How did his money blind me?

Like most fledgling entrepreneurs, I wanted a mentor. I met as many nearby folks with "entrepreneur" in their LinkedIn biographies for coffee.

These meetings taught me a lot, and I'd suggest them to any new creator. Attention! Meeting with many experienced entrepreneurs means getting contradictory advice. One entrepreneur will tell you to do X, then the next one you talk to may tell you to do Y, which are sometimes opposites. You'll have to chose which suggestion to take after the chats.

I experienced this. Same afternoon, I had two coffee meetings with experienced entrepreneurs. The first meeting was with a billionaire entrepreneur who took his company public.

I met him in a swanky hotel lobby and ordered a drink I didn't pay for. As a fledgling entrepreneur, money was scarce.

During the meeting, I demoed the software I'd built, he liked it, and we spent the hour discussing what features would make it a success. By the end of the meeting, he requested I include a killer feature we both agreed would attract buyers. The feature was complex and would require some time. The billionaire I was sipping coffee with in a beautiful hotel lobby insisted people would love it, and that got me enthusiastic.

The second meeting was with a young entrepreneur who had recently raised a small amount of investment and looked as eager to pitch me as I was to pitch him. I forgot his name. I mostly recall meeting him in a filthy coffee shop in a bad section of town and buying his pricey cappuccino. Water for me.

After his pitch, I demoed my app. When I was done, he barely noticed. He questioned my customer acquisition plan. Who was my client? What did they offer? What was my plan? Etc. No decent answers.

After our meeting, he insisted I spend more time learning my market and selling. He ignored my questions about features. Don't worry about features, he said. Customers will request features. First, find them.

Putting your faith in results over relevance

Problems plagued my afternoon. I met with two entrepreneurs who gave me differing advice about how to proceed, and I had to decide which to pursue. I couldn't decide.

Ultimately, I followed the advice of the billionaire.

Obviously.

Who wouldn’t? That was the guy who clearly knew more.

A few months later, I constructed the feature the billionaire said people would line up for.

The new feature was unpopular. I couldn't even get the billionaire to answer an email showing him what I'd done. He disappeared.

Within a few months, I shut down the company, wasting all the time and effort I'd invested into constructing the killer feature the billionaire said I required.

Would follow the struggling entrepreneur's advice have saved my company? It would have saved me time in retrospect. Potential consumers would have told me they didn't want what I was producing, and I could have shut down the company sooner or built something they did want. Both outcomes would have been better.

Now I know, but not then. I favored achievement above relevance.

Success vs. relevance

The millionaire gave me advice on building a large, successful public firm. A successful public firm is different from a startup. Priorities change in the last phase of business building, which few entrepreneurs reach. He gave wonderful advice to founders trying to double their stock values in two years, but it wasn't beneficial for me.

The other failing entrepreneur had relevant, recent experience. He'd recently been in my shoes. We still had lots of problems. He may not have achieved huge success, but he had valuable advice on how to pass the closest hurdle.

The money blinded me at the moment. Not alone So much of company success is defined by money valuations, fundraising, exits, etc., so entrepreneurs easily fall into this trap. Money chatter obscures the value of knowledge.

Don't base startup advice on a person's income. Focus on what and when the person has learned. Relevance to you and your goals is more important than a person's accomplishments when considering advice.

Are MVPs outdated and have no place in modern product culture?

Frank Robinson coined "MVP" in 2001. In the same year as the Agile Manifesto, the first Scrum experiment began. MVPs are old.

The concept was created to solve the waterfall problem at the time.

The market was still sour from the .com bubble. The tech industry needed a new approach. Product and Agile gained popularity because they weren't waterfall.

More than 20 years later, waterfall is dead as dead can be, but we are still talking about MVPs. Does that make sense?

What is an MVP?

Minimum viable product. You probably know that, so I'll be brief:

[…] The MVP fits your company and customer. It's big enough to cause adoption, satisfaction, and sales, but not bloated and risky. It's the product with the highest ROI/risk. […] — Frank Robinson, SyncDev

MVP is a complete product. It's not a prototype. It's your product's first iteration, which you'll improve. It must drive sales and be user-friendly.

