My Reviews Archives • The Open Source Research (TOSR Co.)

July 16, 2023July 16, 2023

Discovering Presentations Through Artificial Intelligence

I have done a lot recently around image creation. I have also explored other AI technologies. For this post, I used Tome which is a Artificial Intelligence PowerPoint creation tool. This application will create images but will also present a slide containing text similar to Microsoft’s PowerPoint. My first attempt was to combine my fondness of history with this technology. I choose the Medici since I have once visited Florence, Italy from which the were one of the power families during the Renaissance. My primary focus is the technology and discovering ways to use this technology to incorporate into daily and business usage.

Presentation link: https://tome.app/tosrco/the-medici-effect-how-a-family-of-bankers-shaped-the-reformation-clixlnwa204d0oj3hm7dr8049

The following resulted from a prompt within ChatGPT. The prior paragraph was written by a human (me). Another interesting thing is that the below description and the information in the linked Presentation can be different which highlights the differences between the artificial intelligence applications. I have kept the citations provided by ChatGPT in order to keep the correct attribution to the authors to which ChatGPT references.

Title: The Medici Family’s Profound Influence on the Renaissance

Introduction:

The Medici family, also known as the House of Medici, played a pivotal role in shaping the Renaissance, a period of remarkable scientific, artistic, and cultural rebirth in Europe. Originating in the 13th century, the Medicis amassed wealth and political power through their success in commerce and banking in Florence, Italy. However, it was the rise of Cosimo de Medici in 1434 that catapulted the family’s influence to new heights, transforming Florence into the cradle of the Renaissance [1]. Through their patronage of the arts and humanities, the Medici family revolutionized the intellectual and artistic landscape of Europe, leaving an indelible mark on history.

Cosimo de Medici and the Renaissance Flourish:

Cosimo de Medici, also known as Cosimo the Elder, ascended to political power in 1434, ruling Florence as an uncrowned monarch. Despite leading a simple life, he dedicated himself to the patronage of the humanities. Cosimo’s unwavering support nurtured the talents of renowned artists such as Lorenzo Ghiberti, Filippo Brunelleschi, Donatello, and Fra Angelico [1]. Under his rule, the Italian Renaissance flourished, with Florence emerging as the cultural epicenter of Europe. The Medici family’s significant contributions to the arts transformed the cityscape and birthed enduring masterpieces.

Lorenzo de Medici: A Magnificent Patron:

Lorenzo de Medici, also known as Lorenzo the Magnificent, followed in his father’s footsteps as a passionate supporter of the arts. He personally engaged in poetry and provided patronage to legendary artists like Sandro Botticelli, Leonardo da Vinci, and Michelangelo [1]. Lorenzo’s premature death at the age of 43 left a void, but his legacy endured. His eldest son, Piero, briefly succeeded him but faced public ire due to accepting an unfavorable peace treaty with France, leading to his exile from Florence in 1494 [1]. However, Lorenzo’s younger brother, Giovanni, played a crucial role in the Medici family’s return to power in 1512, opening a new chapter in their influential reign.

Catherine de Medici’s Royal Legacy:

Piero’s son, Lorenzo, reclaimed power in Florence, while his daughter, Catherine de Medici, became queen of France after marrying King Henry II. Three of her four sons went on to rule France [1]. Catherine’s ascent to the French throne expanded the Medici family’s reach and influence beyond Florence. Their connections with European royalty strengthened their position, and the Medicis’ genes intermingled with numerous royal families, perpetuating their impact for generations to come.

Medici Patronage and Artistic Expression:

The Medici family’s patronage of the arts played a pivotal role in the development of Renaissance aesthetics. Their support provided artists with the means to pursue their creative endeavors, leading to groundbreaking artistic achievements. One notable example is the San Marco Altarpiece, commissioned shortly after the Medici family’s return to Florence. Painted by Fra Angelico, the altarpiece showcases religious devotion while aligning the Medici family with the Dominican order, emphasizing their societal ideals [3].

The Palazzo Medici, a testament to Medici authority, blends traditional and progressive architectural styles. Designed by Michelozzo, it symbolizes the family’s influence, harmonizing elements from the early Renaissance and classical periods [3]. Inside the palace, the classical-style courtyard, adorned with ancient sculptures and reliefs, further links the Medicis to a prosperous artistic era.

