i think it's the information-sharing/generalization part that makes what we'd call understanding happen
GENERATIVE AI
-
Everyone Should Learn to Code with AI Assistance
By
–
Everyone should learn to code with AI! At AI Fund, everyone – not just engineers – can vibe code or use AI assistance to code. This has been great for our creativity and productivity. I hope more teams will empower everyone to build with AI. Please watch the video for details. pic.twitter.com/rsGC1QSKHL
— Andrew Ng (@AndrewYNg) 3 juin 2025Everyone should learn to code with AI! At AI Fund, everyone – not just engineers – can vibe code or use AI assistance to code. This has been great for our creativity and productivity. I hope more teams will empower everyone to build with AI. Please watch the video for details.
-
Test Examples Extractability and Privacy in Saturated AI Models
By
–
– hrunning these experiments in a clean setting with perfectly deduplicated texts tells us a lot about privacy: – once capacity is sufficiently saturated, the **test examples** are slightly more extractable than the training examples — maybe extraction is a bit of a myth?
– the -
Scaling Laws Show LLMs Cannot Memorize Training Data
By
–
and finally we can compute membership inference success rate across all our models, ending up with this scaling law main takeaway: models trained on massive datasets (e.g. every LLM that comes out) can't memorize their training data there's simply not enough capacity
-
Model Memorization Fixed Capacity Training Saturation Analysis
By
–
we train all of our models until they "saturate" which usually happens around 1M steps using a very large batch size models memorize the same amount, regardless of training datasize meaning they have fixed capacity and instead "spread it thinner" when trained on more examples
-
GPT Model Capacity Analysis: Linear Trend at 3.6 Bits-Per-Parameter
By
–
we then compute the capacity of different models (GPT models with varying numbers of layers and hidden dimensions) averaged over hundreds of models in fp32, we get the following curve, indicating a linear trend of around 3.6 bits-per-parameter, regardless of the exact details:
-
AI Filmmaking Workflow and Feedback Opportunity
By
–
Pick one template, swap the nouns, run it, iterate 3×, and post the best 8 s cut. Drop your first AI-generated video below. I’ll give feedback. Let’s push the frontier of AI filmmaking together.
-
Prompt Engineering Tips for Fixing Common Generative AI Visual Artifacts
By
–
Common fails → quick fixes Jerky motion ⇒ add “cinematic 24 fps with motion interpolation.” Melting faces ⇒ include “stable identity, photoreal human, no distortion.” Lighting mismatch ⇒ “global HDR grade” or relight with Runway “Relight” filter.
-
AI-Powered Video Post-Production Workflow
By
–
Post-production flow 1. Generate A-roll clips (4–8 s each).
2. Export to DaVinci/Premiere or Capcut; slice at beat markers.
3. Add LUT + grain overlay for cohesion.
4. Upscale to 4 K with Topaz or Runway Super-Res.
5. Caption burn-in with your brand font. -
Advanced Prompt Engineering Techniques for AI Models
By
–
8. Advanced prompt tricks • Parentheses weighting: (highly detailed:1.4) boosts fidelity.
• Negative prompts: “no watermark, no distorted digits.”
• Rhyme schemes for audio: “Up-tempo hip-hop beat, 95 BPM, minor scale,” gets Veo 3’s music generator on point.