Today's newsletter features an interview with Harvard Medical School's Dr. Kohane (@zakkohane) on the integration of AI and medicine. (GPT-4 Passed the US Medical Licensing Board with 90%+ accuracy) Drops in 2 hours – Sign up here:
@aibreakfast
-
Meta launches the Segment Anything Model (SAM)
By
–
Meta's Segment Anything Model (SAM) recognizes different elements in a picture and creates individual "masks" for them with stunning accuracy
— AI Breakfast (@AiBreakfast) 10 avril 2023
This (open source) model has been trained on a massive dataset of 11 million images and 1.1 billion masks
Demo: https://t.co/Wk5nkxnCz5 pic.twitter.com/cvjpOCx6PVMeta's Segment Anything Model (SAM) recognizes different elements in a picture and creates individual "masks" for them with stunning accuracy. This (open source) model has been trained on a massive dataset of 11 million images and 1.1 billion masks. Demo: https://segment-anything.com/demo
-
Anthropic takes aim at OpenAI with a $5 billion plan
By
–
Anthropic has a $5B, 4-yr plan to overtake OpenAI. Their own pitch deck says they're building a model called “Claude-Next” that's 10x more capable than ChatGPT. Anthropic was founded by OpenAI's former VP of Research, and Google owns a 10% stake in them. Can they do it?
-
Custom ChatGPT API with Google Search
By
–
A custom ChatGPT API with Google Search connects GPT-3.5 to the internet for real-time information retrieval.
— AI Breakfast (@AiBreakfast) 8 avril 2023
(Uses your own API keys. Project in beta.)
Link: https://t.co/IeFCB2KRbH pic.twitter.com/R19wj843pFA custom ChatGPT API with Google Search connects GPT-3.5 to the internet for real-time information retrieval. (Uses your own API keys. Project in beta.) Link: https://
heygpt.chat -
Evolution of AI in 12-24 Months
By
–
Reminder: ChatGPT was released just 4 months ago. What will we be looking at in 12 or 24 months? Today: Text-to-Text
Text-to-Image
Text-to-Speech 2024/25: Text-to-Product
Text-to-Solution
Text-to-Entertainment Beyond that, we might swap the "Text" part to "Thought" -
3D model generating plausible views from a single image
By
–
This is a 3D-aware diffusion-based model for generating diverse and "plausible" novel views from a single input image, incorporating geometry priors (assumptions about the geometric properties of objects) to build the world from the image.
-
New 3D diffusion models from a single image
By
–
Hyperrealistic 3D scenes from a single image 👀
— AI Breakfast (@AiBreakfast) 7 avril 2023
New research from NVIDIA & Stanford: Generative Novel View Synthesis with 3D-Aware Diffusion Models
These videos are generated from a single photo ↓ pic.twitter.com/TI37A6EHfhHyperrealistic 3D scenes from a single image New research from NVIDIA & Stanford: Generative Novel View Synthesis with 3D-Aware Diffusion Models These videos are generated from a single photo ↓
-
AI generating 3D frames in real time
By
–
This is awesome:
— AI Breakfast (@AiBreakfast) 6 avril 2023
Frame it in 3D, and AI creates it.
Very cool project from https://t.co/Y5kWvocD6y pic.twitter.com/FBCxGKgP18This is awesome: Frame it in 3D, and AI creates it. Very cool project from http://
plask.ai -
Uncensored GPT-4 revealed in an OpenAI article
By
–
If you want to see what the uncensored, unrestricted GPT-4 was like, OpenAI published a paper about it. They gave it some very graphic prompts – and showed how the early GPT-4 responded. It's wild. (Jump to pg. 44) link to paper: https:// cdn.openai.com/papers/gpt-4-s ystem-card.pdf …