Why breakthrough technologies must be accessible

ChatGPT has exploded. Over 1 million people have used the app, and coding sites like Stack Overflow have banned its answers. It's huge.

I wouldn't have called that as an AI researcher. ChatGPT uses the same GPT-3 technology that's been around for over two years.

More than impressive technology, ChatGPT 3 shows how access makes breakthroughs usable. OpenAI has finally made people realize the power of AI by packaging GPT-3 for normal users.

We think of Thomas Edison as the inventor of the lightbulb, not because he invented it, but because he popularized it.

Going forward, AI companies that make using AI easy will thrive.

Use-case importance

Most modern AI systems use massive language models. These language models are trained on 6,000+ years of human text.

GPT-3 ate 8 billion pages, almost every book, and Wikipedia. It created an AI that can write sea shanties and solve coding problems.

Nothing new. I began beta testing GPT-3 in 2020, but the system's basics date back further.

Tools like GPT-3 are hidden in many apps. Many of the AI writing assistants on this platform are just wrappers around GPT-3.

Lots of online utilitarian text, like restaurant menu summaries or city guides, is written by AI systems like GPT-3. You've probably read GPT-3 without knowing it.

Accessibility

Why is ChatGPT so popular if the technology is old?

ChatGPT makes the technology accessible. Free to use, people can sign up and text with the chatbot daily. ChatGPT isn't revolutionary. It does it in a way normal people can access and be amazed by.

Accessibility isn't easy. OpenAI's Sam Altman tweeted that opening ChatGPT to the public increased computing costs.

Each chat costs "low-digit cents" to process. OpenAI probably spends several hundred thousand dollars a day to keep ChatGPT running, with no immediate business case.

Academic researchers and others who developed GPT-3 couldn't afford it. Without resources to make technology accessible, it can't be used.

Retrospective

This dynamic is old. In the history of science, a researcher with a breakthrough idea was often overshadowed by an entrepreneur or visionary who made it accessible to the public.

We think of Thomas Edison as the inventor of the lightbulb. But really, Vasilij Petrov, Thomas Wright, and Joseph Swan invented the lightbulb. Edison made technology visible and accessible by electrifying public buildings, building power plants, and wiring.

Edison probably lost a ton of money on stunts like building a power plant to light JP Morgan's home, the NYSE, and several newspaper headquarters.

People wanted electric lights once they saw their benefits. By making the technology accessible and visible, Edison unlocked a hugely profitable market.

Similar things are happening in AI. ChatGPT shows that developing breakthrough technology in the lab or on B2B servers won't change the culture.

AI must engage people's imaginations to become mainstream. Before the tech impacts the world, people must play with it and see its revolutionary power.

As the field evolves, companies that make the technology widely available, even at great cost, will succeed.

OpenAI's compute fees are eye-watering. Revolutions are costly.

Deep learning-based generative art is being researched. As usual, self-learning is better. Some models, like OpenAI's DALL-E 2, require registration and can only be used online, but others can be used locally, which is usually more enjoyable for curious users. I'll demonstrate the Stable Diffusion model's operation on a standard PC.

Let’s get started.

What It Does

Stable Diffusion uses numerous components:

• A generative model trained to produce images is called a diffusion model. The model is incrementally improving the starting data, which is only random noise. The model has an image, and while it is being trained, the reversed process is being used to add noise to the image. Being able to reverse this procedure and create images from noise is where the true magic is (more details and samples can be found in the paper).

• An internal compressed representation of a latent diffusion model, which may be altered to produce the desired images, is used (more details can be found in the paper). The capacity to fine-tune the generation process is essential because producing pictures at random is not very attractive (as we can see, for instance, in Generative Adversarial Networks).

• A neural network model called CLIP (Contrastive Language-Image Pre-training) is used to translate natural language prompts into vector representations. This model, which was trained on 400,000,000 image-text pairs, enables the transformation of a text prompt into a latent space for the diffusion model in the scenario of stable diffusion (more details in that paper).

