AI Twitter Bookmarks - Q4 2022 - Non-Technical

The number of AI papers on arXiv per month grows exponentially with doubling rate of 24 months.

How can we cope with this? AI itself can help, by predicting & suggesting new research directions.

Predicting the Future of AI with AI: https://t.co/FB3CBzFOfN pic.twitter.com/2sxc17LZHp

— Mario Krenn (@MarioKrenn6240) October 4, 2022

Presets make a huge difference in the @StableDiffusion #Photoshop plugin. Here are examples of Astrophotography and Architecture. Now you can import and export in CSV format. I should be able to release this week. pic.twitter.com/jtNTOgaOLh

— Christian Cantrell (@cantrell) December 27, 2022

Bird SQL — Twitter search powered by OpenAI Codex.

Stroke your vanity. Read the single least appreciated Elon Musk tweet, currently at 3 likes. Find points of agreement between yourself and Gary Marcus.https://t.co/p5SFaD8nSo pic.twitter.com/O1U1VT1LSd

— Riley Goodside (@goodside) December 15, 2022

I spent the weekend playing with ChatGPT, MidJourney, and other AI tools… and by combining all of them, published a children’s book co-written and illustrated by AI!

Here’s how! 🧵 pic.twitter.com/0UjG2dxH7Q

— Ammaar Reshi (@ammaar) December 9, 2022

My girlfriend envisioned a world where anyone could become the hero of the story.

That's why we launched a project called storytimes(.)ai

Now your children can become the main characters.

This is how it works: 🧵 pic.twitter.com/Y9nHfft9GQ

— Max Hertan (@maxhertan) November 27, 2022

This task took me 30 minutes to work through in my head from inception to working notebook. It took ChatGPT 20 seconds and I did this all on my phone, pasted into @Replit mobile app in the middle of the English countryside. 2 years ago you could barely setup a venv this fast. pic.twitter.com/Q8Toajobdj

— Callum (@Callum__Kemp) December 4, 2022

Future Diffusion: A fine-tuned Stable Diffusion 2.0 model trained on high quality 3D images with a futuristic Sci-Fi theme.

An early prototype of a Dreambooth fine-tuned #StableDiffusion2 base 512x512 model.https://t.co/zIZ7hX3u6f pic.twitter.com/FujaKhz3jb

— hardmaru (@hardmaru) November 25, 2022

Shipped: you can now semantically search the latent space of all your liked/bookmarked tweets, browsing history, highlights, and more with @SehnApp.

Have been playing around with it all day and is a ton of fun. Some previews: pic.twitter.com/9CFUqOJ1BF

— Eli (@ebadgio_) October 31, 2022

I read a lot of books for my podcast & blog.

But often I can't finding the particular passage I'm looking for.

Ctrl-F doesn't work unless you know the exact phrase.

So I built search for ebooks using OpenAI's embeddings API.

Link below to use.

Works surprisingly well! pic.twitter.com/UFDjDhZ507

— Dwarkesh Patel (@dwarkesh_sp) October 30, 2022

My jaw hit the floor when this finished. @LumaLabsAI 's app is fully legit y'all even in a beta state. Took this at the CU Museum and with the fossil under glass. But bloody hell at this detail y'all. #NeRF technology is gonna blow up the 3D world. More soon ✌️ pic.twitter.com/51WcFkhBmQ

— Kynthic.Eth (@KynthicMedia) October 30, 2022

i'm announcing the company we've been building.

it's called rime.

here's a tiny demo of what truly generative text-to-speech can accomplish. pic.twitter.com/WbaShCGcth

— lily clifford (@lilyjclifford) October 17, 2022

When using GPT-3 for classification, rather than generating a single label you can use the token logprobs to get the probability of each possible label. This gives you confidence scores on your answers. pic.twitter.com/IHvhOEiOCs

— Riley Goodside (@goodside) October 15, 2022

Dramatron paper out!

