How to reframe irritation and go forward

“… before you can figure out where you are going, you need to know where you are, and once you know and accept where you are, you can design your way to where you want to be.” — Bill Burnett and Dave Evans

“You’ve been here before. But there are some new ingredients this time. What can tell yourself that will make you understand that now isn’t just like last year? That there’s something new in this August.”

My coach paused. I sighed, inhaled deeply, and considered her question.

What could I say? I simply needed a plan from her so everything would fall into place and I could be the happy, successful person I want to be.

Time passed. My mind was exhausted from running all morning, all summer, or the last five years, searching for what to do next and how to get there.

Calmer, I remembered that my coach's inquiry had benefited me throughout the summer. The month before our call, I read Designing Your Work Life — How to Thrive and Change and Find Happiness at Work from Standford University’s Bill Burnett and Dave Evans.

A passage in their book felt like a lifeline: “We have something important to say to you: Wherever you are in your work life, whatever job you are doing, it’s good enough. For now. Not forever. For now.”

As I remembered this book on the coaching call, I wondered if I could embrace where I am in August and say my job life is good enough for now. Only temporarily.

I've done that since. I'm getting unstuck.

Here's how you can take the first step in any area where you feel stuck.

How to acquire the perspective of "Good enough for now" for yourself

We’ve all heard the advice to just make the best of a bad situation. That´s not bad advice, but if you only make the best of a bad situation, you are still in a bad situation. It doesn’t get to the root of the problem or offer an opportunity to change the situation. You’re more cheerfully navigating lousiness, which is an improvement, but not much of one and rather hard to sustain over time.” — Bill Burnett and Dave Evans

Reframing Burnett at Evans says good enough for now is the key to being happier at work. Because, as they write, a designer always has options.

Choosing to believe things are good enough for now is liberating. It helps us feel less victimized and less judged. Accepting our situation helps us become unstuck.

Let's break down the process, which designers call constructing your way ahead, into steps you can take today.

Writing helps get started. First, write down your challenge and why it's essential to you. If pen and paper help, try this strategy:

• Make the decision to accept the circumstance as it is. Designers always begin by acknowledging the truth of the situation. You now refrain from passing judgment. Instead, you simply describe the situation as accurately as you can. This frees us from negative thought patterns that prevent us from seeing the big picture and instead keep us in a tunnel of negativity.

• Look for a reframing right now. Begin with good enough for the moment. Take note of how your body feels as a result. Tell yourself repeatedly that whatever is occurring is sufficient for the time being. Not always, but just now. If you want to, you can even put it in writing and repeatedly breathe it in, almost like a mantra.

• You can select a reframe that is more relevant to your situation once you've decided that you're good enough for now and have allowed yourself to believe it. Try to find another perspective that is possible, for instance, if you feel unappreciated at work and your perspective of I need to use and be recognized for all my new skills in my job is making you sad and making you want to resign. For instance, I can learn from others at work and occasionally put my new abilities to use.

• After that, leave your mind and act in accordance with your new perspective. Utilize the designer's bias for action to test something out and create a prototype that you can learn from. Your beginning point for creating experiences that will support the new viewpoint derived from the aforementioned point is the new perspective itself. By doing this, you recognize a circumstance at work where you can provide value to yourself or your workplace and then take appropriate action. Send two or three coworkers from whom you wish to learn anything an email, for instance, asking them to get together for coffee or a talk.

Choose tiny, doable actions. You prioritize them at work.

Let's assume you're feeling disconnected at work, so you make a list of folks you may visit each morning or invite to lunch. If you're feeling unmotivated and tired, take a daily walk and treat yourself to a decent coffee.

This may be plenty for now. If you want to take this procedure further, use Burnett and Evans' internet tools and frameworks.

Developing the daily practice of reframing

“We’re not discontented kids in the backseat of the family minivan, but how many of us live our lives, especially our work lives, as if we are?” — Bill Burnett and Dave Evans

I choose the good enough for me perspective every day, often. No quick fix. Am a failing? Maybe a little bit, but I like to think of it more as building muscle.

This way, every time I tell myself it's ok, I hear you. For now, that muscle gets stronger.

Hopefully, reframing will become so natural for us that it will become a habit, and not a technique anymore.

If you feel like you’re stuck in your career or at work, the reframe of Good enough, for now, might be valuable, so just go ahead and try it out right now.

And while you’re playing with this, why not think of other areas of your life too, like your relationships, where you live — even your writing, and see if you can feel a shift?

A must-read for everyone

Jobs is well-known. You probably know the tall, thin guy who wore the same clothing every day. His influence is unavoidable. In fewer than 40 years, Jobs' innovations have impacted computers, movies, cellphones, music, and communication.

Steve Jobs may be more imaginative than the typical person, but if we can use some of his ingenuity, ambition, and good traits, we'll be successful. This essay explains how to follow his guidance and success secrets.

