Are you one of these Zoom dogs?

The Person Who Is Apparently Always on Mute

Waffles thinks he can overpower the mute button by shouting loudly.

The person who believed their camera to be off

Barkley's used to remote work, but he hasn't mastered the "Stop Video" button. Everyone is affected.

Who is driving for some reason, exactly?

Why is Pumpkin always late? Who knows? Shouldn't she be driving? If you could hear her over the freeway, she'd answer these questions.

The Person With the Amazing Bookcase

Cicero likes to use SAT-words like "leverage" and "robust" in Zoom sessions, presumably from all the books he wants you to see behind him.

The Individual Who Is Unnecessarily Dressed

We hope Bandit is going somewhere beautiful after this meeting, or else he neglected the quarterly earnings report and is overcompensating to distract us.

The person who works through lunch in between zoom calls

Barksworth has back-to-back meetings all day, so you can watch her eat while she talks.

The Person Who Is A Little Too Comfy

Hercules thinks Zoom meetings happen between sleeps. He'd appreciate everyone speaking more quietly.

The Person Who Answered the Phone Outside

Frisbee has a gorgeous backyard and lives in a place with great weather year-round, and she wants you to think about that during the daily team huddle.

Who Wants You to Pay Attention to Their Pet

Snickers hasn't listened to you in 20 minutes unless you tell her how cute her kitten is.

One who is, for some reason, positioned incorrectly on the screen

Nelson's meetings consist primarily of attempting to figure out how he positioned his laptop so absurdly.

The person who says too many goodbyes

Zeus waves farewell like it's your first day of school while everyone else searches for the "Leave Meeting" button. It's nice.

He who has a poor internet connection

Ziggy's connectivity problems continue... She gives a long speech as everyone waits awkwardly to inform her they missed it.

The Clearly Multitasking Person

Tinkerbell can play fetch during the monthly staff meeting if she works from home, but that's not a good idea.

The Person Using Zoom as a Makeup and Hair Mirror

If Gail and Bob knew Zoom had a "hide self view" option, they'd be distraught.

The person who feels at ease with simply leaving

Rusty bails when a Zoom conference is over. Rusty's concept is decent.

My colleague Ilana Ovental and I examined pandemic media coverage of education at the end of last year. That analysis examined coverage changes. We tracked K-12 topic attention over the previous two decades using Lexis Nexis. See the results here.

I was struck by how cleanly the past two decades can be divided up into three (or three and a half) eras of school reform—a framing that can help us comprehend where we are and how we got here. In a time when epidemic, political unrest, frenetic news cycles, and culture war can make six months seem like a lifetime, it's worth pausing for context.

If you look at the peaks in the above graph, the 21st century looks to be divided into periods. The decade-long rise and fall of No Child Left Behind began during the Bush administration. In a few years, NCLB became the dominant K-12 framework. Advocates and financiers discussed achievement gaps and measured success with AYP.

NCLB collapsed under the weight of rigorous testing, high-stakes accountability, and a race to the bottom by the Obama years. Obama's Race to the Top garnered attention, but its most controversial component, the Common Core State Standards, rose quickly.

Academic standards replaced assessment and accountability. New math, fiction, and standards were hotly debated. Reformers and funders chanted worldwide benchmarking and systems interoperability.

We went from federally driven testing and accountability to government encouraged/subsidized/mandated (pick your verb) reading and math standardization. Last year, Checker Finn and I wrote The End of School Reform? The 2010s populist wave thwarted these objectives. The Tea Party, Occupy Wall Street, Black Lives Matter, and Trump/MAGA all attacked established institutions.

Consequently, once the Common Core fell, no alternative program emerged. Instead, school choice—the policy most aligned with populist suspicion of institutional power—reached a half-peak. This was less a case of choice erupting to prominence than of continuous growth in a vacuum. Even with Betsy DeVos' determined, controversial efforts, school choice received only half the media attention that NCLB and Common Core did at their heights.

Recently, culture clash-fueled attention to race-based curriculum and pedagogy has exploded (all playing out under the banner of critical race theory). This third, culture war-driven wave may not last as long as the other waves.

Even though I don't understand it, the move from slow-building policy debate to fast cultural confrontation over two decades is notable. I don't know if it's cyclical or permanent, or if it's about schooling, media, public discourse, or all three.

One final thought: After doing this work for decades, I've noticed how smoothly advocacy groups, associations, and other activists adapt to the zeitgeist. In 2007, mission statements focused on accomplishment disparities. Five years later, they promoted standardization. Language has changed again.