Co-Writing Screenplays and Theatre Scripts with Language Models: An Evaluation by Industry Professionals

📝: https://t.co/i4PqH4m4iI

Dramatron is system that enables writers to co-create theatre / film scripts with a language model, @DeepMind's Chinchilla pic.twitter.com/NdyDZgCjWz

— Kory Mathewson (@korymath) September 30, 2022

I won second place in the @bellingcat hackathon at the weekend, with a GitHub Actions/Issues-based tool for archiving captions from online video using @OpenAI Whisper

I wrote about the project here https://t.co/17eps2HYit

— Simon Willison (@simonw) September 30, 2022

Created a Speech-to-Youtube Search App with @Gradio , @huggingface Spaces & OpenAI's Whisper.
Compared to Google voice input and, in this instance, Whisper(OpenSource🤩) performs way better than Google's service! See the video attached and explore the App! https://t.co/4D1snUtEL1 pic.twitter.com/BkfeeNpSDp

— Yuvi (@yvrjsharma) September 25, 2022

Write tweets with the help of AI using the style of any account. Type the username, write the information of the tweet and let #GPT3 do the rest. Built with @Fermat_ws . Try it out here: https://t.co/qMb96cJH7N pic.twitter.com/N44tk63CYK

— Marc Aguilar (⨍) (@marcagve) September 23, 2022

Just published the code for my NodeJS palette generator tool with Stable Diffusion:

→ https://t.co/QvFIj3Wstt

Below: comparing different color quantization modules for palette extraction. pic.twitter.com/yXj8KxWkCY

— Matt DesLauriers (@mattdesl) September 13, 2022

This morning I've been playing around with the @huggingface's #StableDiffusion Textual Inversion Concepts Library https://t.co/ecWiFMEKak working out how best to train my own concepts.
Would be great to force consistency with image generation to improve storytelling. #AIart pic.twitter.com/2ZLbcjjw12

— TomLikesRobots (@TomLikesRobots) September 11, 2022

To learn how to make your own chatbots in this style using OpenAI's GPT‑3 API, see my minimal example here: https://t.co/qss5jg3jSM

— Riley Goodside (@goodside) December 26, 2022

McKinsey bot is almost complete.

First we practiced with alignment vs. autonomy

And now ChatGPT can make its own 🤣 pic.twitter.com/XcPKZmN53m

— John Cutler (@johncutlefish) December 23, 2022

pic.twitter.com/hwo26pgcfE

— aashiq (@aashiq) December 5, 2022

I guess GPT-3 is old news, but playing with OpenAI’s new chatbot is mindblowing. https://t.co/so1TuXMQB0

We’re witnessing the death of the college essay in realtime. Here’s the response to a prompt from one of my 200-level history classes at Amherst

Solid A- work in 10 seconds pic.twitter.com/z1KPxiAc1O

— Corry Wang (@corry_wang) December 1, 2022

OpenAI's new ChatGPT explains the worst-case time complexity of the bubble sort algorithm, with Python code examples, in the style of a fast-talkin' wise guy from a 1940's gangster movie: pic.twitter.com/MjkQ5OAIlZ

— Riley Goodside (@goodside) December 1, 2022

I kinda like this one even more! pic.twitter.com/0rioxNdfov

— Nick Moran (@NickEMoran) November 30, 2022

Using zero-shot chain-of-thought combined with the "format trick", Victoria can evaluate Johnny against her entire checklist with a single GPT-3 prompt. Applying CoT again, she obtains a single binary classification and learns that Johnny is not her person. pic.twitter.com/Ueh5hqTIfK

— Riley Goodside (@goodside) October 27, 2022

"You are GPT-3", revised: A long-form GPT-3 prompt for assisted question-answering with accurate arithmetic, string operations, and Wikipedia lookup. Generated IPython commands (in green) are pasted into IPython and output is pasted back into the prompt (no green). pic.twitter.com/CFVkufPjhf