1. Repetition is necessary for success.

Be patient and diligent to master something. Practice makes perfect. This is why older workers are often more skilled.

When should you repeat a task? When you're confident and excited to share your product. It's when to stop tweaking and repeating.

Jobs stated he'd make the crowd sh** their pants with an iChat demo.

Use this in your daily life.

• Start with the end in mind. You can put it in writing and be as detailed as you like with your plan's schedule and metrics. For instance, you have a goal of selling three coffee makers in a week.

• Break it down, break the goal down into particular tasks you must complete, and then repeat those tasks. To sell your coffee maker, you might need to make 50 phone calls.

• Be mindful of the amount of work necessary to produce the desired results. Continue doing this until you are happy with your product.

2. Acquire the ability to add and subtract.

How did Picasso invent cubism? Pablo Picasso was influenced by stylised, non-naturalistic African masks that depict a human figure.

Artists create. Constantly seeking inspiration. They think creatively about random objects. Jobs said creativity is linking things. Creative people feel terrible when asked how they achieved something unique because they didn't do it all. They saw innovation. They had mastered connecting and synthesizing experiences.

Use this in your daily life.

• On your phone, there is a note-taking app. Ideas for what you desire to learn should be written down. It may be learning a new language, calligraphy, or anything else that inspires or intrigues you.

• Note any ideas you have, quotations, or any information that strikes you as important.

• Spend time with smart individuals, that is the most important thing. Jim Rohn, a well-known motivational speaker, has observed that we are the average of the five people with whom we spend the most time.

• Learning alone won't get you very far. You need to put what you've learnt into practice. If you don't use your knowledge and skills, they are useless.

3. Develop the ability to refuse.

Steve Jobs deleted thousands of items when he created Apple's design ethic. Saying no to distractions meant upsetting customers and partners.

John Sculley, the former CEO of Apple, said something like this. According to Sculley, Steve’s methodology differs from others as he always believed that the most critical decisions are things you choose not to do.

Use this in your daily life.

• Never be afraid to say "no," "I won't," or "I don't want to." Keep it simple. This method works well in some situations.

• Give a different option. For instance, X might be interested even if I won't be able to achieve it.

• Control your top priority. Before saying yes to anything, make sure your work schedule and priority list are up to date.

4. Follow your passion

“Follow your passion” is the worst advice people can give you. Steve Jobs didn't start Apple because he suddenly loved computers. He wanted to help others attain their maximum potential.

Great things take a lot of work, so quitting makes sense if you're not passionate. Jobs learned from history that successful people were passionate about their work and persisted through challenges.

Use this in your daily life.

• Stay away from your passion. Allow it to develop daily. Keep working at your 9-5-hour job while carefully gauging your level of desire and endurance. Less risk exists.

• The truth is that if you decide to work on a project by yourself rather than in a group, it will take you years to complete it instead of a week. Instead, network with others who have interests in common.

• Prepare a fallback strategy in case things go wrong.

Success, this small two-syllable word eventually gives your life meaning, a perspective. What is success?  For most, it's achieving their ambitions. However, there's a catch. Successful people aren't always happy.

Furthermore, where do people’s goals and achievements end? It’s a never-ending process. Success is a journey, not a destination. We wish you not to lose your way on this journey.

Don’t give a number

Google pays its software engineers generously. While many of their employees are competent, they disregard a critical skill to maximize their pay.

Negotiation.

If Google employees have never negotiated, they're as helpless as anyone else.

In this piece, I'll reveal a compensation negotiation tip that will set you apart.

The Fallacy of Negotiating

How do you negotiate your salary? “Just give them a number twice the amount you really want”. - Someplace on the internet

Above is typical negotiation advice. If you ask for more than you want, the recruiter may meet you halfway.

It seems logical and great, but here's why you shouldn't follow that advice.

Haitian hostage rescue

In 1977, an official's aunt was kidnapped in Haiti. The kidnappers demanded $150,000 for the aunt's life. It seems reasonable until you realize why kidnappers want $150,000.

FBI detective and negotiator Chris Voss researched why they demanded so much.

“So they could party through the weekend”

When he realized their ransom was for partying, he offered $4,751 and a CD stereo. Criminals freed the aunt.

These thieves gave 31.57x their estimated amount and got a fraction. You shouldn't trust these thieves to negotiate your compensation.

What happened?

Negotiating your offer and Haiti

This narrative teaches you how to negotiate with a large number.

You can and will be talked down.

If a recruiter asks your wage expectation and you offer double, be ready to explain why.

If you can't justify your request, you may be offered less. The recruiter will notice and talk you down.