Legacy and Lasting Impact:

The Medici family’s influence extended far beyond their patronage of the arts. They left an enduring impact on European history, with their descendants ascending to papal thrones and ruling over prominent nations [1]. Their role as tastemakers, cultural influencers, and patrons of learning contributed to the diffusion of Renaissance ideals across Europe. The Medici family’s impact on the Renaissance cannot be overstated; their legacy forever altered the course of history.

Conclusion:

The Medici family’s rise to power in Florence during the Renaissance era propelled the city to new heights of intellectual, artistic, and cultural brilliance. Through their patronage, the Medicis fostered an environment conducive to innovation and creativity, supporting legendary artists and leaving an indelible mark on Europe’s cultural landscape. Their influence spread beyond Florence, as their descendants held positions of power in the church and ruled over influential nations. The Medici family’s profound impact on the Renaissance cemented their legacy as one of the most important dynasties in history.

References: [1] History.com. Medici Family: Cosimo, Lorenzo & Catherine. Available at: [1] [2] Through Eternity Tours. The Florence of the Medici: Power, Intrigue and Art in the Renaissance City. Available at: [2] [3] Barnebys. How the Medici Family Influenced Renaissance Florence. Available at: [3]

July 7, 2023July 7, 2023

A New AI Image Generation Enthusiast | Exploring Ice Hockey Image Variations with Midjourney v5.2 and Stable Diffusion XL 0.9

Currently, I am early in my journey with Artificial Intelligence. Most of my experience to date is through image content creation using Midjourney. Midjourney is one of the popular text-to-image models. The others are Dall-E 2 via OpenAI and Stable Diffusion by StabilityAI, and Adobe Firefly (based on a dataset of Adobe Stock). My experience with Midjourney and Stable Diffusion has been very interesting. The current versions of the applications that I have used is Midjourney v5.2 and now the recently released Stable Diffusion XL 0.9 (accessed through DreamStudio).

I am getting more acclimated and have been producing better images. There is a learning curve, primarily with using parameters as well as prompt structure to get the best results. However, there are still some shortcomings. Each of the image models do an overall great job on producing images based on my prompts. There are differences in the results pending the model I had used. Taking my interest in sports and combining it with technology, I have focused on hockey as the primary subject matter. I have also created prompts in a more abstract creation based on seeing animals in my morning runs on multiple days.

Let me state that I got fantastic images right from the start even though my prompts were not the best. This is truly a very powerful median to use and I have learned a great deal growing through this process. However, there are places where both language models struggled. By and large, the models struggled with the players hockey sticks in either showing them or even omitting them entirely. It took me some time to massage my prompt to mostly get the desired result.

At first glance, the picture above looks great and it does in an overall view which is a great example to use. The devil is in the details with this image. Looking closer, the closest player’s stick is not in the players hands combined with the players in the background missing sticks entirely. Other difficulties that I had come across deals with the number of players placed on the ice is some images. There were images where there were players with 15 players and more than two teams on the ice despite the prompt giving parameters for two teams and a limit of the number of players in my prompt. This is where trial and error came to play, all I had to do is find some more specific wording (prompts) and placed in a specific structure along with learning Midjourney’s parameters command fixed this issue to a good extent. There were some images that had issues but they were far fewer in number. Another difficulty I ran into was the uniforms and numbers. If there were numbers on the jersey in two places, they did not match or the number on the helmet did not match the number of the jersey. For these, just had to use the trial-and-error method to massage the prompt as explained before. The biggest issue was that the jersey’s did match for players on the same team in my original prompts. This is one where I had to add extra information to the original prompt specifically stating that the jersey’s for the same team must be the same. If I would have known this, the process to get to good results would have been much more efficient.

This picture was the result of correcting the prompts that I mentioned in the previous paragraph. For this picture, I would have to adjust the prompt to be more specific so the same color helmet would be used for all players on the same team. One thing that I learned is that just because I think of something as being logical and specific is not what works best for the language model to use to provide the desired results. In this way, it is still very much a “garbage in, garbage out” process. This means the resulting image is only as good as what you use. It takes time but anyone can improve. Now, I also have the advantage of my wife being a Visual Information Specialist (Graphic Design) and run the images by her. So far not to shabby for this history major and Government Contract researcher.

Up to this point I have concentrated on Midjourney in this article/post. I have mentioned that I have used other image creation applications. The following picture is from DreamStudio by Stability which uses the latest version of Stable Diffusion. This version is Stable Diffusion XL v. 0.9.

