Generative AI hype: some thoughts

The sudden surge in "generative AI" startups and projects feels like the inverse of the recent "web3" boom. Both came from hyped-up pots. But while web3 hyped idealistic tech and an easy way to make money, generative AI hypes unsettling tech and questions whether it can be used to make money.

Web3 is technology looking for problems to solve, while generative AI is technology creating almost too many solutions. Web3 has been evangelists trying to solve old problems with new technology. As Generative AI evolves, users are resolving old problems in stunning new ways.

It's a jab at web3, but it's true. Web3's hype, including crypto, was unhealthy. Always expected a tech crash and shakeout. Tech that won't look like "web3" but will enhance "web2"

But that doesn't mean AI hype is healthy. There'll be plenty of bullshit here, too. As moths to a flame, hype attracts charlatans. Again, the difference is the different starting point. People want to use it. Try it.

With the beta launch of Dall-E 2 earlier this year, a new class of consumer product took off. Midjourney followed suit (despite having to jump through the Discord server hoops). Twelve more generative art projects. Lensa, Prisma Labs' generative AI self-portrait project, may have topped the hype (a startup which has actually been going after this general space for quite a while). This week, ChatGPT went off-topic.

This has a "fake-it-till-you-make-it" vibe. We give these projects too much credit because they create easy illusions. This also unlocks new forms of creativity. And faith in new possibilities.

As a user, it's thrilling. We're just getting started. These projects are not only fun to play with, but each week brings a new breakthrough. As an investor, it's all happening so fast, with so much hype (and ethical and societal questions), that no one knows how it will turn out. Web3's demand won't be the issue. Too much demand may cause servers to melt down, sending costs soaring. Companies will try to mix rapidly evolving tech to meet user demand and create businesses. Frustratingly difficult.

Anyway, I wanted an excuse to post some Lensa selfies.

These are really weird. I recognize them as me or a version of me, but I have no memory of them being taken. It's surreal, out-of-body. Uncanny Valley.

Recent investigations have shown that generative AI can boost hackers and misinformation spreaders.

Since its inception in late November 2022, OpenAI's ChatGPT has entertained and assisted many online users in writing, coding, task automation, and linguistic translation. Given this versatility, it is maybe unsurprising but nonetheless regrettable that fraudsters and mis-, dis-, and malinformation (MDM) spreaders are also considering ChatGPT and related AI models to streamline and improve their operations.

Malign actors may benefit from ChatGPT, according to a WithSecure research. ChatGPT promises to elevate unlawful operations across many attack channels. ChatGPT can automate spear phishing attacks that deceive corporate victims into reading emails from trusted parties. Malware, extortion, and illicit fund transfers can result from such access.

ChatGPT's ability to simulate a desired writing style makes spear phishing emails look more genuine, especially for international actors who don't speak English (or other languages like Spanish and French).

This technique could let Russian, North Korean, and Iranian state-backed hackers conduct more convincing social engineering and election intervention in the US. ChatGPT can also create several campaigns and various phony online personas to promote them, making such attacks successful through volume or variation. Additionally, image-generating AI algorithms and other developing techniques can help these efforts deceive potential victims.

Hackers are discussing using ChatGPT to install malware and steal data, according to a Check Point research. Though ChatGPT's scripts are well-known in the cyber security business, they can assist amateur actors with little technical understanding into the field and possibly develop their hacking and social engineering skills through repeated use.

Additionally, ChatGPT's hacking suggestions may change. As a writer recently indicated, ChatGPT's ability to blend textual and code-based writing might be a game-changer, allowing the injection of innocent content that would subsequently turn out to be a malicious script into targeted systems. These new AI-powered writing- and code-generation abilities allow for unique cyber attacks, regardless of viability.

OpenAI fears ChatGPT usage. OpenAI, Georgetown University's Center for Security and Emerging Technology, and Stanford's Internet Observatory wrote a paper on how AI language models could enhance nation state-backed influence operations. As a last resort, the authors consider polluting the internet with radioactive or misleading data to ensure that AI language models produce outputs that other language models can identify as AI-generated. However, the authors of this paper seem unaware that their "solution" might cause much worse MDM difficulties.