Reasonably,

• a tiny bit more than the present amount you earn

• a small premium over an alternative offer

• a little less than the role's allotted amount

Real-World Illustration

Recruiter: What’s your expected salary? Candidate: (I know the role is usually $100,000) $200,000 Recruiter: How much are you compensated in your current role? Candidate: $90,000 Recruiter: We’d be excited to offer you $95,000 for your experiences for the role.

So Why Do They Even Ask?

Recruiters ask for a number to negotiate a lower one. Asking yourself limits you.

You'll rarely get more than you asked for, and your request can be lowered.

The takeaway from all of this is to never give an expected compensation.

Tell them you haven't thought about it when you applied.

We've fought for humanity's sake. We need equilibrium.

We live in a world of opposites (black/white, up/down, love/hate), thus life is a game of achieving equilibrium. We have a universe of paradoxes within ourselves, not just in physics.

Individually, you balance your intellect and heart, but as a species, we're full of polarities. They might be gentle and compassionate, then ruthless and unsympathetic.

We desire for connection so much that we personify non-human beings and objects while turning to violence and hatred toward others. These contrasts baffle me. Will we find balance?

Anthropomorphization

Assigning human-like features or bonding with objects is common throughout childhood. Cartoons often give non-humans human traits. Adults still anthropomorphize this trait. Researchers agree we start doing it as infants and continue throughout life.

Humans of all ages are good at humanizing stuff. We build emotional attachments to weather events, inanimate objects, animals, plants, and locales. Gods, goddesses, and fictitious figures are anthropomorphized.

Cast Away, starring Tom Hanks, features anthropization. Hanks is left on an island, where he builds an emotional bond with a volleyball he calls Wilson.

We became emotionally invested in Wilson, including myself.

Why do we do it, though?

Our instincts and traits helped us survive and thrive. Our brain is alert to other people's thoughts, feelings, and intentions to assist us to determine who is safe or hazardous. We can think about others and our own mental states, or about thinking. This is the Theory of Mind.

Neurologically, specialists believe the Theory of Mind has to do with our mirror neurons, which exhibit the same activity while executing or witnessing an action.

Mirror neurons may contribute to anthropization, but they're not the only ones. In 2021, Harvard Medical School researchers at MGH and MIT colleagues published a study on the brain's notion of mind.

“Our study provides evidence to support theory of mind by individual neurons. Until now, it wasn’t clear whether or how neurons were able to perform these social cognitive computations.”

Neurons have particular functions, researchers found. Others encode information that differentiates one person's beliefs from another's. Some neurons reflect tale pieces, whereas others aren't directly involved in social reasoning but may multitask contributing factors.

Combining neuronal data gives a precise portrait of another's beliefs and comprehension. The theory of mind describes how we judge and understand each other in our species, and it likely led to anthropomorphism. Neuroscience indicates identical brain regions react to human or non-human behavior, like mirror neurons.

Some academics believe we're wired for connection, which explains why we anthropomorphize. When we're alone, we may anthropomorphize non-humans.

Humanizing non-human entities may make them deserving of moral care, according to another theory. Animamorphizing something makes it responsible for its actions and deserves punishments or rewards. This mental shift is typically apparent in our connections with pets and leads to deanthropomorphization.

Dehumanization

Dehumanizing involves denying someone or anything ethical regard, the opposite of anthropomorphizing.

Dehumanization occurs throughout history. We do it to everything in nature, including ourselves. We experiment on and torture animals. We enslave, hate, and harm other groups of people.

Race, immigrant status, dress choices, sexual orientation, social class, religion, gender, politics, need I go on? Our degrading behavior is promoting fascism and division everywhere.

Dehumanizing someone or anything reduces their agency and value. Many assume they're immune to this feature, but tests disagree.

It's inevitable. Humans are wired to have knee-jerk reactions to differences. We are programmed to dehumanize others, and it's easier than we'd like to admit.

Why do we do it, though?

Dehumanizing others is simpler than humanizing things for several reasons. First, we consider everything unusual as harmful, which has helped our species survive for hundreds of millions of years. Our propensity to be distrustful of others, like our fear of the unknown, promotes an us-vs.-them mentality.

Since WWII, various studies have been done to explain how or why the holocaust happened. How did so many individuals become radicalized to commit such awful actions and feel morally justified? Researchers quickly showed how easily the mind can turn gloomy.

Stanley Milgram's 1960s electroshock experiment highlighted how quickly people bow to authority to injure others. Philip Zimbardo's 1971 Stanford Prison Experiment revealed how power may be abused.

The us-versus-them attitude is natural and even young toddlers act on it. Without a relationship, empathy is more difficult.

It's terrifying how quickly dehumanizing behavior becomes commonplace. The current pandemic is an example. Most countries no longer count deaths. Long Covid is a major issue, with predictions of a handicapped tsunami in the future years. Mostly, we shrug.