This figure shows all data flow:

The weights file size for Stable Diffusion model v1 is 4 GB and v2 is 5 GB, making the model quite huge. The v1 model was trained on 256x256 and 512x512 LAION-5B pictures on a 4,000 GPU cluster using over 150.000 NVIDIA A100 GPU hours. The open-source pre-trained model is helpful for us. And we will.

Install

Before utilizing the Python sources for Stable Diffusion v1 on GitHub, we must install Miniconda (assuming Git and Python are already installed):

wget https://repo.anaconda.com/miniconda/Miniconda3-py39_4.12.0-Linux-x86_64.sh
chmod +x Miniconda3-py39_4.12.0-Linux-x86_64.sh
./Miniconda3-py39_4.12.0-Linux-x86_64.sh
conda update -n base -c defaults conda

Install the source and prepare the environment:

git clone https://github.com/CompVis/stable-diffusion
cd stable-diffusion
conda env create -f environment.yaml
conda activate ldm
pip3 install transformers --upgrade

Download the pre-trained model weights next. HiggingFace has the newest checkpoint sd-v14.ckpt (a download is free but registration is required). Put the file in the project folder and have fun:

python3 scripts/txt2img.py --prompt "hello world" --plms --ckpt sd-v1-4.ckpt --skip_grid --n_samples 1

Almost. The installation is complete for happy users of current GPUs with 12 GB or more VRAM. RuntimeError: CUDA out of memory will occur otherwise. Two solutions exist.

Running the optimized version

Try optimizing first. After cloning the repository and enabling the environment (as previously), we can run the command:

python3 optimizedSD/optimized_txt2img.py --prompt "hello world" --ckpt sd-v1-4.ckpt --skip_grid --n_samples 1

Stable Diffusion worked on my visual card with 8 GB RAM (alas, I did not behave well enough to get NVIDIA A100 for Christmas, so 8 GB GPU is the maximum I have;).

Running Stable Diffusion without GPU

If the GPU does not have enough RAM or is not CUDA-compatible, running the code on a CPU will be 20x slower but better than nothing. This unauthorized CPU-only branch from GitHub is easiest to obtain. We may easily edit the source code to use the latest version. It's strange that a pull request for that was made six months ago and still hasn't been approved, as the changes are simple. Readers can finish in 5 minutes:

• Replace if attr.device!= torch.device(cuda) with if attr.device!= torch.device(cuda) and torch.cuda.is available at line 20 of ldm/models/diffusion/ddim.py ().

• Replace if attr.device!= torch.device(cuda) with if attr.device!= torch.device(cuda) and torch.cuda.is available in line 20 of ldm/models/diffusion/plms.py ().

• Replace device=cuda in lines 38, 55, 83, and 142 of ldm/modules/encoders/modules.py with device=cuda if torch.cuda.is available(), otherwise cpu.

• Replace model.cuda() in scripts/txt2img.py line 28 and scripts/img2img.py line 43 with if torch.cuda.is available(): model.cuda ().

Run the script again.

Testing

Test the model. Text-to-image is the first choice. Test the command line example again:

python3 scripts/txt2img.py --prompt "hello world" --plms --ckpt sd-v1-4.ckpt --skip_grid --n_samples 1

The slow generation takes 10 seconds on a GPU and 10 minutes on a CPU. Final image:

Hello world is dull and abstract. Try a brush-wielding hamster. Why? Because we can, and it's not as insane as Napoleon's cat. Another image:

Generating an image from a text prompt and another image is interesting. I made this picture in two minutes using the image editor (sorry, drawing wasn't my strong suit):

I can create an image from this drawing:

python3 scripts/img2img.py --prompt "A bird is sitting on a tree branch" --ckpt sd-v1-4.ckpt --init-img bird.png --strength 0.8

It was far better than my initial drawing:

I hope readers understand and experiment.

Stable Diffusion UI

Developers love the command line, but regular users may struggle. Stable Diffusion UI projects simplify image generation and installation. Simple usage:

• Unpack the ZIP after downloading it from https://github.com/cmdr2/stable-diffusion-ui/releases. Linux and Windows are compatible with Stable Diffusion UI (sorry for Mac users, but those machines are not well-suitable for heavy machine learning tasks anyway;).