At the MVP stage, you should know your product's core value, audience, and price. We are way deep into early adoption territory.

What about all the things that come before?

Modern product discovery

Eric Ries popularized the term with The Lean Startup in 2011. (Ries would work with the concept since 2008, but wide adoption came after the book was released).

Ries' definition of MVP was similar to Robinson's: "Test the market" before releasing anything. Ries never mentioned money, unlike Jobs. His MVP's goal was learning.

“Remove any feature, process, or effort that doesn't directly contribute to learning” — Eric Ries, The Lean Startup

Product has since become more about "what" to build than building it. What started as a learning tool is now a discovery discipline: fake doors, prototyping, lean inception, value proposition canvas, continuous interview, opportunity tree... These are cheap, effective learning tools.

Over time, companies realized that "maximum ROI divided by risk" started with discovery, not the MVP. MVPs are still considered discovery tools. What is the problem with that?

Time to Market vs Product Market Fit

Waterfall's Time to Market is its biggest flaw. Since projects are sliced horizontally rather than vertically, when there is nothing else to be done, it’s not because the product is ready, it’s because no one cares to buy it anymore.

MVPs were originally conceived as a way to cut corners and speed Time to Market by delivering more customer requests after they paid.

Original product development was waterfall-like.

Time to Market defines an optimal, specific window in which value should be delivered. It's impossible to predict how long or how often this window will be open.

Product Market Fit makes this window a "state." You don’t achieve Product Market Fit, you have it… and you may lose it.

Take, for example, Snapchat. They had a great time to market, but lost product-market fit later. They regained product-market fit in 2018 and have grown since.

An MVP couldn't handle this. What should Snapchat do? Launch Snapchat 2 and see what the market was expecting differently from the last time? MVPs are a snapshot in time that may be wrong in two weeks.

MVPs are mini-projects. Instead of spending a lot of time and money on waterfall, you spend less but are still unsure of the results.

MVPs aren't always wrong. When releasing your first product version, consider an MVP.

Minimum viable product became less of a thing on its own and more interchangeable with Alpha Release or V.1 release over time.

Modern discovery technics are more assertive and predictable than the MVP, but clarity comes only when you reach the market.

MVPs aren't the starting point, but they're the best way to validate your product concept.

SaaS spending is 40/40/20. 40% of operating expenses should be R&D, 40% sales and marketing, and 20% G&A. We wanted to see the statistics behind the rules of thumb. Since October 2017, 73 SaaS startups have gone public. Perhaps the rule of thumb should be 30/50/20. The data is below.

30/50/20. R&D accounts for 26% of opex, sales and marketing 48%, and G&A 22%. We think R&D/S&M/G&A should be 30/50/20.

There are outliers. There are exceptions to rules of thumb. Dropbox spent 45% on R&D whereas Zoom spent 13%. Zoom spent 73% on S&M, Dropbox 37%, and Bill.com 28%. Snowflake spent 130% of revenue on S&M, while their EBITDA margin is -192%.

G&A shouldn't stand out. Minimize G&A spending. Priorities should be product development and sales. Cloudflare, Sendgrid, Snowflake, and Palantir spend 36%, 34%, 37%, and 43% on G&A.

Another myth is that COGS is 20% of revenue. Median and averages are 29%.

Where is the profitability? Data-driven operating income calculations were simplified (Revenue COGS R&D S&M G&A). 20 of 73 IPO businesses reported operational income. Median and average operating income margins are -21% and -27%.

As long as you're growing fast, have outstanding retention, and marquee clients, you can burn cash since recurring income that doesn't churn is a valuable annuity.

The data was compelling overall. 30/50/20 is the new 40/40/20 for more established SaaS enterprises, unprofitability is alright as long as your business is expanding, and COGS can be somewhat more than 20% of revenue.

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.