I used the same prompt to come up with this image as I have written about the Midjourney prompts. At first glance, this player seems almost perfect. Yes, it is in a different style with it being more of a computer-generated type over a picture type of image. This can be changed with editing the prompt or changing the settings in the application. The big error in this image also revolves around the hockey stick. Where the stick meets the players legs you will see that the stick shaft is bent almost in a serpentine type of shape. This is a big error for hockey players where their stick shafts only come in straight. This also generates ideas that lead one to better prompts. Before seeing this result, I would not have considered to put instructions in the prompt related to stick shape and instructions on the length of the stick for the language to consider. After seeing this image, I edited my prompt to receive better results. The point here, is that there is a great learning experience that can be applied for anyone to learn and adjust their skills.

The different text-to-image languages offer similar results, with their own spin on things, to a close enough extent for my use case. Each will find that what fits them best. The point of this article/post is to show what my novice experience has been particularly with my interest in sports and technology. I have had a blast learning this new technology using skills I have developed through my career and expanding in the world of both prompt engineering and image creation from ChatGPT to the text-to-image language models previously discussed. Next Steps for me is looking into how to utilize other forms artificial intelligence applications to increase my marketable skillset.

Now I did mention ChatGPT in the last paragraph which was the first mention of this technology in this article/post. As I have said in my other articles and post that I have written, I am a history major and take the term cite your sources seriously. With new technology, specifically Artificial Intelligence, the line between human generated content and machine generated content has blurred significantly. Thus, I like to point out each content type in my writings. Everything above this point has been human generated (written by me) and everything below was provided from a prompt by ChatGPT.

Per ChatGPT: When using the Midjourney v5.2 and Stable Diffusion XL v0.9 text-to-image artificial intelligence learning models to create images related to ice hockey, there are several aspects to consider.

What Midjourney v5.2 did well:

Improved image detail: Midjourney v5.2 showed significant progress in generating images with enhanced detail compared to its previous versions (v3 and v4). It added more realistic skin textures, facial features, cinematic lighting, reflections, glares, shadows, and expressive angles or overviews of a scene.

Realistic hands: One notable improvement in Midjourney v5.2 is its ability to generate realistic human hands with five fingers, which was a challenge for previous AI art generators. This feature adds credibility and realism to the generated ice hockey-related images.

Issues with Midjourney v5.2:

Overwhelming perfection: Some AI art fans have expressed that the images generated by Midjourney v5.2 can be “too perfect” and photorealistic to the point where they lack a sense of novelty or surprise. This precision and lack of imperfections in the results can reduce the excitement and thrill of using the AI model to generate images [[1]].

Regarding Stable Diffusion XL v0.9, the provided information doesn’t specifically mention its performance in the context of ice hockey images. However, it is worth noting the general differences between Midjourney and Stable Diffusion for better understanding:

Differences between Midjourney and Stable Diffusion:

Machine learning model: Midjourney utilizes a proprietary machine learning model, while Stable Diffusion has its source code available as open-source.

Accessibility: Stable Diffusion can be downloaded and run on your own computer, given that it meets the requirements, whereas Midjourney can only be used through the Discord chat app and requires an internet connection.

Cost: Midjourney requires a minimum monthly cost of $10 for limited image generations, while Stable Diffusion can be used for free by running it on your own hardware or by paying a nominal fee for online services.

Customization options: Stable Diffusion offers thousands of downloadable custom models, providing more options for users, whereas Midjourney has a limited selection of models to choose from.

Ease of use vs. advanced features: Midjourney may seem easier to use due to its simpler settings, while Stable Diffusion offers more advanced features and customization options, making it suitable for both beginners and experienced users.

Since the information provided does not specifically highlight the issues or strengths of Stable Diffusion XL v0.9 in generating ice hockey-related images, it’s not possible to provide a comprehensive evaluation of its performance in this context.

Overall, Midjourney v5.2 excelled in improving image detail and generating realistic human hands for ice hockey-related images. However, it also faced criticism for producing images that were perceived as “too perfect” and lacking the element of surprise. Meanwhile, Stable Diffusion XL v0.9 differs from Midjourney in terms of its machine learning model, accessibility, cost, customization options, and the balance between ease of use and advanced features, but its specific performance in generating ice hockey-related images is not mentioned in the provided information.

*A human written addendum to the ChatGPT information. After several conversations back and forth between I and ChatGPT, the conclusion was that the current knowledge base did not include the differences that I have written and have not posted to the internet yet about ice hockey images. And yes, I do use a Chrome extension that attaches ChatGPT to the Internet. Thus, the best answer was to provide ChatGPT’s initial response that was based on the basic differences between Midjourney and Stable Diffusion versions I have written about. I guess that I am ahead of the curve.