• Start the script.

Done. The web browser UI makes configuring various Stable Diffusion features (upscaling, filtering, etc.) easy:

V2.1 of Stable Diffusion

I noticed the notification about releasing version 2.1 while writing this essay, and it was intriguing to test it. First, compare version 2 to version 1:

• alternative text encoding. The Contrastive LanguageImage Pre-training (CLIP) deep learning model, which was trained on a significant number of text-image pairs, is used in Stable Diffusion 1. The open-source CLIP implementation used in Stable Diffusion 2 is called OpenCLIP. It is difficult to determine whether there have been any technical advancements or if legal concerns were the main focus. However, because the training datasets for the two text encoders were different, the output results from V1 and V2 will differ for the identical text prompts.

• a new depth model that may be used to the output of image-to-image generation.

• a revolutionary upscaling technique that can quadruple the resolution of an image.

• Generally higher resolution Stable Diffusion 2 has the ability to produce both 512x512 and 768x768 pictures.

The Hugging Face website offers a free online demo of Stable Diffusion 2.1 for code testing. The process is the same as for version 1.4. Download a fresh version and activate the environment:

conda deactivate  
conda env remove -n ldm  # Use this if version 1 was previously installed
git clone https://github.com/Stability-AI/stablediffusion
cd stablediffusion
conda env create -f environment.yaml
conda activate ldm

Hugging Face offers a new weights ckpt file.

The Out of memory error prevented me from running this version on my 8 GB GPU. Version 2.1 fails on CPUs with the slow conv2d cpu not implemented for Half error (according to this GitHub issue, the CPU support for this algorithm and data type will not be added). The model can be modified from half to full precision (float16 instead of float32), however it doesn't make sense since v1 runs up to 10 minutes on the CPU and v2.1 should be much slower. The online demo results are visible. The same hamster painting with a brush prompt yielded this result:

It looks different from v1, but it functions and has a higher resolution.

The superresolution.py script can run the 4x Stable Diffusion upscaler locally (the x4-upscaler-ema.ckpt weights file should be in the same folder):

python3 scripts/gradio/superresolution.py configs/stable-diffusion/x4-upscaling.yaml x4-upscaler-ema.ckpt

This code allows the web browser UI to select the image to upscale:

The copy-paste strategy may explain why the upscaler needs a text prompt (and the Hugging Face code snippet does not have any text input as well). I got a GPU out of memory error again, although CUDA can be disabled like v1. However, processing an image for more than two hours is unlikely:

Stable Diffusion Limitations

When we use the model, it's fun to see what it can and can't do. Generative models produce abstract visuals but not photorealistic ones. This fundamentally limits The generative neural network was trained on text and image pairs, but humans have a lot of background knowledge about the world. The neural network model knows nothing. If someone asks me to draw a Chinese text, I can draw something that looks like Chinese but is actually gibberish because I never learnt it. Generative AI does too! Humans can learn new languages, but the Stable Diffusion AI model includes only language and image decoder brain components. For instance, the Stable Diffusion model will pull NO WAR banner-bearers like this:

V1:

V2.1:

The shot shows text, although the model never learned to read or write. The model's string tokenizer automatically converts letters to lowercase before generating the image, so typing NO WAR banner or no war banner is the same.

I can also ask the model to draw a gorgeous woman:

V1:

V2.1:

The first image is gorgeous but physically incorrect. A second one is better, although it has an Uncanny valley feel. BTW, v2 has a lifehack to add a negative prompt and define what we don't want on the image. Readers might try adding horrible anatomy to the gorgeous woman request.

If we ask for a cartoon attractive woman, the results are nice, but accuracy doesn't matter:

V1:

V2.1:

Another example: I ordered a model to sketch a mouse, which looks beautiful but has too many legs, ears, and fingers:

V1:

V2.1: improved but not perfect.

V1 produces a fun cartoon flying mouse if I want something more abstract:

I tried multiple times with V2.1 but only received this:

The image is OK, but the first version is closer to the request.

