I'm gadget-obsessed. So I shared my top 6 desk gadgets.

These gadgets improve my workflow and are handy for working from home.

Without further ado...

Computer light bar Xiaomi Mi

I've previously recommended the Xiaomi Mi Light Bar, and I still do. It's stylish and convenient.

The Mi bar is a monitor-mounted desk lamp. The lamp's hue and brightness can be changed with a stylish wireless remote.

Changeable hue and brightness make it ideal for late-night work.

Desk Mat 2.

I wasn't planning to include a desk surface in this article, but I find it improves computer use.

The mouse feels smoother and is a better palm rest than wood or glass.

I'm currently using the overkill Razer Goliathus Extended Chroma RGB Gaming Surface, but I like RGB.

Using a desk surface or mat makes computer use more comfortable, and it's not expensive.

Third, the Logitech MX Master 3 Mouse

The Logitech MX Master 3 or any from the MX Master series is my favorite mouse.

The side scroll wheel on these mice is a feature I've never seen on another mouse.

Side scroll wheels are great for spreadsheets and video editing. It would be hard for me to switch from my Logitech MX Master 3 to another mouse. Only gaming is off-limits.

Google Nest 4.

Without a smart assistant, my desk is useless. I'm currently using the second-generation Google Nest Hub, but I've also used the Amazon Echo Dot, Echo Spot, and Apple HomePod Mini.

As a Pixel 6 Pro user, the Nest Hub works best with my phone.

My Nest Hub plays news, music, and calendar events. It also lets me control lights and switches with my smartphone. It plays YouTube videos.

Google Pixel Stand, No. 5

A wireless charger on my desk is convenient for charging my phone and other devices while I work. My desk has two wireless chargers. I have a Satechi aluminum fast charger and a second-generation Google Pixel Stand.

If I need to charge my phone and earbuds simultaneously, I use two wireless chargers. Satechi chargers are well-made and fast. Micro-USB is my only complaint.

The Pixel Stand converts compatible devices into a smart display for adjusting charging speeds and controlling other smart devices. My Pixel 6 Pro charges quickly. Here's my video review.

6. Anker Power Bank

Anker's 65W charger is my final recommendation. This online find was a must-have. This can charge my laptop and several non-wireless devices, perfect for any techie!

The charger has two USB-A ports and two USB-C ports, one with 45W and the other with 20W, so it can charge my iPad Pro and Pixel 6 Pro simultaneously.

Summary

These are some of my favorite office gadgets. My kit page has an updated list.

Links to the products mentioned in this article are in the appropriate sections. These are affiliate links.

You're up! Share the one desk gadget you can't live without and why.

McMurtry Spéirling outperforms Tesla in speed and efficiency.

EVs were ridiculously slow for decades. However, the 2008 Tesla Roadster revealed that EVs might go extraordinarily fast. The Tesla Model S Plaid and Rimac Nevera are the fastest-accelerating road vehicles, despite combustion-engined road cars dominating the course. A little-known firm beat Tesla and Rimac in the 0-60 race, beat F1 vehicles on a circuit, and boasts a 350-mile driving range. The McMurtry Spéirling is completely insane.

Mat Watson of CarWow, a YouTube megastar, was recently handed a Spéirling and access to Silverstone Circuit (view video above). Mat ran a quarter-mile on Silverstone straight with former F1 driver Max Chilton. The little pocket-rocket automobile touched 100 mph in 2.7 seconds, completed the quarter mile in 7.97 seconds, and hit 0-60 in 1.4 seconds. When looking at autos quickly, 0-60 times can seem near. The Tesla Model S Plaid does 0-60 in 1.99 seconds, which is comparable to the Spéirling. Despite the meager statistics, the Spéirling is nearly 30% faster than Plaid!

My vintage VW Golf 1.4s has an 8.8-second 0-60 time, whereas a BMW Z4 3.0i is 30% faster (with a 0-60 time of 6 seconds). I tried to beat a Z4 off the lights in my Golf, but the Beamer flew away. If they challenge the Spéirling in a Model S Plaid, they'll feel as I did. Fast!

Insane quarter-mile drag time. Its road car record is 7.97 seconds. A Dodge Demon, meant to run extremely fast quarter miles, finishes so in 9.65 seconds, approximately 20% slower. The Rimac Nevera's 8.582-second quarter-mile record was miles behind drag racing. This run hampered the Spéirling. Because it was employing gearing that limited its top speed to 150 mph, it reached there in a little over 5 seconds without accelerating for most of the quarter mile! McMurtry can easily change the gearing, making the Spéirling run quicker.