Part of this is unavoidable and healthy. Chasing currents can also make companies look unprincipled, promote scepticism, and keep them spinning the wheel. Bearing in mind that these tides ebb and flow may give educators, leaders, and activists more confidence to hold onto their values and pause when they feel compelled to follow the crowd.

ZERO GRACE/ZERO MALICE

Big companies and wealthy people make stupid mistakes too.

Your ancestors kept snakes and drank bad water. You (probably) don't because you've learnt from their failures via instinct+, the ultimate life-lessons streaming network in your head. Instincts foretell the future. If you approach a lion, it'll eat you. Our society's nuanced/complex decisions have surpassed instinct. Human growth depends on how we handle these issues. 80% of people believe they are above-average drivers, yet few believe they make many incorrect mistakes that make them risky. Stupidity hurts others like death. Basic Laws of Human Stupidity by Carlo Cipollas:

1. Everyone underestimates the prevalence of idiots in our society.

2. Any other trait a person may have has no bearing on how likely they are to be stupid.

3. A dumb individual is one who harms someone without benefiting themselves and may even lose money in the process.

4. Non-dumb people frequently underestimate how destructively powerful stupid people can be.

5. The most dangerous kind of person is a moron.

Professor Cippola defines stupid as bad for you and others. We underestimate the corporate world's and seemingly successful people's ability to make bad judgments that harm themselves and others. Success is an intoxication that makes you risk-aggressive and blurs your peripheral vision.

Stupid companies and decisions:

Big Dumber

Big-company bad ideas have more bulk and inertia. The world's most valuable company recently showed its board a VR headset. Jony Ive couldn't destroy Apple's terrible idea in 2015. Mr. Ive said that VR cut users off from the outer world, made them seem outdated, and lacked practical uses. Ives' design team doubted users would wear headsets for lengthy periods.

VR has cost tens of billions of dollars over a decade to prove nobody wants it. The next great SaaS startup will likely come from Florence, not Redmond or San Jose.

Apple Watch and Airpods have made the Cupertino company the world's largest jewelry maker. 10.5% of Apple's income, or $38 billion, comes from wearables in 2021. (seven times the revenue of Tiffany & Co.). Jewelry makes you more appealing and useful. Airpods and Apple Watch do both.

Headsets make you less beautiful and useful and promote isolation, loneliness, and unhappiness among American teenagers. My sons pretend they can't hear or see me when on their phones. VR headsets lack charisma.

Coinbase disclosed a plan to generate division and tension within its workplace weeks after Apple was pitched $2,000 smokes. The crypto-trading platform is piloting a program that rates staff after every interaction. If a coworker says anything you don't like, you should tell them how to improve. Everyone gets a 110-point scorecard. Coworkers should evaluate a person's rating while deciding whether to listen to them. It's ridiculous.

Organizations leverage our superpower of cooperation. This encourages non-cooperation, period. Bridgewater's founder Ray Dalio designed the approach to promote extreme transparency. Dalio has 223 billion reasons his managerial style works. There's reason to suppose only a small group of people, largely traders, will endure a granular scorecard. Bridgewater has 20% first-year turnover. Employees cry in bathrooms, and sex scandals are settled by ignoring individuals with poor believability levels. Coinbase might take solace that the stock is 80% below its initial offering price.

Poor Stupid

Fools' ledgers are valuable. More valuable are lists of foolish rich individuals.

Robinhood built a $8 billion corporation on financial ignorance. The firm's median account value is $240, and its stock has dropped 75% since last summer. Investors, customers, and society lose. Stupid. Luna published a comparable list on the blockchain, grew to $41 billion in market cap, then plummeted.

A podcast presenter is recruiting dentists and small-business owners to invest in Elon Musk's Twitter takeover. Investors pay a 7% fee and 10% of the upside for the chance to buy Twitter at a 35% premium to the current price. The proposal legitimizes CNBC's Trade Like Chuck advertising (Chuck made $4,600 into $460,000 in two years). This is stupid because it adds to the Twitter deal's desperation. Mr. Musk made an impression when he urged his lawyers to develop a legal rip-cord (There are bots on the platform!) to abandon the share purchase arrangement (for less than they are being marketed by the podcaster). Rolls-Royce may pay for this list of the dumb affluent because it includes potential Cullinan buyers.

Worst company? Flowcarbon, founded by WeWork founder Adam Neumann, operates at the convergence of carbon and crypto to democratize access to offsets and safeguard the earth's natural carbon sinks. Can I get an ayahuasca Big Gulp?