Stable Diffusion struggles to draw letters, fingers, etc. However, abstract images yield interesting outcomes. A rural landscape with a modern metropolis in the background turned out well:

V1:

V2.1:

Generative models help make paintings too (at least, abstract ones). I searched Google Image Search for modern art painting to see works by real artists, and this was the first image:

I typed "abstract oil painting of people dancing" and got this:

V1:

V2.1:

It's a different style, but I don't think the AI-generated graphics are worse than the human-drawn ones.

The AI model cannot think like humans. It thinks nothing. A stable diffusion model is a billion-parameter matrix trained on millions of text-image pairs. I input "robot is creating a picture with a pen" to create an image for this post. Humans understand requests immediately. I tried Stable Diffusion multiple times and got this:

This great artwork has a pen, robot, and sketch, however it was not asked. Maybe it was because the tokenizer deleted is and a words from a statement, but I tried other requests such robot painting picture with pen without success. It's harder to prompt a model than a person.

I hope Stable Diffusion's general effects are evident. Despite its limitations, it can produce beautiful photographs in some settings. Readers who want to use Stable Diffusion results should be warned. Source code examination demonstrates that Stable Diffusion images feature a concealed watermark (text StableDiffusionV1 and SDV2) encoded using the invisible-watermark Python package. It's not a secret, because the official Stable Diffusion repository's test watermark.py file contains a decoding snippet. The put watermark line in the txt2img.py source code can be removed if desired. I didn't discover this watermark on photographs made by the online Hugging Face demo. Maybe I did something incorrectly (but maybe they are just not using the txt2img script on their backend at all).

Conclusion

The Stable Diffusion model was fascinating. As I mentioned before, trying something yourself is always better than taking someone else's word, so I encourage readers to do the same (including this article as well;).

Is Generative AI a game-changer? My humble experience tells me:

• I think that place has a lot of potential. For designers and artists, generative AI can be a truly useful and innovative tool. Unfortunately, it can also pose a threat to some of them since if users can enter a text field to obtain a picture or a website logo in a matter of clicks, why would they pay more to a different party? Is it possible right now? unquestionably not yet. Images still have a very poor quality and are erroneous in minute details. And after viewing the image of the stunning woman above, models and fashion photographers may also unwind because it is highly unlikely that AI will replace them in the upcoming years.

• Today, generative AI is still in its infancy. Even 768x768 images are considered to be of a high resolution when using neural networks, which are computationally highly expensive. There isn't an AI model that can generate high-resolution photographs natively without upscaling or other methods, at least not as of the time this article was written, but it will happen eventually.

• It is still a challenge to accurately represent knowledge in neural networks (information like how many legs a cat has or the year Napoleon was born). Consequently, AI models struggle to create photorealistic photos, at least where little details are important (on the other side, when I searched Google for modern art paintings, the results are often even worse;).

• When compared to the carefully chosen images from official web pages or YouTube reviews, the average output quality of a Stable Diffusion generation process is actually less attractive because to its high degree of randomness. When using the same technique on their own, consumers will theoretically only view those images as 1% of the results.

Anyway, it's exciting to witness this area's advancement, especially because the project is open source. Google's Imagen and DALL-E 2 can also produce remarkable findings. It will be interesting to see how they progress.

Unleash the power of the Option key ⌥

#1 Open a link in the Private tab first.

Previously, if I needed to open a Safari link in a private window, I would:

• copied the URL with the right click command,

• choose File > New Private Window to open a private window, and

• clicked return after pasting the URL.

I've found a more straightforward way.

Right-clicking a link shows this, right?

Hold option (⌥) for:

Click Open Link in New Private Window while holding.

Finished!

#2. Instead of searching for specific characters, try this

You may use unicode for business or school. Most people Google them when they need them.

That is lengthy!

You can type some special characters just by pressing ⌥ and a key.