McMurtry did this how? First, the Spéirling is a tiny single-seater EV with a 60 kWh battery pack, making it one of the lightest EVs ever. The 1,000-hp Spéirling has more than one horsepower per kg. The Nevera has 0.84 horsepower per kg and the Plaid 0.44.

However, you cannot simply construct a car light and power it. Instead of accelerating, it would spin. This makes the Spéirling a fan car. Its huge fans create massive downforce. These fans provide the Spéirling 2 tonnes of downforce while stationary, so you could park it on the ceiling. Its fast 0-60 time comes from its downforce, which lets it deliver all that power without wheel spin.

It also possesses complete downforce at all speeds, allowing it to tackle turns faster than even race vehicles. Spéirlings overcame VW IDRs and F1 cars to set the Goodwood Hill Climb record (read more here). The Spéirling is a dragstrip winner and track dominator, unlike the Plaid and Nevera.

The Spéirling is astonishing for a single-seater. Fan-generated downforce is more efficient than wings and splitters. It also means the vehicle has very minimal drag without the fan. The Spéirling can go 350 miles per charge (WLTP) or 20-30 minutes at full speed on a track despite its 60 kWh battery pack. The G-forces would hurt your neck before the battery died if you drove around a track for longer. The Spéirling can charge at over 200 kW in about 30 minutes. Thus, driving to track days, having fun, and returning is possible. Unlike other high-performance EVs.

Tesla, Rimac, or Lucid will struggle to defeat the Spéirling. They would need to build a fan automobile because adding power to their current vehicle would make it uncontrollable. The EV and automobile industries now have a new, untouchable performance king.

And morning water bottle struggles.

Welcome to my TikTok, where I share my stay-at-home life! I'll show you my usual day from morning to night.

I rise early to prepare my guy iced coffee. I make matcha, my favorite drink. I also fill our water bottles, which takes time and effort, so I record and describe the procedure. As you see me perform the unthinkable by putting a water bottle in a soda machine, you'll see my magnificent but unowned condo. My lover has everything, including:

1. In the living room, a sizable velvet alabaster divan. I was unable to use the words white or sofa in place of alabaster or a divan since they are insufficiently elegant and do not adequately convey how opulent the item is. The price tag on the divan was another huge feature; I'm sure my lover wouldn't purchase any furniture for less than $20k because it would be beneath him.

2. A plush Swiss coffee-colored Tabriz carpet. Once more, white is a color associated with the underclass; for us, the wealthy, it's alabaster or swiss coffee. Sorry, my boyfriend is wealthy; I'm truly in the same situation. And yet, I’m the one whos freeloading off of him, not you haha!

3. Soft translucent powder is the hue of the vinyl wallcoverings. I merely made up the name of that hue, but I have to maintain the online character I've established. There is no room for adopting language typical of peasant people; I must reiterate that I am wealthy while they are not.

I rest after filling our water bottles. I'm really fatigued from chores. My boyfriend is skeptical about hiring a housekeeper and cook. Does he assume I'm a servant or maid? I can't be overly demanding or throw a tantrum since he may replace me with a younger version. Leonardo Di Caprio's fault!

After the break, I bring my lover a water bottle. He's off to work with my best wishes. After cleaning the shower, I text my BF saying I broke a nail. He charged $675 for a crystal-topped shellac manicure. Lucky me!

After this morning's crazy choirs, especially the water bottle one, I'm famished. I dress quickly and go to the neighborhood organic-vegan-gluten-free-sugar-free-plasma-free-GMO-free-HBO-free breakfast place. Most folks can't afford $17.99 for a caffeine-free-mushroom-plus-mud-and-electrolytes morning beverage. It goes nicely with my matcha. Eggs Benedict cost $68. English muffins are off-limits. I can't make myself obese. My partner said he'd swap me for a 19-year-old Eastern European if I keep eating bacon.

I leave no tip since tipping is too much pressure and math for me, so I go shopping.

My shopping adventures have gotten monotonous. 47 designer bags and 114 bag covers Birkins need their own luggage. My babies! I've never caught my BF with a baby. I have sleeping medications and a turkey baster. Tatiana is much younger and thinner than me, so I can't lose him to her. The goal is to become a stay-at-home wife shortly. A turkey baster is essential.

After spending $955 on La Mer lotions and getting a crystal manicure, I nap. Before my boyfriend's return, I can nap for 5 hours.

I wake up around 4 pm — it’s time to prepare dinner. Yes, I said “prepare for dinner,” not “prepare dinner.” I have crystals on my nails! Do you really think I would cook? No way.

My husband's arrival still requires much work. I clean the kitchen, get cutlery and napkins. I order UberEats while my BF is 30-45 minutes away.

