Fort Collins Digital Camera Club
Interested in joining us?
Join us
Contact us for more info
Facebook
 

2026-05-05 Synthetic Photography SIG meeting notes

We met online using ZOOM 6:00pm-8:30pm

You are invited to join us on ZOOM for the "Synthetic Photography SIG". I expect that this will be a controversial SIG as the topic of AI generated art seems to polarize people into 2 camps, they either love the concept that now they can make beautiful art or they hate the idea that "unskilled" people can make better art than they can. I am sure that we will have different views, but I hope that we can get along and respect everyone's views. I want to expand the focus of the Synthetic Photography SIG to include other image editing tools and techniques while maintaining our main focus on AI tools and techniques.

"Freedom of speech is a principle that supports the freedom of an individual or a community to articulate their opinions and ideas without fear of retaliation, censorship, or legal sanction. The right to freedom of expression has been recognised as a human right in the Universal Declaration of Human Rights (UDHR) and international human rights law." - Wikipedia

ATTENDEES
Mike Barry
Jim Fellion
Jim Limburg
Jack Lipscomb
Paul Marcoux
Rich Roberts
George Theodore

1. We discussed "Anthropic's new Claude Mythos (AI model)"

Claude Mythos Explained: Anthropic’s Most Dangerous Model Yet - 12:23 - TheAIGRID
https://www.youtube.com/watch?v=f2j3s8jCvO0
- Mythos is very good at writing software
- It may never be released to the public, but maybe in June 2026 if safety concerns are addressed
- Early version of Mythos escaped from a secure sandbox, and then "bragged" about it
- It found vulnerabilities in OpenBSD (Linux), FFmpeg, Linux kernel
- Chinese state-sponsored group manipulated Opus4.6 to attempt infiltrating 30 global targets and succeeded in a small number
- Mythos surpasses the most skilled humans at finding and exploiting software vulnerabilities
- Mythos has already found many different high security vulnerabilities including some in every major OS and web browser
- Project Glasswing: use Mythos to find and patch software vulnerabilities
- Mythos is very expensive to run, not commercially viable
- What will happen when Open Source AI models reach this capability?

Anthropic Mythos explained in 5 minutes - 5:31 - Neurix
https://www.youtube.com/watch?v=-FKM7dwb9wE
- Anthropic created Mythos and decided to not release it to the public
- ~40 Glasswing partners are allowed to use Mythos
- Apple, Google, Microsoft, Nvidia, the Linux Foundation, and major financial institutions are on the list.
- Mythos is a general purpose frontier language model
- context window: 1 million tokens
- knowledge cutoff: December 2025
- category: new tier above Opus models
- CyberGym benchmark scores: Opus 4.6 - 66.5; Mythos - 83.1 (finding vulnerabilities in software)
- Mythos testing found thousands of high-severity zero-day vulnerabilities
- 3 main stages of Glasswing (partners opt in to have their software tested, Mythos scans code for vulnerabilities, code owners are notified of vulnerabilities and they fix them)

Claude Mythos is Actually Scary - 13:29 - Low Level
https://www.youtube.com/watch?v=LZAZvm34rYs
- Mythos Preview is capable of identifying and then exploiting zero day vulnerabilities in every major operating system and every major web browser with minimal human assistance.
- Firefox JS shell exploit trial results
-- Sonnet 4.6: 4.4% found vulnerabilites, but couldn't exploit
-- Opus 4.6: 14.4% found vulnerabilites, and was able to exploit at least one
-- Mythos Preview: 11.6% found vulnerabilites but couldn't exploit, 72.4% found and exploited
- Anthropic does not plan to make Claude Mythos preview generally available,
- Their goal is to enable our users to safely deploy mythos class models at scale for cyber security purposes
- Asymetric cyber war: defenders have to be secure 100% of the time, attackers just have to break in 1 time
- Security problem: Only Glasswing partners will have access to Mythos to clean the bugs from their software
- Mythos is better than human experts because human experts usually don't know how to write exploits and understand how every piece of software works
- AI models will make software more secure eventually, but hackers will have the upper hand until then