Neumann raised $70 million with their yogababble drink. More than half of the consideration came from selling GNT. Goddess Nature Token. I hope the company gets an S-1. Or I'll start a decentralized AI Meta Renewable NFTs company. My Community Based Ebitda coin will fund the company. Possible.

Stupidity inside oneself

This weekend, I was in NYC with my boys. My 14-year-old disappeared. He's realized I'm not cool and is mad I let the charade continue. When out with his dad, he likes to stroll home alone and depart before me. Friends told me hell would return, but I was surprised by how fast the eye roll came.

Not so with my 11-year-old. We went to The Edge, a Hudson Yards observation platform where you can see the city from 100 storeys up for $38. This is hell's seventh ring. Leaning into your boys' interests is key to engaging them (dad tip). Neither loves Crossfit, WW2 history, or antitrust law.

We take selfies on the Thrilling Glass Floor he spots. Dad, there's a bar! Coke? I nod, he rushes to the bar, stops, runs back for money, and sprints back. Sitting on stone seats, drinking Atlanta Champagne, he turns at me and asks, Isn't this amazing? I'll never reach paradise.

Later that night, the lads are asleep and I've had two Zacapas and Cokes. I SMS some friends about my day and how I feel about sons/fatherhood/etc. How I did. They responded and approached. The next morning, I'm sober, have distance from my son, and feel ashamed by my texts. Less likely to impulsively share my emotions with others. Stupid again.

Mainstream economics haven't noticed.

What come to mind when I say the word "economics"?

Probably GDP, unemployment, and inflation.

If you've ever watched the news or listened to an economist, they'll use data like these to defend a political goal.

The issue is that these statistics are total bunk.

I'm being provocative, but I mean it:

• The economy is not measured by GDP.

• How many people are unemployed is not counted in the unemployment rate.

• Inflation is not measured by the CPI.

All orthodox economists' major economic statistics are either wrong or falsified.

Government institutions create all these stats. The administration wants to reassure citizens the economy is doing well.

GDP does not reflect economic expansion.

GDP measures a country's economic size and growth. It’s calculated by the BEA, a government agency.

The US has the world's largest (self-reported) GDP, growing 2-3% annually.

If GDP rises, the economy is healthy, say economists.

Why is the GDP flawed?

GDP measures a country's yearly spending.

The government may adjust this to make the economy look good.

GDP = C + G + I + NX

C = Consumer Spending

G = Government Spending

I = Investments (Equipment, inventories, housing, etc.)

NX = Exports minus Imports

GDP is a country's annual spending.

The government can print money to boost GDP. The government has a motive to increase and manage GDP.

Because government expenditure is part of GDP, printing money and spending it on anything will raise GDP.

They've done this. Since 1950, US government spending has grown 8% annually, faster than GDP.

In 2022, government spending accounted for 44% of GDP. It's the highest since WWII. In 1790-1910, it was 3% of GDP.

Who cares?

The economy isn't only spending. Focus on citizens' purchasing power or quality of life.

Since GDP just measures spending, the government can print money to boost GDP.

Even if Americans are poorer than last year, economists can say GDP is up and everything is fine.

How many people are unemployed is not counted in the unemployment rate.

The unemployment rate measures a country's labor market. If unemployment is high, people aren't doing well economically.

The BLS estimates the (self-reported) unemployment rate as 3-4%.

Why is the unemployment rate so high?

The US government surveys 100k persons to measure unemployment. They extrapolate this data for the country.

They come into 3 categories:

• Employed

People with jobs are employed … duh.

• Unemployed

People who are “jobless, looking for a job, and available for work” are unemployed

• Not in the labor force

The “labor force” is the employed + the unemployed.

The unemployment rate is the percentage of unemployed workers.

Problem is unemployed definition. You must actively seek work to be considered unemployed.

You're no longer unemployed if you haven't interviewed in 4 weeks.

This shit makes no goddamn sense.

Why does this matter?

You can't interview if there are no positions available. You're no longer unemployed after 4 weeks.

In 1994, the BLS redefined "unemployed" to exclude discouraged workers.

If you haven't interviewed in 4 weeks, you're no longer counted in the unemployment rate.

If unemployment were measured by total unemployed, it would be 25%.

Because the government wants to keep the unemployment rate low, they modify the definition.

If every US resident was unemployed and had no job interviews, economists would declare 0% unemployment. Excellent!

Inflation is not measured by the CPI.

The BLS measures CPI. This month was the highest since 1981.

CPI measures the cost of a basket of products across time. Food, energy, shelter, and clothes are included.

A 9.1% CPI means the basket of items is 9.1% more expensive.

What is the CPI problem?

Here's a more detailed explanation of CPI's flaws.

