Dec 7, 20247 minS1
<p>0:00 Introduction </p> <p>0:20 Limitations of traditional SfM and SLAM techniques.</p> <p>0:57 Shortcomings of existing neural network methods.</p> <p>1:07 MegaSaM's approach: balance of accuracy, speed, and...
education
AI Illuminated
Hosted by The AI Illuminators · education · EN-CA
About this Podcast:
A new way to keep up with AI research. Delivered to your ears. Illuminated by AI.Part of the GenAI4Good initiative.
Pitch Analysis
Required Pod Score for this podcast. PitchCentric checks your profile against host openness, topical fit, and audience signals before you generate a pitch.
Contact path
Public web form
Booking probability
19%
Guest openness
Selective
80/100
Required Score
Sign up to generate a grounded pitch for AI Illuminated.
Signup to Generate a PitchQuestions about AI Illuminated
What is AI Illuminated about?
A new way to keep up with AI research. Delivered to your ears. Illuminated by AI.Part of the GenAI4Good initiative.
Who hosts AI Illuminated?
AI Illuminated is hosted by The AI Illuminators. The show has published 25 episodes to date.
What has AI Illuminated covered recently?
Recent episodes include "MegaSaM: Accurate, Fast, and Robust Structure and Motion from Casual Dynamic Videos" (7 December 2024, 7 minutes), "SVDQuant: Absorbing Outliers by Low-Rank Components for 4-Bit Diffusion Models" (12 November 2024, 7 minutes), and "D3RoMa: Disparity Diffusion-based Depth Sensing for Material-Agnostic Robotic Manipulation" (11 November 2024, 15 minutes).
How often does AI Illuminated publish?
AI Illuminated publishes roughly a few times a week, with episodes averaging 10 minutes across 25 episodes.
How hard is it to get booked on AI Illuminated?
Booking probability for AI Illuminated is 19%, and the host is rated Selective on guests. AI Illuminated requires a Pod Score of 80 or higher. AI Illuminated's contact route on record: public web form.
Who should pitch AI Illuminated?
The strongest fits for AI Illuminated are Scientific Research & Development. AI Illuminated has an education audience.
About the host
The AI Illuminators hosts AI Illuminated, an education podcast with 25 episodes published.
Host email not verified for this podcastSign in to unlock →
Recent episodes
Our AI reads these to draft pitches. Use them as grounding for a pitch that cites a real guest and a specific topic.
Nov 12, 20247 minS1<p>[00:00] SVD-Quant: 4-bit diffusion model quantization</p> <p>[00:27] Challenge: Outlier sensitivity in 4-bit quantization</p> <p>[00:59] Solution: Smoothing + SVD approach</p> <p>[01:37] Technical: SVD's role in...
Episode #23
D3RoMa: Disparity Diffusion-based Depth Sensing for Material-Agnostic Robotic Manipulation
Nov 11, 202415 minS1<p>[00:00] Intro</p> <p>[00:18] Current limitations in depth-sensing technology</p> <p>[00:56] D3RoMa's diffusion model approach to depth estimation</p> <p>[01:47] Integration of geometric constraints in the model</p>...
Nov 9, 202411 minS1<p>[00:00] Intro</p> <p>[00:21] Key problem: Poor generalization in robotic learning</p> <p>[00:51] HPT: New transformer architecture for robotics</p> <p>[00:59] Core components of HPT architecture</p> <p>[01:44] Scale...
Nov 4, 20247 minS1<p>[00:00] Introduction to Hover: Neural Whole Body Controller for Humanoids</p> <p>[00:15] Problem: Current controllers lack versatility across tasks</p> <p>[00:50] Human motion imitation as a unified control...
View More
Nov 3, 20249 minS1<p>[00:00] Intro</p> <p>[00:24] Tackles RL challenges using a visual backbone, efficient RL, and human feedback.</p> <p>[01:20] Pretrained backbone boosts stability and exploration efficiency.</p> <p>[02:06] RLPD...
Nov 2, 20248 minS1<p>[00:00] Paper intro: Zero-shot robotic manipulation via local policies</p> <p>[00:26] Key challenges: Limited generalization and sim-to-real transfer</p> <p>[01:03] Local policies: Task decomposition through...
Episode #18
MENTOR: Mixture-of-Experts Network with Task-Oriented Perturbation for Visual Reinforcement Learning
Oct 30, 20249 minS1<p>[00:00] Introduction to Mentor system for visual RL</p> <p>[00:29] Problem: Sample inefficiency in robotic learning</p> <p>[00:59] Innovation: Mixture of Experts (MoE) architecture</p> <p>[01:55] Results: MoE...