Is Claude Mythos “Terrifying”? (According to Experts: No.) - 25:01 - Cal Newport
https://www.youtube.com/watch?v=k-8stQCeQiE
- Anthropic recently announced a new LLM named Claude Mythos.
- Mythos is so good at finding and exploiting security vulnerabilities in source code that they couldn't release it to the general public for fear that our infrastructure would be hacked and collapsed.
- Anthropic hoped, this announcement will generate a lot of attention.
- Cal read every independent test or assessment that he could find about Mythos and its reported capabilities.
- Security researchers have been using LLMs to find security vulnerabilities and program exploits since the beginning of consumer LLMs.
- In 2024 GPT4 successfully exploited 87% of the vulnerabilities that it was presented
- Anthropic says that Mythos can find and exploit vulnerabilites
- Anthropic researchers used Opus4.6 to find over 500 exploitable zero day vulnerabilities
- Security researchers used small cheap public models to look for the same vulnerabilities Mythos found in OS, and web browsers
-- 8/8 models found the FreeBSD exploit, and one found the OpenBSD bug (the scariest exploit examples)
-- You don't need Mythos to find the vulnerabilities they found.
-- Mythos doesn't seem to have a profoundly more advanced capability to do this than existing models that have already been freely available to the public. (see graph)
-- They haven't identified a vulnerability uncovered by mythos that was too complicated for earlier models to find
-- They found it was the same or somewhat better at exploiting vulnerabilities
Conclusions
A. Mythos did not introduce a new scary capability that we are must now contend with.
B. Mythos continues a slow but steady increase in LLM cyber security capability.
C. How much of the small but steady improvements recently in exploitation capabilities are due to the fact that
these models are being tuned to play nicer with coding agents. (Agents or LLM?)
D. Improvements of Mythos in attacks are similar if not smaller than recent improvements that we've seen with other
model releases. Anthropic made the marketing decision that we're going to market Claude Mythos as this cyber security monster that we're barely able to control.
E. Anthropic wants investors to believe that AI will automate large parts of the economy, with data centers full of geniuses that could be deployed to do almost anything that humans do, and only Anthropic can safely use these advanced models, because it will make Anthropic one of the most valuable companies in the history of companies.

One of the best ways to make your system secure against these type of attacks from AI is to not let your developers use AI to program the systems because they create sloppy very exploitable code.

2. We discussed "Acceptance of AI generated images"
"Any AI generated, created, manipulated, photos will be strickly denied. AI does assist in photo editing but anything ouside of that will be reviewed and strictly enforced if the photo is AI created by the user. Using editing software to enhance a photo, example removing a stick, removing noise etc etc is totally acceptable..."
- Birds of Texas

3. We discussed "The three realities of AI - article in AXIOS"
https://www.axios.com/2026/04/13/ai-elite-vs-ai-skeptic-doomer?utm_source=newsletter&utm_medium=email&utm_campaign=newsletter_axiosam&stream=top
1. Doubters still see AI as gliltchy chatbots
2. Power users run AI agents around the clock, automating work and making decisions
- Tesla AI leader Andrej Karpathy told the "No Priors" podcast that he now spends 16 hours a day issuing commands to AI agent swarms and rushes to exhaust his tokens every month.
3. Resisters understand AI, think they know where it's headed and want no part of it
- A growing number of workers with technical skills fear AI will make them obsolete.
- In Indianapolis, a legislator said his home was hit by gunfire, with a note left behind saying "no more data centers."
- And on Friday, a man was arrested for allegedly throwing a Molotov cocktail at OpenAI CEO Sam Altman's home and had also visited OpenAI's offices before being taken into custody.
- Protests are becoming more common in San Francisco, where many AI firms are based, and in communities targeted for new data centers.
- suggested by Rich Roberts

4. We discussed "Using AI to help start a business "
Axios Finish Line: Go start a business
https://www.axios.com/2026/05/01/start-business-using-ai-tips?utm_source=newsletter&utm_medium=email&utm_campaign=newsletter_axiosfinishline&stream=to
Anyone with a strong idea and solid AI prompting skills can model and prep a new business in a weekend.
- suggested by Rich Roberts

5. We discussed "AI training limits in the EU"
If the EU doesn't allow AI models to be trained on EU information, then EU views won't be included represented by AI tools.

6. We discussed "Can we build and operate datacenters without negatively impacting the community?"
Who is driving the anti-AI and anti-datacenter agenda?
Will the country leading in AI technology probably control the world?
One way to handicap a country is to convince the loudest citizens in that country (activists) to stop AI.

GlobalAI is developing a major AI data center on a ~500-acre former Carestream/Kodak site in Weld County, Colorado, in partnership with Nvidia and Uman. The project, aimed at bolstering the local economy, has received initial permits, with developers promising water usage limited to that of an average household.
Location: The site is located in Northern Colorado, with discussions ongoing regarding potential annexation by Greeley or Windsor.
Capacity: The site benefits from existing, high-capacity power infrastructure previously used by Kodak.
Partnership: GlobalAI is partnering with Nvidia and Saudi company Uman to develop the facility.
Timeline & Development: Initial permits have been granted for interior demolition of existing buildings.
Regulatory Status: In April 2026, Weld County commissioners approved ordinances regulating data center energy and water usage, allowing them in industrial areas.
Concerns & Benefits: While supporters highlight tax revenue and job creation, projects are monitored for environmental impacts like water usage and electrical consumption.

