AI Today

About AI Today

Welcome to AI Today TechTalk – where we geek out about the coolest, craziest, and most mind-blowing stuff happening in the world of Artificial Intelligence! 🚀 This is your AI crash course, snackable podcast-style. Think of it as your weekly dose of cutting-edge research, jaw-dropping breakthroughs, and “Wait, AI can do THAT?!” moments. We take the techy, brain-bending papers and news, break them down, and serve them up with a side of humor and a whole lot of fun. Whether you’re an AI superfan, a tech wizard, or just someone who loves knowing what’s next in the tech world, this channel has s

Publisher

AI Today Tech Talk

Category

technology

Language

en

Explicit

No

First episode

27 Nov 2024

Latest episode

7 Feb 2025

Latest episodes

30 episodes in the feed.

1. 

   7 Feb 2025

   SFT Memorizes, RL Generalizes: A Comparative Study of Foundation Model | #ai #2025 #genai #google

   Paper: https://arxiv.org/pdf/2501.17161 (https://www.youtube.com/redirect?event=video_description&redir_token=QUFFLUhqbjluM0EybmhlZVZhdFpGalFINlpkZWlLYlc0QXxBQ3Jtc0tsNHBDQXhjVzR6Um5lVDBBNmd0YmJva05yVUpJSjRKVUpVbnpCUHVYQ1psZlZldFRYRHByUHlPWGU5RHdlQ0dxdjZZUGl4ZzhBWlNSRUV2cU1fZWtjLVh4dGtiX1NqN01pY1hvRmU2SnFyaGJqd000NA&q=https%3A%2F%2Farxiv.org%2Fpdf%2F2501.17161&v=QyZ8NzIVhSs) This research paper compares supervised fine-tuning (SFT) and reinforcement learning (RL) for post-training foundation models. Using novel and existing tasks involving arithmetic and spatial reasoning, the study finds that RL promotes better generalization to unseen data, unlike SFT which tends to memorize training data. Further analysis reveals RL enhances visual recognition capabilities in multimodal models, while SFT aids in stabilizing RL training by improving output formatting. The paper also explores the impact of increased inference-time computation on generalization. #ai (https://www.youtube.com/hashtag/ai), #artificialintelligence (https://www.youtube.com/hashtag/artificialintelligence), #arxiv (https://www.youtube.com/hashtag/arxiv), #research (https://www.youtube.com/hashtag/research), #paper (https://www.youtube.com/hashtag/paper), #publication (https://www.youtube.com/hashtag/publication), #llm (https://www.youtube.com/hashtag/llm), #genai (https://www.youtube.com/hashtag/genai), #generativeai (https://www.youtube.com/hashtag/generativeai), #largevisualmodels (https://www.youtube.com/hashtag/largevisualmodels), #largelanguagemodels (https://www.youtube.com/hashtag/largelanguagemodels), #largemultimodalmodels (https://www.youtube.com/hashtag/largemultimodalmodels), #nlp (https://www.youtube.com/hashtag/nlp), #text (https://www.youtube.com/hashtag/text), #machinelearning (https://www.youtube.com/hashtag/machinelearning), #ml (https://www.youtube.com/hashtag/ml), #nvidia (https://www.youtube.com/hashtag/nvidia), #openai (https://www.youtube.com/hashtag/openai), #anthropic (https://www.youtube.com/hashtag/anthropic), #microsoft (https://www.youtube.com/hashtag/microsoft), #google (https://www.youtube.com/hashtag/google), #technology (https://www.youtube.com/hashtag/technology), #cuttingedge (https://www.youtube.com/hashtag/cuttingedge), #meta (https://www.youtube.com/hashtag/meta), #llama (https://www.youtube.com/hashtag/llama), #chatgpt (https://www.youtube.com/hashtag/chatgpt), #gpt (https://www.youtube.com/hashtag/gpt), #elonmusk (https://www.youtube.com/hashtag/elonmusk), #samaltman (https://www.youtube.com/hashtag/samaltman), #deployment (https://www.youtube.com/hashtag/deployment), #engineering (https://www.youtube.com/hashtag/engineering), #scholar (https://www.youtube.com/hashtag/scholar), #science (https://www.youtube.com/hashtag/science), #apple (https://www.youtube.com/hashtag/apple), #samsung (https://www.youtube.com/hashtag/samsung), #turing (https://www.youtube.com/hashtag/turing), #aiethics (https://www.youtube.com/hashtag/aiethics), #innovation (https://www.youtube.com/hashtag/innovation), #futuretech (https://www.youtube.com/hashtag/futuretech), #deeplearning (https://www.youtube.com/hashtag/deeplearning), #datascience (https://www.youtube.com/hashtag/datascience), #computervision (https://www.youtube.com/hashtag/computervision), #autonomoussystems (https://www.youtube.com/hashtag/autonomoussystems), #robotics (https://www.youtube.com/hashtag/robotics), #dataprivacy (https://www.youtube.com/hashtag/dataprivacy), #cybersecurity (https://www.youtube.com/hashtag/cybersecurity), #digitaltransformation (https://www.youtube.com/hashtag/digitaltransformation), #quantumcomputing (https://www.youtube.com/hashtag/quantumcomputing), #aiapplications (https://www.youtube.com/hashtag/aiapplications), #aiethics (https://www.youtube.com/hashtag/aiethics), #techleadership (https://www.youtube.com/hashtag/techleadership), #technews (https://www.youtube.com/hashtag/technews), #aiinsights (https://www.youtube.com/hashtag/aiinsights), #aiindustry (https://www.youtube.com/hashtag/aiindustry), #aiadvancements (https://www.youtube.com/hashtag/aiadvancements), #futureai (https://www.youtube.com/hashtag/futureai), #airesearchers (https://www.youtube.com/hashtag/airesearchers)