For instance

• ⌥+2 -> ™ (Trademark)
• ⌥+0 -> ° (Degree)
• ⌥+G -> © (Copyright)
• ⌥+= -> ≠ (Not equal to)
• ⌥+< -> ≤ (Less than or equal to)
• ⌥+> -> ≥ (Greater then or equal to)
• ⌥+/ -> ÷ (Different symbol for division)

#3 Activate Do Not Disturb silently.

Do Not Disturb when sharing my screen is awkward for me (because people may think Im trying to hide some secret notifications).

Here's another method.

Hold ⌥ and click on Time (at the extreme right on the menu-bar).

Now, DND is activated (secretly!). To turn it off, do it again.

Note: This works only for DND focus.

#4. Resize a window starting from its center

Although this is rarely useful, it is still a hidden trick.

When you resize a window, the opposite edge or corner is used as the pivot, right?

However, if you want to resize it with its center as the pivot, hold while doing so.

#5. Yes, Cut-Paste is available on Macs as well (though it is slightly different).

I call it copy-move rather than cut-paste. This is how it works.

Carry it out.

Choose a file (by clicking on it), then copy it (⌘+C).

Go to a new location on your Mac. Do you use ⌘+V to paste it? However, to move it, press ⌘+⌥+V.

This removes the file from its original location and copies it here. And it works exactly like cut-and-paste on Windows.

#6. Instantly expand all folders

Set your Mac's folders to List view.

Assume you have one folder with multiple subfolders, each of which contains multiple files. And you wanted to look at every single file that was over there.

How would you do?

You're used to clicking the ⌄ glyph near the folder and each subfolder to expand them all, right? Instead, hold down ⌥ while clicking ⌄ on the parent folder.

This is what happens next.

Everything expands.

View/Copy a file's path as an added bonus

If you want to see the path of a file in Finder, select it and hold ⌥, and you'll see it at the bottom for a moment.

To copy its path, right-click on the folder and hold down ⌥ to see this

Click on Copy <"folder name"> as Pathname to do it.

#7 "Save As"

I was irritated by the lack of "Save As" in Pages when I first got a Mac (after 15 years of being a Windows guy).

It was necessary for me to save the file as a new file, in a different location, with a different name, or both.

Unfortunately, I couldn't do it on a Mac.

However, I recently discovered that it appears when you hold ⌥ when in the File menu.

Yay!

Amazon's Rings of Power is the most costly TV series ever made. This is merely a down payment towards Amazon's grand goal.

Here's a video:

Amazon bought J.R.R. Tolkien's fantasy novels for $250 million in 2017. This agreement allows Amazon to create a Tolkien series for Prime Video.

The business spent years developing and constructing a Lord of the Rings prequel. Rings of Power premiered on September 2, 2022.

It drew 25 million global viewers in 24 hours. Prime Video's biggest debut.

An Exorbitant Budget

The most expensive. First season cost $750 million to $1 billion, making it the most costly TV show ever.

Jeff Bezos has spent years looking for the next Game of Thrones, a critically and commercially successful original series. Rings of Power could help.

Why would Amazon bet $1 billion on one series?

It's Not Just About the Streaming War

It's simple to assume Amazon just wants to win. Since 2018, the corporation has been fighting Hulu, Netflix, HBO, Apple, Disney, and NBC. Each wants your money, talent, and attention. Amazon's investment goes beyond rivalry.

Subscriptions Are the Bait

Audible, Amazon Music, and Prime Video are subscription services, although the company's fundamental business is retail. Amazon's online stores contribute over 50% of company revenue. Subscription services contribute 6.8%. The company's master plan depends on these subscriptions.

Streaming videos on Prime increases membership renewals. Free trial participants are more likely to join. Members buy twice as much as non-members.

Amazon Studios doesn't generate original programming to earn from Prime Video subscriptions. It aims to retain and attract clients.

Amazon can track what you watch and buy. Its algorithm recommends items and services. Mckinsey says you'll use more Amazon products, shop at Amazon stores, and watch Amazon entertainment.

In 2015, the firm launched the first season of The Man in the High Castle, a dystopian alternate history TV series depicting a world ruled by Nazi Germany and Japan after World War II.

