Videos de Ciencia
Videos sobre avances científicos y descubrimientos.
Ciencia 255 videos
IDL Lect 3C Interpretable Models Before Deep Learning | Feature Importance, Naïve Bayes & Dec. Trees
In this lecture, Prof. Dilip K. Prasad explains how classical machine learning models provided interpretable explanations before deep learning became dominant. The lecture introduces non-graphical explanation methods such as feature importance, probability-based explanations, and decision rules. It discusses why models such as Naïve Bayes, Random Forests, and decision trees are easier to inspect compared with many deep neural networks, and how their structure can help users understand predictions. The lecture also explains important limitations, including why feature importance does not necessarily mean causality and why complex models may require additional explanation tools. It connects classical interpretable models with modern Explainable AI approaches, including surrogate models and other XAI strategies. This lecture is part of a course series on interpretable and explainable artificial intelligence.
Clustering Explained — k-Means, Hierarchical & DBSCAN Visually | Master AI & ML Ep 14
Clustering is where machine learning goes unsupervised — no labels, no correct answers, just data and the question: is there hidden structure here? In this episode we animate k-means from scratch, tackle the deceptively hard problem of choosing k, explore hierarchical clustering and DBSCAN, and confront the hardest question in all of clustering: how do you know if it worked? In this episode: → The unsupervised shift — what changes when there's no target variable → Real use cases: customer segmentation, anomaly detection, document grouping, image compression → k-means step by step — animated convergence from random centroids to stable clusters → Choosing k — the elbow method and silhouette score explained → When k-means fails — elongated clusters and concentric rings → Hierarchical clustering — the dendrogram explained visually → DBSCAN — density-based clustering with built-in outlier detection → Evaluating clustering without ground truth — the hardest problem in unsupervised ML This is Episode 14 of Master AI & Machine Learning — Module 3: Core ML Algorithms, Episode 4 of 6. ────────────────────────────── 📋 FULL COURSE PLAYLIST ⬅ Ep 13 — Decision Trees & Random Forests ➡ Ep 15 — Neural Networks Explained 🌐 TechnovativeAI → www.technovativeai.com ────────────────────────────── ⏱ TIMESTAMPS 00:00 — Hook: the unsupervised shift 00:30 — What clustering is for — real use cases 01:30 — k-means step by step — animated 03:30 — Choosing k — elbow method and silhouette score 04:45 — When k-means fails 05:45 — Hierarchical clustering and DBSCAN 07:00 — How do you evaluate clustering without ground truth? 08:00 — Next episode & CTA ────────────────────────────── Series of Thoughts · Presented by TechnovativeAI #Clustering #kMeans #UnsupervisedLearning #DBSCAN #MachineLearning #MLalgorithms #LearnAI #TechnovativeAI #SeriesOfThoughts #DataScience clustering machine learning, k-means explained, k-means algorithm visual, hierarchical clustering explained, DBSCAN explained, unsupervised learning explained, customer segmentation ML, elbow method k-means, silhouette score explained, clustering without labels, anomaly detection clustering, how k-means works, dendrogram explained, clustering algorithms compared, unsupervised ML tutorial, TechnovativeAI, Series of Thoughts, learn AI, ML algorithm visual, k-means vs DBSCAN
O ROBÔ QUE A UBTECH CONSTRUIU PARA SUBSTITUIR A SOLIDÃO