2. 

   30 Jan 2025

   Deepseek Janus-Pro: Unified Multimodal Understanding and Generation | #ai #2025 #genai #deepseek

   Paper: https://github.com/deepseek-ai/Janus/blob/main/janus_pro_tech_report.pdf Github: https://github.com/deepseek-ai/Janus/tree/main?tab=readme-ov-file The paper introduces Janus-Pro, an improved multimodal model building upon its predecessor, Janus. Janus-Pro boasts enhanced performance in both multimodal understanding and text-to-image generation due to optimized training strategies, expanded datasets (including synthetic aesthetic data), and a larger model size (1B and 7B parameter versions). The architecture uses decoupled visual encoding for improved efficiency and performance across various benchmarks. Results show significant gains over previous state-of-the-art models, although limitations remain in resolution and fine detail. The code and models are publicly available. #ai, #artificialintelligence, #arxiv, #research, #paper, #publication, #llm, #genai, #generativeai, #largevisualmodels, #largelanguagemodels, #largemultimodalmodels, #nlp, #text, #machinelearning, #ml, #nvidia, #openai, #anthropic, #microsoft, #google, #technology, #cuttingedge, #meta, #llama, #chatgpt, #gpt, #elonmusk, #samaltman, #deployment, #engineering, #scholar, #science, #apple, #samsung, #turing, #aiethics, #innovation, #futuretech, #deeplearning, #datascience, #computervision, #autonomoussystems, #robotics, #dataprivacy, #cybersecurity, #digitaltransformation, #quantumcomputing, #aiapplications, #aiethics, #techleadership, #technews, #aiinsights, #aiindustry, #aiadvancements, #futureai, #airesearchers

3. 