Wagyu steaks with Matsutake mushroom soup today. I pick desserts for my lover but not myself. Eastern European threat?

When my BF gets home from work, we eat. I don't believe in tipping UberEats drivers. If he wants to appreciate life's finer things, he should locate a rich woman.

After eating, we plan our getaway. I requested Aruba's fanciest hotel for winter and expect a butler. We're bickering over who gets the butler. We may need two.

Day's end, I'm exhausted. Stay-at-home girlfriends put in a lot of time and work. Work and duties are never-ending.

Before bed, I shower and use a liquid gold mask in my 27-step makeup procedure. It's a French luxury brand, not La Mer.

Here's my day.

Note: I like satire and absurd trends. Stay-at-home-girlfriend TikTok videos have become popular recently.

I don't shame or support such agreements; I'm just an observer. Thanks for reading.

https://tomconnor.me/wp-content/uploads/2021/08/10x-Engineer-Mental-Models.pdf

I keep returning to the same mental models and tricks after writing and reading about a wide range of topics.

Top 12 mental models

12.

Survival bias - We perceive the surviving population as remarkable, yet they may have gotten there through sheer grit.

Survivorship bias affects us in many situations. Our retirement fund; the unicorn business; the winning team. We often study and imitate the last one standing. This can lead to genuine insights and performance improvements, but it can also lead us astray because the leader may just be lucky.

11.

The Helsinki Bus Theory - How to persevere Buss up!

Always display new work, and always be compared to others. Why? Easy. Keep riding. Stay on the fucking bus.

10.

Until it sticks… Turning up every day… — Artists teach engineers plenty. Quality work over a career comes from showing up every day and starting.

9.

WRAP decision making process (Heath Brothers)

Decision-making WRAP Model:

W — Widen your Options

R — Reality test your assumptions

A — Attain Distance

P — Prepare to be wrong or Right

8.

Systems for knowledge worker excellence - Todd Henry and Cal Newport write about techniques knowledge workers can employ to build a creative rhythm and do better work.

Todd Henry's FRESH framework:

1. Focus: Keep the start in mind as you wrap up.

2. Relationships: close a loop that's open.

3. Pruning is an energy.

4. Set aside time to be inspired by stimuli.

5. Hours: Spend time thinking.

7.

Black Box Thinking…..

BBT is learning from mistakes. Science has transformed the world because it constantly updates its theories in light of failures. Complexity guarantees failure. Do we learn or self-justify?

6.

The OODA Loop - Competitive advantage

O: Observe: collect the data. Figure out exactly where you are, what’s happening.

O: Orient: analyze/synthesize the data to form an accurate picture.

D: Decide: select an action from possible options

A: Action: execute the action, and return to step (1)

Boyd's approach indicates that speed and agility are about information processing, not physical reactions. They form feedback loops. More OODA loops improve speed.

5.

Know your Domain 

Leaders who try to impose order in a complex situation fail; those who set the stage, step back, and allow patterns to develop win.

https://vimeo.com/640941172?embedded=true&source=vimeo_logo&owner=11999906

4.

The Three Critical Gaps

• Information Gap - The discrepancy between what we know and what we would like to know

• Gap in Alignment - What individuals actually do as opposed to what we wish them to do

• Effects Gap - the discrepancy between our expectations and the results of our actions

3.

Theory of Constraints — The Goal  - To maximize system production, maximize bottleneck throughput.

• Goldratt creates a five-step procedure:

1. Determine the restriction

2. Improve the restriction.

3. Everything else should be based on the limitation.

4. Increase the restriction

5. Go back to step 1 Avoid letting inertia become a limitation.

Any non-constraint improvement is an illusion.

2.

Serendipity and the Adjacent Possible - Why do several amazing ideas emerge at once? How can you foster serendipity in your work?

You need specialized abilities to reach to the edge of possibilities, where you can pursue exciting tasks that will change the world. Few people do it since it takes a lot of hard work. You'll stand out if you do.

Most people simply lack the comfort with discomfort required to tackle really hard things. At some point, in other words, there’s no way getting around the necessity to clear your calendar, shut down your phone, and spend several hard days trying to make sense of the damn proof.

1.

Boundaries of failure - Rasmussen's accident model.

Rasmussen modeled this. It has economic, workload, and performance boundaries.

The economic boundary is a company's profit zone. If the lights are on, you're within the economic boundaries, but there's pressure to cut costs and do more.

Performance limit reflects system capacity. Taking shortcuts is a human desire to minimize work. This is often necessary to survive because there's always more labor.

Both push operating points toward acceptable performance. Personal or process safety, or equipment performance.

If you exceed acceptable performance, you'll push back, typically forcefully.

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.