In 2020, we panicked. Remember everyone's caution? Now Long Covid is ruining more lives, threatening to disable an insane amount of our population for months or their entire lives.

There's little research. Experts can't even classify or cure it. The people should be outraged, but most have ceased caring. They're over covid.

We're encouraged to find a method to live with a terrible pandemic that will cause years of damage. People aren't worried about infection anymore. They shrug and say, "We'll all get it eventually," then hope they're not one of the 30% who develops Long Covid.

We can correct course before further damage. Because we can recognize our urges and biases, we're not captives to them. We can think critically about our thoughts and behaviors, then attempt to improve. We can recognize our deficiencies and work to attain balance.

Changing perspectives

We're currently attempting to find equilibrium between opposites. It's superficial to defend extremes by stating we're only human or wired this way because both imply we have no control.

Being human involves having self-awareness, and by being careful of our thoughts and acts, we can find balance and recognize opposites' purpose.

Extreme anthropomorphizing and dehumanizing isolate and imperil us. We anthropomorphize because we desire connection and dehumanize because we're terrified, frequently of the connection we crave. Will we find balance?

Katrina Paulson ponders humanity, unanswered questions, and discoveries. Please check out her newsletters, Curious Adventure and Curious Life.

New standards help NFTs reach their full potential.

NFTs lack interoperability and functionality. They have great potential but are mostly speculative. To maximize NFTs, we need flexible smart contracts.

Current requirements are too restrictive.

Most NFTs are based on ERC-721, which makes exchanging them easy. CryptoKitties, a popular online game, used the 2017 standard to demonstrate NFTs' potential.

This simple standard includes a base URI and incremental IDs for tokens. Add the tokenID to the base URI to get the token's metadata.

This let creators collect NFTs. Many NFT projects store metadata on IPFS, a distributed storage network, but others use Google Drive. NFT buyers often don't realize that if the creators delete or move the files, their NFT is just a pointer.

This isn't the standard's biggest issue. There's no way to validate NFT projects.

Creators are one of the most important aspects of art, but nothing is stored on-chain.

ERC-721 contracts only have a name and symbol.

Most of the data on OpenSea's collection pages isn't from the NFT's smart contract. It was added through a platform input field, so it's in the marketplace's database. Other websites may have different NFT information.

In five years, your NFT will be just a name, symbol, and ID.

Your NFT doesn't mention its creators. Although the smart contract has a public key, it doesn't reveal who created it.

The NFT's creators and their reputation are crucial to its value. Think digital fashion and big brands working with well-known designers when more professionals use NFTs. Don't you want them in your NFT?

Would paintings be as valuable if their artists were unknown? Would you believe it's real?

Buying directly from an on-chain artist would reduce scams. Current standards don't allow this data.

Most creator profiles live on centralized marketplaces and could disappear. Current platforms have outpaced underlying standards. The industry's standards are lagging.

For NFTs to grow beyond pointers to a monkey picture file, we may need to use new Web3-based standards.

Introducing NFTs 2.0

Fabian Vogelsteller, creator of ERC-20, developed new web3 standards. He proposed LSP7 Digital Asset and LSP8 Identifiable Digital Asset, also called NFT 2.0.

NFT and token metadata inputs are extendable. Changes to on-chain metadata inputs allow NFTs to evolve. Instead of public keys, the contract can have Universal Profile addresses attached. These profiles show creators' faces and reputations. NFTs can notify asset receivers, automating smart contracts.

LSP7 and LSP8 use ERC725Y. Using a generic data key-value store gives contracts much-needed features:

• The asset can be customized and made to stand out more by allowing for unlimited data attachment.

• Recognizing changes to the metadata

• using a hash reference for metadata rather than a URL reference

This base will allow more metadata customization and upgradeability. These guidelines are:

• Genuine and Verifiable Now, the creation of an NFT by a specific Universal Profile can be confirmed by smart contracts.

• Dynamic NFTs can update Flexible & Updatable Metadata, allowing certain things to evolve over time.

• Protected metadata Now, secure metadata that is readable by smart contracts can be added indefinitely.

• Better NFTS prevent the locking of NFTs by only being sent to Universal Profiles or a smart contract that can interact with them.

Summary

NFTS standards lack standardization and powering features, limiting the industry.

ERC-721 is the most popular NFT standard, but it only represents incremental tokenIDs without metadata or asset representation. No standard sender-receiver interaction or security measures ensure safe asset transfers.

NFT 2.0 refers to the new LSP7-DigitalAsset and LSP8-IdentifiableDigitalAsset standards.

They have new standards for flexible metadata, secure transfers, asset representation, and interactive transfer.

With NFTs 2.0 and Universal Profiles, creators could build on-chain reputations.

NFTs 2.0 could bring the industry's needed innovation if it wants to move beyond trading profile pictures for speculation.

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.