A UBTECH acaba de lançar o U one — o primeiro robô humanoide de tamanho real à venda para consumidores comuns. Com pele de silicone, cabelo real, 88 graus de liberdade e inteligência artificial que reconhece emoções, o U1 é a aposta mais ousada do mercado de robótica em décadas. Neste vídeo, analisamos em profundidade o que está por trás desse lançamento: quem é a UBTECH, como funciona a tecnologia do U1 da linha UWORLD, o que ele consegue fazer — e o que ainda não consegue —, os impactos concretos para idosos, cuidadores e o mercado global de robótica, e as questões éticas que esse produto levanta sobre privacidade, apego emocional e regulação. Em menos de uma semana, mais de 2.100 pessoas fizeram reserva a 30 mil dólares. As entregas começam em setembro de 2026. O mundo da robótica de consumo nunca mais será o mesmo. Gasgoo Auto News — "UBTECH Launches Pre-sales for Full-size Hyper-bionic Humanoid Robot on JD.com" https://autonews.gasgoo.com/articles/icv/ubtech-launches-pre-sales-for-full-size-hyper-bionic-humanoid-robot-on-jdcom-2061828693847244801 Pandaily — "UBTECH's First Full-Size Biomimetic Humanoid Robot Surpasses 1,000 Pre-Orders in 3 Days" https://pandaily.com/ubtech-biomimetic-humanoid-robot-presale-jun2026 Origin of Bots — "UBTECH Robotics to Launch Full-Size Hyper-Realistic Humanoid Robot U1 on June 30" https://www.originofbots.com/news/ubtech-robotics-to-launch-full-size-hyper-realistic-humanoid-robot-u1-on-june-30 TechEBlog — "UBTECH's UWORLD U1 Humanoids Bring Emotional Companionship Into Focus" https://www.techeblog.com/ubtech-uworld-u1-robots-humanoid/ Mike Kalil — "UBTECH's New UWorld U1 Lifelike Humanoid Robots" https://mikekalil.com/blog/ubtech-u1-lifelike-humanoids/ Let's Data Science — "UBTECH launches UWORLD U1 humanoid robot companions" https://letsdatascience.com/news/ubtech-launches-uworld-u1-humanoid-robot-companions-39ee02ca Humanoid Guide — "UWORLD U1 humanoid robot opens preorders in China" https://humanoid.guide/ubtech-opens-preorders-for-uworld-u1-humanoid-robot-in-china/ INFO Guangdong — "UBTECH Robotics leads global market with 2025 revenue and full-sized humanoid robot sales hitting record highs" https://info.newsgd.com/node_a42013034a/a0c724a624.shtml Gasgoo — "UBTECH 2025 Report Card: Revenue from Full-Size Humanoid Robots Grows Over 22-Fold" https://autonews.gasgoo.com/articles/icv/ubtech-2025-report-card-revenue-from-full-size-humanoid-robots-grows-over-22-fold-2039900685372407808 PR Newswire — "UBTECH Humanoid Robot Walker S2 Begins Mass Production and Delivery, with Orders Exceeding 800 Million Yuan" https://www.prnewswire.com/news-releases/ubtech-humanoid-robot-walker-s2-begins-mass-production-and-delivery-with-orders-exceeding-800-million-yuan-302616924.html JMIR Aging — "Evaluating the Effectiveness of Digital Social Robots in Reducing Loneliness Among Community-Dwelling Older Adults in Japan" https://aging.jmir.org/2025/1/e74422 The Gerontologist / Oxford Academic — "Wired for companionship: a meta-analysis on social robots filling the void of loneliness in later life" https://academic.oup.com/gerontologist/article/65/12/gnaf219/8268528 Frontiers in Computer Science — "Long-term effect of the absence of a companion robot on older adults: A preliminary pilot study" https://www.frontiersin.org/journals/computer-science/articles/10.3389/fcomp.2023.1129506/full #Ubtech #Robotica #InteligenciaArtificial
🏥 IA na Medicina para Quem Tem Paciente em Casa: O Jeito Certo de Usar
Fique sempre atualizado e cuide da sua saúde com o @NEUROCIRURGIABR! Nosso compromisso é oferecer informações confiáveis sobre neurocirurgia, condições neurológicas e as principais notícias em saúde. Descubra conteúdos que abrangem: Neurocirurgia: Tumores cerebrais, aneurismas, doenças da coluna vertebral e muito mais. Saúde em Foco: Dicas essenciais de prevenção, bem-estar e as novidades da medicina. Vacinação: Informações sobre campanhas, a importância da imunização e a luta contra doenças. Doenças da Atualidade: Atualizações sobre surtos, epidemias e medidas de prevenção para enfermidades comuns. Autor: Julio Pereira - Neurocirugião - São Paulo 🇧🇷 CRM/SP 163.113 Consulta: http://drjuliopereira.com.br/ Consultório: (11)99503-8838 (WhatsApp) / (11)4200-2300 Atuando no Hospital Sírio-Libanês e na BP – A Beneficência Portuguesa de São Paulo. A neurocirurgia é a especialidade médica que se dedica ao tratamento de doenças do sistema nervoso central e periférico, abrangendo desde recém-nascidos até adultos e, recentemente, fetos. Atuamos em uma ampla gama de condições, como: Hidrocefalia Tumores cerebrais e da coluna vertebral Doenças vasculares (aneurismas, malformações arteriovenosas) Doenças degenerativas (doença de