Literally False News

The public argument about ChatGPTs content-generation has focused on originality, bias, and academic honesty, but broader global issues are at stake. ChatGPT can influence public opinion, troll individuals, and interfere in local and national elections by creating and automating enormous amounts of social media material for specified audiences.

ChatGPT's capacity to generate textual and code output is crucial. ChatGPT can write Python scripts for social media bots and give diverse content for repeated posts. The tool's sophistication makes it irrelevant to one's language skills, especially English, when writing MDM propaganda.

I ordered ChatGPT to write a news piece in the style of big US publications declaring that Ukraine is on the verge of defeat in its fight against Russia due to corruption, desertion, and exhaustion in its army. I also gave it a fake reporter's byline and an unidentified NATO source's remark. The outcome appears convincing:

Worse, terrible performers can modify this piece to make it more credible. They can edit the general's name or add facts about current wars. Furthermore, such actors can create many versions of this report in different forms and distribute them separately, boosting its impact.

In this example, ChatGPT produced a news story regarding (fictional) greater moviegoer fatality rates:

Editing this example makes it more plausible. Dr. Jane Smith, the putative author of the medical report, might be replaced with a real-life medical person or a real victim of this supposed medical hazard.

Can deceptive texts be found? Detecting AI text is behind AI advancements. Minor AI-generated text alterations can upset these technologies.

Some OpenAI individuals have proposed covert methods to watermark AI-generated literature to prevent its abuse. AI models would create information that appears normal to humans but would follow a cryptographic formula that would warn other machines that it was AI-made. However, security experts are cautious since manually altering the content interrupts machine and human detection of AI-generated material.

How to Prepare

Cyber security and IT workers can research and use generative AI models to fight spear fishing and extortion. Governments may also launch MDM-defence projects.

In election cycles and global crises, regular people may be the most vulnerable to AI-produced deceit. Until regulation or subsequent technical advances, individuals must recognize exposure to AI-generated fraud, dating scams, other MDM activities.

A three-step verification method of new material in suspicious emails or social media posts can help identify AI content and manipulation. This three-step approach asks about the information's distribution platform (is it reliable? ), author (is the reader familiar with them? ), and plausibility given one's prior knowledge of the topic.

Consider a report by a trusted journalist that makes shocking statements in their typical manner. AI-powered fake news may be released on an unexpected platform, such as a newly created Facebook profile. However, if it links to a known media source, it is more likely to be real.

Though hard and subjective, this verification method may be the only barrier against manipulation for now.

AI language models:

How to Recognize an AI-Generated Article ChatGPT, the popular AI-powered chatbot, can and likely does generate medium.com-style articles.

AI-Generated Text Detectors Fail. Do This. Online tools claim to detect ChatGPT output. Even with superior programming, I tested some of these tools. pub

Why Original Writers Matter Despite AI Language Models Creative writers may never be threatened by AI language models.

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.

The same way Amazon killed bookstores. You just can’t see it yet.

Shultz globalized coffee. Before Starbucks, coffee sucked.

All accounts say 1970s coffee was awful.

Starbucks had three stores selling ground Indonesian coffee in the 1980s.

What a show!

A year after joining the company at 29, Shultz traveled to Italy for R&D.

He noticed the coffee shops' sense of theater and community and realized Starbucks was in the wrong business.

Integrating coffee and destination created a sense of community in the store.

Brilliant!

He told Starbucks' founders about his experience.

They disapproved.

For two years.

Shultz left and opened an Italian coffee shop chain like any good entrepreneur.

Starbucks ran into financial trouble, so the founders offered to sell to Shultz.

Shultz bought Starbucks in 1987 for $3.8 million, including six stores and a payment plan.

Starbucks is worth $100.79Billion, per Google Finance.

26,500 times Shultz's initial investment

Starbucks is releasing its own NFT Platform under Shultz and his early Vision.

This year, Starbucks Odyssey launches. The new digital experience combines a Loyalty Rewards program with NFT.

The side chain Polygon-based platform doesn't require a Crypto Wallet. Customers can earn and buy digital assets to unlock incentives and experiences.