In summary, CPI is manipulated to be understated.

Housing costs are understated to manipulate CPI. Housing accounts for 33% of the CPI because it's the biggest expense for most people.

This signifies it's the biggest CPI weight.

Rather than using actual house prices, the Bureau of Labor Statistics essentially makes shit up. You can read more about the process here.

Surprise! It’s bullshit

The BLS stated Shelter's price rose 5.5% this month.

House prices are up 11-21%. (Source 1, Source 2, Source 3)

Rents are up 14-26%. (Source 1, Source 2)

Why is this important?

If CPI included housing prices, it would be 12-15 percent this month, not 9.1 percent.

9% inflation is nuts. Your money's value halves every 7 years at 9% inflation.

Worse is 15% inflation. Your money halves every 4 years at 15% inflation.

If everyone realized they needed to double their wage every 4-5 years to stay wealthy, there would be riots.

Inflation drains our money's value so the government can keep printing it.

The Solution

Most individuals know the existing system doesn't work, but can't explain why.

People work hard yet lag behind. The government lies about the economy's data.

In reality:

• GDP has been down since 2008

• 25% of Americans are unemployed

• Inflation is actually 15%

People might join together to vote out kleptocratic politicians if they knew the reality.

Having reliable economic data is the first step.

People can't understand the situation without sufficient information. Instead of immigrants or billionaires, people would blame liar politicians.

Here’s the vision:

A decentralized, transparent, and global dashboard that tracks economic data like GDP, unemployment, and inflation for every country on Earth.

Government incentives influence economic statistics.

ShadowStats has already started this effort, but the calculations must be transparent, decentralized, and global to be effective.

If interested, email me at trevorstark02@gmail.com.

Here are some links to further your research:

1. MIT Billion Prices Project

2. 1729 Decentralized Inflation Dashboard Project

3. Balaji Srinivasan on “Fiat Information VS. Crypto Information”

It’s simply not conceivable, or at least not credible.

One of the most touted selling points of Crypto has always been this grandiose idea of decentralization. Bitcoin first arose in 2009 after the housing crisis and subsequent crash that came with it. It aimed to solve this supposed issue of centralization. Nobody “owns” Bitcoin in theory, so the idea then goes that it won’t be subject to the same downfalls that led to the 2008 crash or similarly speculative events that led to the 2008 disaster. The issue is the banks, not the human nature associated with the greedy individuals running them.

Subsequent blockchains have attempted to fix many of the issues of Bitcoin by increasing capacity, decreasing the costs and processing times associated with Bitcoin, and expanding what can be done with their blockchains. Since nobody owns Bitcoin, it hasn’t really been able to be expanded on. You have people like Vitalk Buterin, however, that actively work on Ethereum though.

The leap from Bitcoin to Ethereum was a massive leap toward centralization, and the trend has only gotten worse. In fact, crypto has since become almost exclusively centralized in recent years.

Decentralization is only good in theory

It’s a good idea. In fact, it’s a wonderful idea. However, like other utopian societies, individuals misjudge human nature and greed. In a perfect world, decentralization would certainly be a wonderful idea because sure, people may function as their own banks, move payments immediately, remain anonymous, and so on. However, underneath this are a couple issues:

• You can already send money instantaneously today.

• They are not decentralized.

• Decentralization is a bad idea.

• Being your own bank is a stupid move.

Let’s break these down. Some are quite simple, but lets have a look.

Sending money right away

One thing with crypto is the idea that you can send payments instantly. This has pretty much been entirely solved in current times. You can transmit significant sums of money instantly for a nominal cost and it’s instantaneously cleared. Venmo was launched in 2009 and has since increased to prominence, and currently is on most people's phones. I can directly send ANY amount of money quickly from my bank to another person's Venmo account.

Comparing that with ETH and Bitcoin, Venmo wins all around. I can send money to someone for free instantly in dollars and the only fee paid is optional depending on when you want it.

Both Bitcoin and Ethereum are subject to demand. If the blockchains have a lot of people trying to process transactions fee’s go up, and the time that it takes to receive your crypto takes longer. When Ethereum gets bad, people have reported spending several thousand of dollars on just 1 transaction.

These transactions take place via “miners” bundling and confirming transactions, then recording them on the blockchain to confirm that the transaction did indeed happen. They charge fees to do this and are also paid in Bitcoin/ETH. When a transaction is confirmed, it's then sent to the other users wallet. This within itself is subject to lots of controversy because each transaction needs to be confirmed 6 times, this takes massive amounts of power, and most of the power is wasted because this is an adversarial system in which the person that mines the transaction gets paid, and everyone else is out of luck. Also, these could theoretically be subject to a “51% attack” in which anyone with over 51% of the mining hash rate could effectively control all of the transactions, and reverse transactions while keeping the BTC resulting in “double spending”.