Parkinson) Traumatismos cranioencefálicos e lesões raquimedulares Epilepsia Dores crônicas Distúrbios do movimento (distonias, síndrome de Tourette) Alguns transtornos psiquiátricos (psicocirurgia) Nosso objetivo é oferecer soluções cirúrgicas inovadoras e eficazes, visando melhorar a qualidade de vida dos pacientes. http://weneuro.com/ http://radiocirurgia.com.br/ Nossas Mídias Sociais Blog: http://www.neurocirurgiabr.com Instagram: http://www.instagram.com/neurocirurgiabr Twitter: http://www.twitter.com/neurocirurgiabr Youtube: http://www.youtube.com/neurocirurgiabr TikTok: https://www.tiktok.com/@neurocirurgiabr Meus livros publicados: 1) Após um tumor cerebral 2) Escrito em letra de médico 3) Ultimas Cartas de um médico 4) Suturando Ideias 5) Neurocirurgia para o estudante de medicina Quer ter acesso grátis aos livros? É só enviar um e-mail para neurocirurgiabr@gmail.com que mandamos em PDF. Informações sobre alguma doença Neurológica ou sobre Neurocirurgia veja nos sites: 🤕 Tumor Cerebral https://tumor-cerebral.com/ 😷 Radiocirurgia http://radiocirurgia.com.br/ 🧠 Câncer https://www.neurocancer.com/ 🫀 Aneurisma Cerebral https://aneurismacerebral.org/ 👨🏽⚕️ Coluna e Dor https://cirurgianacoluna.com/ 👨🏻🏫 Informações para Estudantes de Medicina http://weneuro.com/ 👨🔬 Artigos Médicos: https://neurocirurgiabr.com/ ⚠️ AVISO LEGAL: As informações disponíveis neste vídeo são estritamente informativas, não substituindo o parecer médico/profissional. Sempre consulte o seu médico sobre qualquer assunto relativo à sua saúde e aos seus tratamentos e medicamentos.
Unitree Robotics, fabricante chinesa de robôs humanoides, abre capital na Bolsa de Valores de Xangai
Robôs humanoides 🤖 A Unitree Robotics, fabricante chinesa de robôs humanoides, teve aprovação em apenas 73 dias para abrir capital na Bolsa de Valores de Xangai. É o processo mais rápido do ano na China, superando o recorde anterior em mais de duas semanas. A empresa faturou 1,7 bilhão de yuans em 2025 e entregou mais de 5.500 robôs humanoides, com receita do setor crescendo 335% em um ano. Hoje, mais da metade da receita da Unitree já vem desses robôs. Parceria entre #BrasildeFato e #ChinaAcademy Acesse o nosso site para mais notícias: www.brasildefato.com.br INSCREVA-SE em nosso canal: www.youtube.com/@brasildefato APOIE O BRASIL DE FATO: https://bdf.onl/y/7KTG Instagram: instagram.com/brasildefato TikTok: tiktok.com/@brasildefato_ LinkedIn: linkedin.com/company/brasildefato Bluesky: bsky.app/profile/brasilfato.bsky.social Threads: threads.net/@brasildefato Facebook: facebook.com/brasildefato
NOTICIAS SEMANALES SOBRE IA. SIN HYPE. CON IMPACTO POLÍTICO, ECONÓMICO Y SOCIAL.
# Noticias Semanales sobre IA. Sin Hype ¿Qué pasó esta semana en inteligencia artificial y por qué importa? En este resumen analizamos algunas de las noticias más relevantes sobre IA, salud, regulación, privacidad, automatización y futuro del trabajo. Sin promesas exageradas y sin alarmismo. Solo contexto, impacto y criterio para comprender cómo la inteligencia artificial está transformando nuestras instituciones, empresas y sociedades. 📌 En este episodio: • La IA ayudó a detectar más de 40 enfermedades raras en Boston Children's Hospital y México comenzó a utilizar inteligencia artificial para fortalecer la detección temprana del cáncer de mama. • Donald Trump firmó una orden ejecutiva para que el gobierno pueda revisar voluntariamente modelos avanzados de IA antes de su lanzamiento y evaluar posibles riesgos de seguridad. • Meta retrocedió parcialmente en un plan para utilizar datos laborales de sus empleados para entrenar sistemas de inteligencia artificial, reabriendo el debate sobre privacidad y consentimiento en el trabajo. • Meta también presentó nuevos agentes empresariales para WhatsApp, Instagram y Messenger capaces de atender clientes, recomendar productos y automatizar ventas. • Argentina debate proyectos para atraer inversiones en infraestructura de IA y habilitar sociedades que puedan operar mediante agentes de inteligencia artificial. Más allá de cada noticia, aparece una tendencia común: la inteligencia artificial ya no es solamente una cuestión tecnológica. También es una conversación sobre salud, regulación, empleo, soberanía tecnológica, privacidad y concentración económica. En Instituto Futuro Humano creemos que comprender la IA requiere algo más que aprender herramientas. Requiere desarrollar criterio para entender su impacto en las personas, las organizaciones y la sociedad. 