Gemini 1.5 Pro is super impressive. I am feeding it @karpathy
's new lesson on Tokenization, and I can easily: – Summarize the video.
– Identify the most important concepts.
– List action items for further exploration. I broke the video down into 30-minute chunks.
@skirano
-
Gemini 1.5 Pro Analyzes Tokenization Video Content
By
–
-
Four-Step AI Image Generation Breakthrough Achieves Consumer-Ready Quality
By
–
I am shocked at how good this is in just 4 steps. The future is so bright. I think for AI to really break into consumer habits, we need the best compromise on quality and speed, and this is getting really close. Way better than LCM or Turbo.
-
Nous-Hermes-2 Quantized to 4-bit for MLX Apple Silicon
By
–
I just quantized this amazing model to 4-bit, with support for the MLX platform so you can run it super fast on Apple Silicon https://
huggingface.co/mlx-community/
Nous-Hermes-2-Mistral-7B-DPO-4bit-MLX
… -
Gemini 1.5 Pro long-context codebase testing access request
By
–
Hey Jeff, I would love access! I want to test Gemini 1.5 Pro when it comes to long-context codebases.
-
Sora invites rollout begins for video generation
By
–
When the Sora invites roll out pic.twitter.com/xrBcoaGaT4
— Pietro Schirano (@skirano) 18 février 2024When the Sora invites roll out
-
Word Positioning in AI Prompts: Does Order Affect Generation?
By
–
I am very curious about the prompt's word positioning related to the temporal effect. Like, if you mentioned the giant cat at the start, would we have seen that early in the scene?
-
Physics Engine Advances Enable Scientific Simulation Applications
By
–
So good! The physics engine looks really solid here. I am excited about the possibility of using this tech to simulate more scientifically specific scenarios. Would love an invite btw!
-
AI Tool Performs Flawlessly for Coding Tasks Overnight
By
–
It's REALLY good. I was using it for coding all night last night, and it didn't get one thing wrong.
-
Mistral-Next LLM Drops: GPT-4 Level Coding Performance Achieved
By
–
Ok, so… today is crazy. There's a mysterious Mistral-Next model that just dropped, and I am testing it for coding—it's GOOD. Possibly GPT-4 level. I even ran a few tests that I asked GPT to evaluate the outputs, and it gave Mistral-Next a perfect score.
-
Prototype Resolution Scaling Techniques for AI Model Optimization
By
–
Also, I believe there is an entire world of proto results leading to better outcomes that we haven't even started to explore, like when they are saying they can prototype at smaller resolutions and then upscale to a higher one. I can see how we could apply this same concept to