RESEARCH - AI Dynamics

Data Storytelling: A Crucial Skill for AI Era

By

@saboo_shubham_

–

07 January 2023 8h44

Data storytelling is indeed a crucial thing!!

→ View original post on X — @saboo_shubham_

7 January 2023

ChatGPT vs. Claude: Comparing AI Models with a Literary Prompt

By

@goodside

–

07 January 2023 3h54

Side-by-side comparison: @OpenAI
's ChatGPT vs. @AnthropicAI
's Claude Each model is asked to compare itself to the machine from Stanisław Lem's "The Cyberiad" (1965) that can create any object whose name begins with "n":

→ View original post on X — @goodside

7 January 2023

Denoising Errors Transformed Into Image Generation via Diffusion

By

@pmddomingos

–

07 January 2023 3h51

When denoising goes haywire and makes stuff up, call it diffusion and use it for image generation.

→ View original post on X — @pmddomingos

7 January 2023

Claude AI Model Compared to ChatGPT for Factual Accuracy

By

@goodside

–

07 January 2023 3h14

Compared to ChatGPT, @AnthropicAI
's new model, Claude, feels somewhat less likely to make up facts, but the tendency is still there. E.g., when discussing what happened on each season of ABC's Lost, its memory is even more confused than most people's:

→ View original post on X — @goodside

7 January 2023

Beyond Text: Multimodal AI’s Stunning Potential Beyond GPT

By

@gdb

–

06 January 2023 23h50

Wrong lesson from GPT paradigm — text is all you need. Other modalities will be incredibly stunning. Can feel it already from DALL-E, which can do something quite useful despite its nascent language ability.

→ View original post on X — @gdb

6 January 2023

Training Costs Stable Diffusion 512 768 4K Models Data

By

@levelsio

–

06 January 2023 23h37

SD 512 cost $600k to train, SD 768 probably $1.3M and 4K would cost $10M I think to train also where do you source billions of images

→ View original post on X — @levelsio

6 January 2023

Training costs for stable diffusion models scale exponentially

By

@levelsio

–

06 January 2023 23h37

SD 512 cost $600k to train, SD 768 probably $1.3M and 4K would cost $10M I think to train also where do you source billions of images

→ View original post on X — @levelsio

6 January 2023

Stable Diffusion Resolution Limitations and Future Improvements

By

@levelsio

–

06 January 2023 23h19

People don't realize that Stable Diffusion is trained on tiny pics of 512x512px! Which is why you're getting so many artefacts, it's like an AI with really bad vision that needs glasses Reso will exponentially increase: the smallest details can be properly simulated then

→ View original post on X — @levelsio

6 January 2023