— Riley Goodside (@goodside) October 17, 2022

GPT-3 adding comments to a neural network written in NumPy, then explaining it with (poorly rhymed) poetry, then rewriting it in PyTorch. Playground link: https://t.co/5VOKXuXR82 pic.twitter.com/Oac0W50Nj0

— Riley Goodside (@goodside) September 2, 2022

I taught #GPTChat how to write Midjourney prompts giving it examples of my previous prompt then had it write me a prompt and this is how it went the prompt and the result. #midjourney #AIArt pic.twitter.com/fSCNYWgGC6

— Shaun Harrison (@shaun_harrison) December 22, 2022

#ChatGPT, invent me an art movement.#MidJourney, show it to me. 😲 pic.twitter.com/fElq5IUgve

— Guy Parsons (@GuyP) December 26, 2022

Good news, we can now all go back to skeuomorphic app icons, thanks to AI ✨ pic.twitter.com/jJjLXelx3v

— Felix Krause (@KrauseFx) December 12, 2022

Someone on Reddit is using stable diffusion to take selfies throughout time - here they are with the Trojan horse pic.twitter.com/yJeh7qQ5Fp

— Andrew Carr (@andrew_n_carr) December 10, 2022

More 768 generations. Tap to expand.

That's it for now. #happythanksgiving fellow travelers. Time to build some magnet towers with the kids.

If you're new here, follow for more #airt and #solarpunk #architecture ✨ pic.twitter.com/bvPBEKLQIK

— Gregory Wieber (@dreamwieber) November 24, 2022

These No Man's Sky Screenshots Do Not Exist #StableDiffusion2 pic.twitter.com/AJMRP9g3X1

— Peter Baylies (@pbaylies) November 27, 2022

Some raw output from #StableDiffusion2

Like actually try it folks, it's only going to get better from here, each of these was generated in a couple of seconds.

We will release some optimised settings soon. 🚀✨ pic.twitter.com/GIMbEAXZmb

— Emad (@EMostaque) November 26, 2022

Best AI art I’ve seen yet was generated in MidJourney by Johnny Derren from the prompt: “Jodorowsky’s TRON”. Holy cow. pic.twitter.com/IdzUjNqTnc

— Greg Newkirk (@nuekerk) November 25, 2022

#midjourney V4 is exceptionally good at generating vector art-type images with clean line art. I can't stop. pic.twitter.com/8jnzBcYX2Q

— surea.i (@sureailabs) November 16, 2022

Astronaut Deaths 😱 pic.twitter.com/3idWy3DpDy

— Andrei (@Riabovitchev) November 12, 2022

Park lights 1

Novel, repeatable faces via model training.#Dreambooth - #StableDiffusion #VisionsofChaos - #CoDeformer - #Topaz #Gigapixel - #LuminarNeo#AIart #machinelearning #ai #artificialintelligence #digitalart #aiartwork #aiartcommunity pic.twitter.com/aujADyBE19

— Maui Mauricio (@Maui_Mauricio) November 7, 2022

Knolling of Gandalf's bag.

According to DALL·E and MidJourney#dalle2 #MidJourney #LOTR pic.twitter.com/djo0i9soJx

— Merzmensch Kosmopol🧑‍🎨🤖 (@Merzmensch) October 19, 2022

stableboost is an awesome new (personal favorite) Stable Diffusion WebUI, great work @tall! It lifts the interaction to population level - you generate many (hundreds/thousands) of prompt/param variations, then search/sort through them by visual look&feel of whatever you're after https://t.co/dVUjifdt0L

— Andrej Karpathy (@karpathy) November 2, 2022

I trained an AI with the best @Poolsuite pictures from their Instagram and here are the results: pic.twitter.com/fyvmHMIg4l

— PromptHero - Stable Diffusion, DALL-E & Midjourney (@prompthero) November 1, 2022

"Oh, you like music? Name every genre.
Oh, let me guess - 'neuropsychedfunkjazzbassfusionmetalwave' wasn't on the list?
OBVIOUSLY you aren't a real music fan 🥱👋
I bet you've never even synthesized your own band members yet 🙄 & dont @ me unless you've generated your own genre!" pic.twitter.com/2zDZzg1R2b

— deKxi@sigmoid.social (@deKxi) October 22, 2022

AI unlocks unlimited product photography. 🤯

Recent advances have made it possible for merchants to turn mobile photos into high quality product images + ad creative.