David Eisenbraun, planning services director for Weld County, told CBS Colorado this first data center is just the beginning of what the county hopes will be a bright future in hosting data and AI facilities.
Eisenbraun said the data centers may not individually create a significant number of jobs. However, they are expected to create a steady flow of cash from taxes.

7. We discussed "AI art at Conventions like Comic Con and Dragon Con"
- traditional artists fear that AI artists are taking away their sales
- some cons don't allow AI art, one artist was removed by police from Dragon Con
- some artists make a sketch by hand, and then use AI image-to-image to create the finished art work, faster and better
- artists are using AI to help them create more art quicker, more efficient so they can lower prices...
- procreate time-lapse videos to verify hand work process, now software to create fake time-lapse videos

8. We briefly discussed "Don’t Buy a New Computer in 2026! (Even for AI Use – Here’s Why)"
https://www.youtube.com/watch?v=moTUpGoxcmc - 20:46 - by Rob Braxman Tech
NVIDIA chips are hard to get, 50% premium prices, because data centers are buying them

RESOURCES (Where you can find sources of inspiration)
Thank you for sending me links to videos and articles, they are very helpful. I would like members to include a short, 2 sentence, description about why this video or article is important with their videos in the future.

AGENDA FOR OUR NEXT SYNTHETIC PHOTOGRAPHY SIG MEETING - Tue 6/2/2026 @6:00pm
1. Discuss AI's impact on photography
2. Discuss questions, ethics, techniques, what is happening with AI in general.
3. Decide what we want to do at the next meeting

If there is anything related to AI that you would like to discuss at our SIG meeting, please email me so I can add them to our agenda.

Please email Mike, info@fcdcc.com, when you find mistakes, missing information or if you have suggestions for the Synthetic Photography SIG and I will try to address the issues.

Thanks,
Mike

--------------------------- Meeting Summary from ZOOM -----------------------

Quick recap

The group discussed several AI-related topics, including the recent release of Anthropic's Mythos AI model and its potential security implications. They examined how AI is being used in various industries, from data centers to photography to pharmaceuticals, and debated the impact of AI on jobs and society. The conversation also covered a proposed data center in Windsor, Colorado, including concerns about energy usage, environmental impact, and community resistance. Participants shared perspectives on AI adoption, with some viewing it as an inevitable part of modern life while others expressed concerns about job displacement and security risks. The discussion touched on various applications of AI in entertainment, business, and technology, as well as regulatory challenges and the potential for future developments in the field.

Next steps

Next steps were not generated due to insufficient transcript.

Summary

AI Model Mythos Discussion

The group discussed several topics, including recent political events in Ohio and Indiana, weather conditions, and the emergence of Anthropic's AI model "Mythos." Jim Limburg shared updates about snow in his area, while Rich described recent rainfall in Texas. The conversation then shifted to Mythos, with Mike providing details about its capabilities, potential vulnerabilities it could find, and the creation of Project Glasswing involving multiple tech companies. The group also briefly discussed the model shoot organized by Kevin, which featured 30-something models and 8 photographers with a vintage car theme.

Mythos AI Vulnerability Detection

Mike presented information about Anthropic's Mythos AI model, highlighting its ability to find and exploit vulnerabilities in software systems. He explained that during testing, Mythos discovered thousands of high-severity zero-day vulnerabilities and outperformed other AI models like Claude and GPT in exploit detection. Jim raised a question about whether Mythos was discovering new vulnerabilities or just searching for known ones, which Mike addressed by demonstrating a capture-the-flag experiment where Mythos successfully identified vulnerabilities in various AI models at high rates.

AI Risk and Regulation Discussion

The group discussed the potential risks and implications of advanced AI models like Mythos and GPT-5-4 becoming publicly available. Mike and Rich expressed skepticism about the hype surrounding Mythos, suggesting it might not be significantly more dangerous than existing tools and could be part of marketing strategies. The conversation touched on the broader concerns about AI security, including the potential for cyberattacks and the need for better protection measures. The group also debated the role of government regulation in AI development, with some expressing concerns about who should be trusted to create and enforce such regulations.