This $72 million production earned two Emmys. It garnered 1.15 million new Prime users globally.

When asked about his Hollywood investment, Bezos said, "A Golden Globe helps us sell more shoes."

Selling more footwear

Amazon secured a deal with DirecTV to air Thursday Night Football in restaurants and bars. First streaming service to have exclusive NFL games.

This isn't just about Thursday night football, says media analyst Ritchie Greenfield. This sells t-shirts. This may be a ticket. Amazon does more than stream games.

The Rings of Power isn't merely a production showcase, either. This sells Tolkien's fantasy novels such Lord of the Rings, The Hobbit, and The Silmarillion.

This tiny commitment keeps you in Amazon's ecosystem.

I'm trying to understand why Bitcoin NFTs aren't ready.

Ordinals, a new Bitcoin protocol, has been controversial. NFTs can be added to Bitcoin transactions using the protocol. They are not tokens or fungible. Bitcoin NFTs are transaction metadata. Yes. They're not owned.

In January, the Ordinals protocol allowed data like photos to be directly encoded onto sats, the smallest units of Bitcoin worth 0.00000001 BTC, on the Bitcoin blockchain. Ordinals does not need a sidechain or token like other techniques. The Ordinals protocol has encoded JPEG photos, digital art, new profile picture (PFP) projects, and even 1993 DOOM onto the Bitcoin network.

Ordinals inscriptions are permanent digital artifacts preserved on the Bitcoin blockchain. It differs from Ethereum, Solana, and Stacks NFT technologies that allow smart contract creators to change information. Ordinals store the whole image or content on the blockchain, not just a link to an external server, unlike centralized databases, which can change the linked image, description, category, or contract identifier.

So far, more than 50,000 ordinals have been produced on the Bitcoin blockchain, and some of them have already been sold for astronomical amounts. The Ethereum-based CryptoPunks NFT collection spawned Ordinal Punk. Inscription 620 sold for 9.5 BTC, or $218,000, the most.

Segwit and Taproot, two important Bitcoin blockchain updates, enabled this. These protocols store transaction metadata, unlike Ethereum, where the NFT is the token. Bitcoin's NFT is a sat's transaction details.

What effects do ordinary values and NFTs have on the Bitcoin blockchain?

Ordinals will likely have long-term effects on the Bitcoin Ecosystem since they store, transact, and compute more data.

Charges Ordinals introduce scalability challenges. The Bitcoin network has limited transaction throughput and increased fees during peak demand. NFTs could make network transactions harder and more expensive. Ordinals currently occupy over 50% of block space, according to Glassnode.

One of the protocols that supported Ordinals Taproot has also seen a huge uptick:

Taproot use increases block size and transaction costs.

This could cause network congestion but also support more L2s with Ordinals-specific use cases. Dune info here.

Storage Needs The Bitcoin blockchain would need to store more data to store NFT data directly. Since ordinals were introduced, blocksize has tripled from 0.7mb to over 2.2mb, which could increase storage costs and make it harder for nodes to join the network.

Use Case Diversity On the other hand, NFTs on the Bitcoin blockchain could broaden Bitcoin's use cases beyond storage and payment. This could expand Bitcoin's user base. This is two-sided. Bitcoin was designed to be trustless, decentralized, peer-to-peer money.

Chain to permanently store NFTs as ordinals will change everything.

Popularity rise This new use case will boost Bitcoin appeal, according to some. This argument fails since Bitcoin is the most popular cryptocurrency. Popularity doesn't require a new use case. Cryptocurrency adoption boosts Bitcoin. It need not compete with Ethereum or provide extra benefits to crypto investors. If there was a need for another chain that supports NFTs (there isn't), why would anyone choose the slowest and most expensive network? It appears contradictory and unproductive.

Nonetheless, holding an NFT on the Bitcoin blockchain is more secure than any other blockchain, but this has little utility.

Bitcoin NFTs are undoubtedly controversial. NFTs are strange and perhaps harmful to Bitcoin's mission. If Bitcoin NFTs are here to stay, I hope a sidechain or rollup solution will take over and leave the base chain alone.