There are tons of other issues with this, but essentially it means: They rely on these third parties to confirm the transactions. Without people confirming these transactions, Bitcoin stalls completely, and if anyone becomes too dominant they can effectively control bitcoin.

Not to mention, these transactions are in Bitcoin and ETH, not dollars. So, you need to convert them to dollars still, and that's several more transactions, and likely to take several days anyway as the centralized exchange needs to send you the money by traditional methods.

They are not distributed

That takes me to the following point. This isn’t decentralized, at all. Bitcoin is the closest it gets because Satoshi basically closed it to new upgrades, although its still subject to:

• Whales

• Miners

It’s vital to realize that these are often the same folks. While whales aren’t centralized entities typically, they can considerably effect the price and outcome of Bitcoin. If the largest wallets holding as much as 1 million BTC were to sell, it’d effectively collapse the price perhaps beyond repair. However, Bitcoin can and is pretty much controlled by the miners. Further, Bitcoin is more like an oligarchy than decentralized. It’s been effectively used to make the rich richer, and both the mining and price is impacted by the rich. The overwhelming minority of those actually using it are retail investors. The retail investors are basically never the ones generating money from it either.

As far as ETH and other cryptos go, there is realistically 0 case for them being decentralized. Vitalik could not only kill it but even walking away from it would likely lead to a significant decline. It has tons of issues right now that Vitalik has promised to fix with the eventual Ethereum 2.0., and stepping away from it wouldn’t help.

Most tokens as well are generally tied to some promise of future developments and creators. The same is true for most NFT projects. The reason 99% of crypto and NFT projects fail is because they failed to deliver on various promises or bad dev teams, or poor innovation, or the founders just straight up stole from everyone. I could go more in-depth than this but go find any project and if there is a dev team, company, or person tied to it then it's likely, not decentralized. The success of that project is directly tied to the dev team, and if they wanted to, most hold large wallets and could sell it all off effectively killing the project. Not to mention, any crypto project that doesn’t have a locked contract can 100% be completely rugged and they can run off with all of the money.

Decentralization is undesirable

Even if they were decentralized then it would not be a good thing. The graphic above indicates this is effectively a rich person’s unregulated playground… so it’s exactly like… the very issue it tried to solve?

Not to mention, it’s supposedly meant to prevent things like 2008, but is regularly subjected to 50–90% drawdowns in value? Back when Bitcoin was only known in niche parts of the dark web and illegal markets, it would regularly drop as much as 90% and has a long history of massive drawdowns.

The majority of crypto is blatant scams, and ALL of crypto is a “zero” or “negative” sum game in that it relies on the next person buying for people to make money. This is not a good thing. This has yet to solve any issues around what caused the 2008 crisis. Rather, it seemingly amplified all of the bad parts of it actually. Crypto is the ultimate speculative asset and realistically has no valuation metric. People invest in Apple because it has revenue and cash on hand. People invest in crypto purely for speculation. The lack of regulation or accountability means this is amplified to the most extreme degree where anything goes: Fraud, deception, pump and dumps, scams, etc. This results in a pure speculative madhouse where, unsurprisingly, only the rich win. Not only that but the deck is massively stacked in against the everyday investor because you can’t do a pump and dump without money.

At the heart of all of this is still the same issues: greed and human nature. However, in setting out to solve the issues that allowed 2008 to happen, they made something that literally took all of the bad parts of 2008 and then amplified it. 2008, similarly, was due to greed and human nature but was allowed to happen due to lack of oversite, rich people's excessive leverage over the poor, and excessive speculation. Crypto trades SOLELY on human emotion, has 0 oversite, is pure speculation, and the power dynamic is just as bad or worse.

Why should each individual be their own bank?

This is the last one, and it's short and basic. Why do we want people functioning as their own bank? Everything we do relies on another person. Without the internet, and internet providers there is no crypto. We don’t have people functioning as their own home and car manufacturers or internet service providers. Sure, you might specialize in some of these things, but masquerading as your own bank is a horrible idea.

I am not in the banking industry so I don’t know all the issues with banking. Most people aren’t in banking or crypto, so they don’t know the ENDLESS scams associated with it, and they are bound to lose their money eventually.

If you appreciate this article and want to read more from me and authors like me, without any limits, consider buying me a coffee: buymeacoffee.com/calebnaysmith

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.