   11 Jan 2025

   Memory Layers at Scale | #ai #2024 #genai #meta

   Paper: https://arxiv.org/pdf/2412.09764 (https://www.youtube.com/redirect?event=video_description&redir_token=QUFFLUhqa3R6amRKcUtJNWZQTjRRMjVodzU0MFVGaVF0Z3xBQ3Jtc0trMVJaeENIWXA5aU1iakV0b1VTZWxrbU5MemVjTzdZZ1ZpSDNCN2tON2lvR1ZkZERLdlRPeGZPc2s5cV83TFZXVVAwZFB2R0RuMTY5ZTN0Q2tDdDBJajFNQlQ5c1M4emRDMnZoaVJCMHVWSndOeVliNA&q=https%3A%2F%2Farxiv.org%2Fpdf%2F2412.09764&v=2wNF8m0Tdjw) This research paper explores the effectiveness of memory layers in significantly enhancing large language models (LLMs). By incorporating a trainable key-value lookup mechanism, memory layers add parameters without increasing computational cost, improving factual accuracy and overall performance on various tasks. The researchers demonstrate substantial gains, especially on factual tasks, even surpassing models with much larger computational budgets and outperforming mixture-of-experts models. They detail improvements in memory layer implementation, achieving scalability with up to 128 billion memory parameters, and discuss various architectural optimizations. The findings strongly advocate for integrating memory layers into future AI architectures. #ai (https://www.youtube.com/hashtag/ai), #artificialintelligence (https://www.youtube.com/hashtag/artificialintelligence), #arxiv (https://www.youtube.com/hashtag/arxiv), #research (https://www.youtube.com/hashtag/research), #paper (https://www.youtube.com/hashtag/paper), #publication (https://www.youtube.com/hashtag/publication), #llm (https://www.youtube.com/hashtag/llm), #genai (https://www.youtube.com/hashtag/genai), #generativeai (https://www.youtube.com/hashtag/generativeai), #largevisualmodels (https://www.youtube.com/hashtag/largevisualmodels), #largelanguagemodels (https://www.youtube.com/hashtag/largelanguagemodels), #largemultimodalmodels (https://www.youtube.com/hashtag/largemultimodalmodels), #nlp (https://www.youtube.com/hashtag/nlp), #text (https://www.youtube.com/hashtag/text), #machinelearning (https://www.youtube.com/hashtag/machinelearning), #ml (https://www.youtube.com/hashtag/ml), #nvidia (https://www.youtube.com/hashtag/nvidia), #openai (https://www.youtube.com/hashtag/openai), #anthropic (https://www.youtube.com/hashtag/anthropic), #microsoft (https://www.youtube.com/hashtag/microsoft), #google (https://www.youtube.com/hashtag/google), #technology (https://www.youtube.com/hashtag/technology), #cuttingedge (https://www.youtube.com/hashtag/cuttingedge), #meta (https://www.youtube.com/hashtag/meta), #llama (https://www.youtube.com/hashtag/llama), #chatgpt (https://www.youtube.com/hashtag/chatgpt), #gpt (https://www.youtube.com/hashtag/gpt), #elonmusk (https://www.youtube.com/hashtag/elonmusk), #samaltman (https://www.youtube.com/hashtag/samaltman), #deployment (https://www.youtube.com/hashtag/deployment), #engineering (https://www.youtube.com/hashtag/engineering), #scholar (https://www.youtube.com/hashtag/scholar), #science (https://www.youtube.com/hashtag/science), #apple (https://www.youtube.com/hashtag/apple), #samsung (https://www.youtube.com/hashtag/samsung), #turing (https://www.youtube.com/hashtag/turing), #aiethics (https://www.youtube.com/hashtag/aiethics), #innovation (https://www.youtube.com/hashtag/innovation), #futuretech (https://www.youtube.com/hashtag/futuretech), #deeplearning (https://www.youtube.com/hashtag/deeplearning), #datascience (https://www.youtube.com/hashtag/datascience), #computervision (https://www.youtube.com/hashtag/computervision), #autonomoussystems (https://www.youtube.com/hashtag/autonomoussystems), #robotics (https://www.youtube.com/hashtag/robotics), #dataprivacy (https://www.youtube.com/hashtag/dataprivacy), #cybersecurity (https://www.youtube.com/hashtag/cybersecurity), #digitaltransformation (https://www.youtube.com/hashtag/digitaltransformation), #quantumcomputing (https://www.youtube.com/hashtag/quantumcomputing), #aiapplications (https://www.youtube.com/hashtag/aiapplications), #aiethics (https://www.youtube.com/hashtag/aiethics), #techleadership (https://www.youtube.com/hashtag/techleadership), #technews (https://www.youtube.com/hashtag/technews), #aiinsights (https://www.youtube.com/hashtag/aiinsights), #aiindustry (https://www.youtube.com/hashtag/aiindustry), #aiadvancements (https://www.youtube.com/hashtag/aiadvancements), #futureai (https://www.youtube.com/hashtag/futureai), #airesearchers (https://www.youtube.com/hashtag/airesearchers)

Reach and audience

Public platform figures. Ratings count people who left a rating, not total listeners.

Spotify

5.0 / 5

1 ratings

Podcast Authority Score: 42 / 100

A composite of feed quality, social presence, YouTube performance and engagement. Read the methodology.

Quality

71

Social presence

0

YouTube

0

Engagement

32

Contact AI Today

Guest appearances

Books guests

Based on episode analysis; this does not confirm that the show is currently accepting guests.