Episode #17
SkillMimicGen: Automated Demonstration Generation for Efficient Skill Learning and Deployment
Oct 29, 20249 minS1<p>[00:00] SkillGen: AI system for robotic learning and automation.</p> <p>[00:19] Core Innovation: Automated dataset generation from minimal human input.</p> <p>[01:12] Skill Segmentation: Smart system for breaking...
Oct 28, 20247 minS1<p>[00:00] Intro to LVSM: Novel transformer for view synthesis</p> <p>[00:14] Problems with existing 3D synthesis methods</p> <p>[00:59] LVSM architecture: encoder-decoder vs decoder-only</p> <p>[01:41] Performance...
Oct 27, 202416 minS1<p>[00:00] Introduction to 3D Gaussian tracking for robotic manipulation</p> <p>[00:26] Limitations of current video prediction methods</p> <p>[01:11] Advantages of 3D Gaussian representation</p> <p>[02:04] Graph Neural...
Episode #14
SPIRE: Synergistic Planning, Imitation, and Reinforcement Learning for Long-Horizon Manipulation
Oct 25, 20248 minS1<p>[00:00] Introduction</p> <p>[00:20] Core limitations in robot manipulation: challenges with RL and IL</p> <p>[01:08] SPIRE's hybrid approach: combining task planning with learning methods</p> <p>[01:44] TAMP-gated...
Oct 24, 202410 minS1<p>[00:00] VILA-U: A unified visual AI model</p> <p>[00:29] Problem: Inefficiency of separate visual modules</p> <p>[01:11] Vision tower: Novel quantization approach</p> <p>[02:09] Training strategy: CLIP-based staged...
Oct 23, 20245 minS1<p>[00:00] Intro</p> <p>[00:31] Challenge: Limited multi-humanoid training data</p> <p>[00:55] CooHOI's two-phase learning framework</p> <p>[01:49] Object dynamics as implicit agent communication</p> <p>[02:25] Bounding...
Oct 22, 20247 minS1<p>[00:00] Introduction to SynFlowNet</p> <p>[00:29] Problem: AI-generated molecules often can't be synthesized</p> <p>[01:17] Solution: SynFlowNet - uses real chemical reactions</p> <p>[02:03] GFlowNets: Enables...
Episode #10
L3DG: Latent 3D Gaussian Diffusion
Oct 21, 202418 minS1<p>[00:00] Intro to L3DG for 3D modeling</p> <p>[00:32] Solving room-sized 3D scene complexity</p> <p>[01:36] VQ-VAE compresses 3D Gaussian representation</p> <p>[02:41] Generative sparse transpose convolution</p>...
Oct 20, 20248 minS1<p>[00:00] Intro</p> <p>[00:33] Combining transformers & diffusion models</p> <p>[01:12] Key design: Scalable attention blocks (AdaLN)</p> <p>[02:30] Efficient observation tokenization</p> <p>[03:45] DiT Block policy...
Oct 19, 202416 minS1<p>[00:00] Introduction to EgoAllo system</p> <p>[00:38] Challenges in egocentric motion estimation</p> <p>[01:20] Importance of spatial/temporal invariance</p> <p>[02:11] Comparison of conditioning...
Oct 19, 20249 minS1<p>[00:00] Intro</p> <p>[00:28] Limitation of existing unified models</p> <p>[00:57] Janus's decoupled visual encoding solution</p> <p>[01:18] Advantages of decoupling</p> <p>[02:03] Janus architecture</p> <p>[02:50]...
Episode #6
One Step Diffusion via Shortcut Models
Oct 19, 20245 minS1<p>[00:00] Introduction</p> <p>[00:23] Computational cost of traditional diffusion models</p> <p>[00:59] Reducing iterations in image generation</p> <p>[01:06] Shortcut models</p> <p>[01:39] Training process and...
Recent guests
Recent guests on AI Illuminated. Study who booked and why before you pitch.
O
Oier Mees
University of California, Berkeley
S
Sudeep Dasari
Carnegie Mellon University
Best industries to pitch AI Illuminated for
PitchCentric classifies AI Illuminated under these industries. Guests from them are the closest fit, and each links to the best podcasts in that industry.
Similar Podcasts
Shows with the most semantically similar episode content. Pitch one, pitch all; producers cluster.
Podcasts like AI Illuminated
Is this podcast yours and you'd like to remove or correct details? Request removal or email privacy@pitchcentric.com.
Questions about the free trial? See the pricing FAQ