They've removed all friction, making it more immersive and convenient than a coffee shop.

Brilliant!

NFTs are the access coupon to their digital community, but they don't highlight the technology.

They prioritize consumer experience by adding non-technical users to Web3. Their collectables are called journey stamps, not NFTs.

No mention of bundled gas fees.

Brady Brewer, Starbucks' CMO, said;

“It happens to be built on blockchain and web3 technologies, but the customer — to be honest — may very well not even know that what they’re doing is interacting with blockchain technology. It’s just the enabler,”

Rewards members will log into a web app using their loyalty program credentials to access Starbucks Odyssey. They won't know about blockchain transactions.

Join the waitlist here

Starbucks has just dealt its rivals a devastating blow.

It generates more than ten times the revenue of its closest competitor Costa Coffee.

The coffee giant is booming.

Starbucks is ahead of its competitors. No wonder.

They have an innovative, adaptable leadership team.

Starbucks' DNA challenges the narrative, especially when others reject their ideas.

I’m off for a cappuccino.

Here's how to be clear.

“I have only made this letter longer because I have not had the time to make it shorter.” — French mathematician, physicist, inventor, philosopher, and writer Blaise Pascal

Concise.

People want this. We tend to repeat ourselves and use unnecessary words.

Being vague frustrates readers. It focuses their limited attention span on figuring out what you're saying rather than your message.

Edit carefully.

“Examine every word you put on paper. You’ll find a surprising number that don’t serve any purpose.” — American writer, editor, literary critic, and teacher William Zinsser

How do you write succinctly?

Here are three ways to polish your writing.

1. Delete

Your readers will appreciate it if you delete unnecessary words. If a word or phrase is essential, keep it. Don't force it.

Many readers dislike bloated sentences. Ask yourself if cutting a word or phrase will change the meaning or dilute your message.

For example, you could say, “It’s absolutely essential that I attend this meeting today, so I know the final outcome.” It’s better to say, “It’s critical I attend the meeting today, so I know the results.”

Key takeaway

Delete actually, completely, just, full, kind of, really, and totally. Keep the necessary words, cut the rest.

2. Just Do It

Don't tell readers your plans. Your readers don't need to know your plans. Who are you?

Don't say, "I want to highlight our marketing's problems." Our marketing issues are A, B, and C. This cuts 5–7 words per sentence.

Keep your reader's attention on the essentials, not the fluff. What are you doing? You won't lose readers because you get to the point quickly and don't build up.

Key takeaway

Delete words that don't add to your message. Do something, don't tell readers you will.

3. Cut Overlap

You probably repeat yourself unintentionally. You may add redundant sentences when brainstorming. Read aloud to detect overlap.

Remove repetition from your writing. It's important to edit our writing and thinking to avoid repetition.

Key Takeaway

If you're repeating yourself, combine sentences to avoid overlap.

4. Simplify

Write as you would to family or friends. Communicate clearly. Don't use jargon. These words confuse readers.

Readers want specifics, not jargon. Write simply. Done.

Most adults read at 8th-grade level. Jargon and buzzwords make speech fluffy. This confuses readers who want simple language.

Key takeaway

Ensure all audiences can understand you. USA Today's 5th-grade reading level is intentional. They want everyone to understand.

5. Active voice

Subjects perform actions in active voice. When you write in passive voice, the subject receives the action.

For example, “the board of directors decided to vote on the topic” is an active voice, while “a decision to vote on the topic was made by the board of directors” is a passive voice.

Key takeaway

Active voice clarifies sentences. Active voice is simple and concise.

Bringing It All Together

Five tips help you write clearly. Delete, just do it, cut overlap, use simple language, and write in an active voice.

Clear writing is effective. It's okay to occasionally use unnecessary words or phrases. Realizing it is key. Check your writing.

Adding words costs.

Write more concisely. People will appreciate it and read your future articles, emails, and messages. Spending extra time will increase trust and influence.

“Not that the story need be long, but it will take a long while to make it short.” — Naturalist, essayist, poet, and philosopher Henry David Thoreau