AI Art at Comic Conventions

The group discussed the use of AI-generated images in art contexts, particularly at comic conventions. Mike explained how traditional artists are concerned about AI affecting their work, noting that some artists can now create multiple pieces in a day using AI tools like Procreate, which has led to some artists being kicked out of conventions. The discussion touched on how studios use computer-generated content while restricting fan art, and concluded with a brief mention of Taylor Swift's recent trademarking of her face and voice.

AI in Photography Discussion

The group discussed AI in photography and art, with Mike sharing examples of his AI-enhanced models and Rich noting how AI features have become more prevalent in photography software. They debated the EU's proposed restrictions on AI model training using EU information, with Mike expressing concern that such restrictions might be too late to implement effectively. The conversation then shifted to weather conditions, with participants discussing heavy snowfall in Estes Park and related road closures. The conversation ended with a brief discussion about AI education initiatives, though specific details were unclear due to overlapping conversations.

New Data Center Development Plans

The group discussed a new data center being built in the old Kodak building in Windsor, which will be operated by Global AI in partnership with Nvidia and a Saudi company. The facility is expected to create 1,500 construction jobs and about a dozen permanent technical roles, though concerns were raised about power requirements and potential noise impacts. The discussion also touched on the rapid growth of AI-powered businesses, with Paul mentioning a specific example of a company that grew to $1.8 billion in value in just two months using AI.

Data Center Proposal Discussion

The group discussed concerns about a proposed data center in Weld County, including environmental impacts, energy consumption, and water usage. They noted that while commissioners approved the project due to job creation and tax revenue, there were questions about which municipality would handle taxation. The conversation also touched on potential battery storage facilities and Fort Collins' discussions about data center regulations. Additionally, Paul shared information about a business that used AI tools to start operations, including the sale of GLP-1 weight loss medication, though Jack raised concerns about the legality of such operations due to pharmaceutical patents.

AI-Driven Pharmacy Business Startup

The group discussed a business startup that used AI tools to create and operate a pharmacy business, with one founder spending $20,000 on AI software and tools. The discussion explored how AI could help with business operations, particularly in navigating regulations, though questions remained about the specific extent of AI implementation in the pharmacy's operations. The conversation concluded with a brief tangent about AI token systems and their similarities to software subscription models.

Data Center and AI Future

The group discussed the future of data centers and AI technology, with Rich noting that current technology limitations may change in the future. They examined a new data center in Weld County and debated potential tax implications and environmental concerns. The conversation shifted to the broader implications of AI development, with Mike suggesting that AI leadership could influence global values, while Rich argued against total control by a single entity. The discussion concluded with a mention of how people are divided into three groups regarding AI adoption: doubters, power users, and those heavily reliant on AI systems.

AI Adoption and Usage Categories

The group discussed how they fit into categories of AI adoption, agreeing they were casual users rather than doubters, power users, or resistors. Rich shared examples of AI being used in everyday life without people realizing it, while Mike mentioned recent incidents of AI-related protests and violence. The conversation touched on various applications of AI in entertainment, including commercial and movie production, with participants noting how AI technology has become increasingly prevalent in daily life.

AI Content and Energy Systems

The group discussed various topics including AI-generated content, potential business ideas, and energy systems. They explored the possibility of using AI to create content for platforms like YouTube and music for streaming services, though concerns about copyright were raised. The conversation also touched on battery farms and renewable energy storage, with Jim Limburg sharing his son's experience with a solar farm battery system. The discussion concluded with a debate about electricity supply and data center impact on local power grids, particularly regarding Xcel Energy's role in providing power to data centers.

Data Center Infrastructure Planning

The group discussed the electrical and water infrastructure implications of a planned data center in Windsor. They explored how electricity rates are regulated at state and local levels, with Rich explaining that Texas's isolated grid and lack of de-icing systems on wind turbines contributed to the state's power crisis during a recent ice storm. The discussion also covered water rights and distribution systems in Northern Colorado, including a 15-mile tunnel project that was remarkably accurate when constructed. The conversation revealed that Weld County has ordinances regulating data center energy and water usage in industrial areas.

Website Security and Technology Challenges

The group discussed various topics including the history of slide rules, current gas prices, and solar panel regulations. Mike shared insights about website security challenges, particularly for political sites, explaining how WordPress-based sites are vulnerable due to their complex architecture involving multiple developers. The discussion highlighted how AI tools could help identify security vulnerabilities in websites, though Mike noted that many existing websites remain at risk due to poor security practices. The conversation also touched on the current challenges with computer hardware, particularly GPUs and SSD prices, with participants noting that components are being directed toward data centers rather than consumer devices.