💬 ¿Cuál de estas noticias te parece más importante y por qué? Dejanos tu opinión en los comentarios. 📚 Instituto Futuro Humano Ayudamos a profesionales de Latinoamérica a comprender, usar y pensar críticamente la inteligencia artificial desde una mirada humana, estratégica y aplicada. Sumate a nuestra comunidad de skoo: https://www.skool.com/ai4good-academy-1487/about Y a nuestra comunidad de whatsapp: https://chat.whatsapp.com/B5NzFrEaQvb4aTlWJ6W4ua #InteligenciaArtificial #IA #NoticiasIA #ChatGPT #OpenAI #MetaAI #WhatsAppBusiness #Automatizacion #SaludDigital #CancerDeMama #EnfermedadesRaras #Trump #RegulacionIA #Privacidad #FuturoDelTrabajo #TransformacionDigital #Tecnologia #Argentina #AgentesIA #InstitutoFuturoHumano
LLMs Explained in 20 Minutes | The Transformer Behind ChatGPT, Gemini & Claude
🚀 LLMs Explained in 10 Minutes | The Transformer Behind ChatGPT, Gemini & Claude Ever wondered how ChatGPT, Gemini, Claude, and other AI assistants actually work? In this video, we'll break down Large Language Models (LLMs) in the simplest way possible. You'll learn how AI evolved from traditional neural networks to the revolutionary Transformer Architecture, the breakthrough that powers modern AI. We'll also explore the Attention Mechanism, the core idea that allows LLMs to understand context, focus on important words, and generate human-like responses. What You'll Learn ✅ What is an LLM (Large Language Model)? ✅ Why RNNs struggled with long context ✅ How Transformer Architecture works ✅ What is the Attention Mechanism? ✅ Why Transformers changed AI forever ✅ How ChatGPT, Gemini, and Claude generate responses ✅ The foundation behind modern Generative AI Whether you're a student, developer, cloud engineer, AI enthusiast, or preparing for AI interviews, this video will help you understand the fundamentals of LLMs without complicated math. 🔥 If you enjoy AI, Generative AI, Agentic AI, Google Cloud, Gemini, MCP, and modern AI architectures, make sure to subscribe for more content. #LLM #ChatGPT #Gemini #Claude #Transformer #GenerativeAI #ArtificialIntelligence #MachineLearning #AIExplained #TechTrapture Playlists Google Agent Development Kit (ADK) https://www.youtube.com/playlist?list=PLLrA_pU9-Gz2HwepRUVpq1TEPuYWo_fSi Learn Airflow https://www.youtube.com/playlist?list=PLLrA_pU9-Gz3i8qw6yakrfJzx75W_vVaH Learn Google Cloud in 2025 https://youtube.com/playlist?list=PLLrA_pU9-Gz2OnBoICkewd9-Fc9Mi0nm7&si=8kkB3ct5wDHCMkoi Data Engineering Hands-on Projects https://www.youtube.com/playlist?list=PLLrA_pU9-Gz2DaQDcY5g9aYczmipBQ_Ek Looking to get in touch? Drop me a line at vishal.bulbule@techtrapture.com Linkedin https://www.linkedin.com/in/vishal-bulbule/ Medium Blog https://medium.com/@VishalBulbule Github Source Code https://github.com/vishal-bulbule
How LLMs Understand your Prompts: Tokenization & Embeddings | Chapter 05
What are vector embeddings and tokenization, and how do they let an LLM understand meaning? This video explains tokenization, embeddings, vector dimensions, cosine similarity and positional embeddings - with a hands-on coding demo. ===== In this video, you will learn ===== • What tokenization is and the main types of tokenizers (incl. Byte Pair Encoding) • What vector embeddings are, and what vectors & dimensions actually mean • How an LLM captures meaning using embeddings • How cosine similarity measures how "close" two pieces of text are • Positional embeddings - how models know word order • A hands-on Python demo: tokenizer + embeddings in code This is Part 5 of the GenAI Fundamentals series - for data engineers, developers, and anyone learning how AI language models actually work. ===== Chapters ===== 00:00 - Introduction 00:32 - What is Tokenization and Vector Embeddings? 03:41 - What are Vectors and Dimensions? 