AI can now learn and reproduce a product on demand. More…👇[1/10]#ai #aiart #stablediffusion #dreambooth pic.twitter.com/sNrRLtIIa1

— Russ Maschmeyer (@StrangeNative) October 11, 2022

Anglofuturist Space Oxbridge pic.twitter.com/Xyzjch5Oqx

— Andrew Cusack (@cusackandrew) October 4, 2022

Made some kitchens today. #midjourneyart #interiordesign pic.twitter.com/e0KNuF90ta

— Olli Siebelt 🇺🇦 (@olli101) September 20, 2022

Day 5 of training my AI pet to do my work for me ✨🤖 pic.twitter.com/dy86lRmJZ7

— Alycia Rainaud (@maalavidaaa) September 8, 2022

imagen video: high definition video generation with diffusion models

paper: https://t.co/kbayCcen2r
blog: https://t.co/ft0v4Mmqm7 pic.twitter.com/Rm7PlFxD1u

— AK (@_akhaliq) October 5, 2022

DreamFusion: Text-to-3D using 2D Diffusion
paper: https://t.co/euVCkNPzB0
abs: https://t.co/FO7nqjLrQA
project page: https://t.co/BTvt1WQqIi

DeepDream on a pretrained 2D diffusion model enables text-to-3D synthesis pic.twitter.com/20zIElXJDN

— AK (@_akhaliq) September 29, 2022

"On the Horizon: Interactive and Compositional Deepfakes"

So you probably know that neural nets can generate videos of people saying stuff they never said. But Microsoft’s chief science officer articulates two threats beyond this that could be way worse: [1/11] pic.twitter.com/U8Hevnh2GG

— Davis Blalock (@davisblalock) September 24, 2022

Currently develop a colab version of dreamfields which could generate 3D model and 360° videos by text & image prompt input. Here is a flower tree video / 3D model generated by text prompts.#dreamfields3D #AIart #3dmodeling #ACG #Blender3d pic.twitter.com/Mu8ntgSm8h

— Simon Meng (@meng_shengyu) September 5, 2022

Try it yourself:https://t.co/ELjf0uNdfN

— Zaesar 📽️ AIFILMS.ai 🤖 (@zaesarius) November 7, 2022

We're seeing a bunch of music generation models coming out, so in 𝚐𝚛𝚊𝚍𝚒𝚘 𝟹.𝟷𝟺, we've released a way for you to create beautiful waveform visualizations

𝚐𝚛.𝚖𝚊𝚔𝚎_𝚠𝚊𝚟𝚎𝚏𝚘𝚛𝚖(𝚊𝚞𝚍𝚒𝚘=..., 𝚋𝚐_𝚒𝚖𝚊𝚐𝚎=...)

Create videos like this, in Python, fast ⚡️ pic.twitter.com/ffGi6X39Vu

— Gradio (@Gradio) December 15, 2022

"AudioLM: a language modeling approach to audio generation", w/ @zalanborsos and colleagues.

AudioLM achieves high-quality and long-term consistency, from audio only. Meaningful speech w/o text, consistent piano w/o MIDI.

📃: https://t.co/TfrW9hui7i
🔊: https://t.co/c0vYQZGUSK pic.twitter.com/bcHy1e3dUz

— Neil Zeghidour (@neilzegh) September 8, 2022

Today we introduce an ML-generated sensory map that relates thousands of molecules and their perceived odors, enabling the prediction of odors from unseen molecules and providing a potential tool to address global health issues like insect-borne disease. https://t.co/wmiq6wPKv5

— Google AI (@GoogleAI) September 6, 2022