen del Estado de la IA 2024:

El informe anual State of AI destaca algunos puntos clave que han sido tendencia el a?o pasado

1. Inversión en Data Centers: Aumento exponencial en la inversión en data centers, con crecientes preocupaciones sobre el consumo energético y la adopción de fuentes renovables o nucleares.
2. Seguridad y ética: Creciente debate sobre la seguridad de los modelos de IA y la necesidad de instituciones que gestionen los riesgos asociados.
3. Nuevas Posibilidades: La IA está revolucionando campos como la medicina y la exploración científica, desde el arte hasta la mecánica cuántica.
4. Dominio de los Modelos Fundacionales: Los modelos propietarios están perdiendo terreno ante los modelos fundacionales, que demuestran versatilidad en diversas áreas como matemáticas, biología y neurociencia.
5. Mercado Global de IA: Estados Unidos y China dominan el mercado, con sanciones económicas que no han limitado el progreso chino. El valor total de las empresas de IA alcanza los 9 trillones de dólares.
6. IA Generativa: Empresas como Pica Art están generando ingresos significativos a través de la creación de contenido, como videos.
7. Regulaciones y Benchmarking: La Unión Europea aplica regulaciones más estrictas, mientras que los benchmarks actuales pueden ser enga?osos debido a su entrenamiento con datos de internet.
8. Arquitecturas de IA: Los transformers siguen siendo la arquitectura predominante, con modelos especializados como Brain LM para predicciones clínicas.
9. Producción de Papers: Estados Unidos y China producen el 55% de los papers de IA a nivel mundial.

• Por su parte el CEO de Anthropic, Darío Amodei, habló en un ensayo sobre el futuro de la IA. Su opinión ha levantado suspicacias ya que propone la renta básica universal para adaptarse al futuro y una coalición para la democracia entre otras ideas.

https://darioamodei.com/machines-of-loving-grace

That’s all for this edition! Stay connected for more exciting updates by subscribing to our newsletter. Don’t forget to like, share, and subscribe to keep up with the latest in AI and tech! ??

Subscribe Now!

---------------------------------------------------------------------------------