06:37 - Why Vectors matter for LLMs? 07:59 - How meaning is captured using Embedding? 12:30 - What is Cosine Similarity? 14:08 - Types of Tokenizers 17:50 - (Hands on) Coding for Tokenizer and Embeddings 26:48 - What are Positional Embeddings? ===== Links ===== Google's "Attention is all You Need" Paper - https://arxiv.org/pdf/1706.03762 Groq Cloud - https://console.groq.com/home GPT Tokenizer - https://platform.openai.com/tokenizer Google Collab - https://colab.research.google.com/ Byte Pair Encoding (BPE) - https://www.geeksforgeeks.org/nlp/byte-pair-encoding-bpe-in-nlp/ Github Code - https://github.com/subhamkharwal/genai-for-data-engineers/blob/master/codes/genai_chap05.ipynb ===== Other Playlists ===== Checkout all other playlists on Data Engineering 👇🏻 https://www.youtube.com/@easewithdata/playlists ===== GitHub Repo ===== https://github.com/subhamkharwal https://github.com/subhamkharwal/genai-for-data-engineers ===== Connect with ME ===== LinkedIn - https://www.linkedin.com/in/subhamkharwal Medium - https://subhamkharwal.medium.com ===== Hashtags ===== #VectorEmbeddings #Tokenization #LLM #GenerativeAI #genai #dataengineering #python #neuralnetworks #machinelearning
How to Build an LLM From Scratch in PyTorch
I implemented a Transformer/LLM from scratch in PyTorch and trained it as a small GPT like language model. The goal of this video is to understand how modern LLMs work under the hood by implementing a language model from scratch, inspired by the Attention Is All You Need paper. We start with the dataset and tokenization, then build embeddings, positional encoding, masked multi head attention, decoder blocks, and the full PyTorch training loop.This is a decoder only Transformer trained with next token prediction, similar in spirit to GPT. For the dataset I use Yu Gi Oh cards, which makes the final output a bit more fun than another Shakespeare example. Code: https://github.com/vossenwout/llm-from-scratch Paper: https://arxiv.org/abs/1706.03762 Timestamps: 00:00 - Building a Transformer from Scratch in PyTorch 01:15 - Why train on Yu-Gi-Oh cards? 03:05 - Downloading and preprocessing the dataset 08:25 - Building tokenizers for our language model 12:40 - Byte Pair Encoding tokenizer explained 19:15 - Creating a PyTorch dataset 20:35 - Next-token prediction explained 21:45 - Transformer input and output shapes 27:40 - Decoder-only Transformers explained 30:55 - Token embeddings and positional encoding 39:10 - Implementing the Transformer decoder 41:30 - Multi-head attention explained 49:40 - Causal masking for GPT-style models 57:05 - Implementing multi-head attention in PyTorch 1:00:20 - Optimizing attention (vectorization) 1:05:00 - Residual connections, dropout, and layer norm 1:09:25 - Feed-forward networks in Transformers 1:13:25 - Stacking decoder blocks 1:16:40 - Building the PyTorch training loop 1:21:35 - Cross entropy loss and optimization 1:29:50 - Validation loss, perplexity, and checkpoints 1:32:00 - Training the Transformer model 1:34:25 - Training results and overfitting 1:35:45 - Building the inference script 1:38:45 - Sampling tokens with temperature 1:41:20 - Generating AI Yu-Gi-Oh cards 1:44:00 - Opening AI-generated booster packs 1:48:30 - Final thoughts #pytorch #llm #deeplearning
Day - 03 : GEN AI + LLM + RAG + Agentic AI Overview by Mr. Ashok
Artificial Intelligence is changing the way software applications are built. In this session, Mr. Ashok explains the most important AI concepts every student and developer should know. In this video, you will learn: ✅ What is Generative AI? ✅ What are Large Language Models? ✅ How RAG works in real-time applications ✅ What is Agentic AI? ✅ Difference between LLM, RAG, and AI Agents ✅ Real-world use cases of Gen AI ✅ Career opportunities in AI This session is useful for students, freshers, working professionals, Java developers, Python developers, and anyone who wants to start learning AI from basics. 📌 Watch the full session and start your AI learning journey today. #GenAI #LLM #RAG #AgenticAI #ArtificialIntelligence #AIOverview #GenerativeAI #PromptEngineering #AIForBeginners #AshokIT #MrAshok #PythonAI #AICareer #MachineLearning #TechLearning