The information provided in this post was obtained through open research and personal experience using various websites. The information must be accessible to all in order to appear in this information piece.

December 11, 2019December 11, 2019

My Contract Opportunities (beta.sam.gov) Review

The release of the Contract Opportunities search function at www.beta.sam.gov occurred on November 12, 2019 to replace the prior portal at www.fbo.gov (FBO) in a step to combine multiple Federal Government Procurement websites into a singular point of entry. While I am still getting used to the new website and acclimated to the new functions as compared to the FBO website that I have come to the conclusion that it is not quite ready for prime time. The new website appears to have gone to a filtering capability and away from an actual search platform. I am not going into any of the decision-making process for this change since I am not privy to this information but coming to this post with an end user mindset. I have many years of search experience using various open source Federal Government procurement websites, such as fbo.gov and FPDS-NG.gov, as well as a proprietary online search and Business Development website. Overall, the redesign has made accessing the expected search results and data more difficult which is the opposite result that should occur when redesigning a data driven website. I do hope that this is not the intended end result but just a temporary first step towards something greater and more useful.

The change to the filter model has made it more difficult to hone into a specific result with keyword filters (searches). The “and” filter (search) seems to have been completely removed/ignored in the Keyword and Title search fields (more on another dislike on the Title field later). When performing a keyword filter, the relevant opportunities will appear at the top of the filter result list in which the user can sort in ascending or descending order by either Title or Last Updated Date Sort By categories. However, searching on multiple keywords to find both of them in the same opportunity will result in many more opportunities that do not contain both, such as performing a search on “protective security officer” AND “south Texas area”. This search example should result in multiple results with those two specific keyword strings. But the end result is the inclusion of other similar opportunities, which may have one or the other keyword but not necessarily both, thus burying the opportunities containing the helpful and desired information. While the initial result in the results list is for Protective Security Officer – South Texas Area, the following opportunities revolve around South Pacific, Nevada, and South Carolina which are irrelevant opportunities when looking at the search string. The total number of opportunities in the filter is 23,212 at the time of this writing. This is an important piece of information since when I remove the checkmark next to Active Only function, in order to tie the current opportunity with any historically released information, the results are more difficult to identify with the greater amount of posted results. This search issue becomes more apparent when searching on keywords that are vague in nature. Another example of this behavior is to search on the keyword string “mega matoc” and sacramento (with the Active Only filter unchecked). This search will result in 8,378 opportunities at the time of this writing. One can search on a single keyword phrase in the Title field, such as “Mega Matoc”, and a singular opportunity will result. However, the usefulness of this is limited. My reasoning is that this is usually a second search of mine and I am looking for other related historical postings in order to find any information that may lead me to identify the original and current opportunity’s incumbent. While technically, the search did work by bringing the one result, it isn’t really useful due to the fact the opportunity names can and do change. Bringing a limited result using a keyword search but greater than a singular result from the title search, having a “and” search in the keyword field would help in showing results to identify specific criteria such as a solicitation number and then use in other databases to find other information such as incumbent information. Notice in the following screen shot. that the keywords that were entered as an and search string is changed to remove the “and” filter and is now simply an “or” filter.

Contract Opportunities at beta.sam.gov has gone further to make accessing the information harder by two methods in my month of use. First, Contract Opportunities eliminated the ability to display additional results. In the www.fbo.gov, the user was able to choose from a range of results in terms of numbers, with the largest amount was 100 opportunities. In the Contracting Opportunities there is no such function so you are limited to 10 opportunities at a time. Again, performing results with vague information (which may be the only info you know) is much more laborious. Taking the 23,212 total filtered results, one has to constantly scroll to the bottom of the website page to click on the next page or next arrow in order to get to the next 10. Secondly, the Clear All button is located at the bottom of the filter fields. Anytime, the user enters criteria in multiple criteria of the filter fields, the most efficient use case will be to scroll the bottom of the page in order to Clear All criteria. The other option would be to click on the x that appears each individual criteria field to remove the filter criteria. This allows for additional and unnecessary steps to access data particularly combined with the First method listed in this section.

There is a neat feature that allows users to change the filter fields, but again, it isn’t quite ready for prime time. Changing the Filters is a good idea particularly when one does not want to use a particular field, such as the Inactive Date field, but its usefulness is limited to the same filter session of your initial inquiry. The original (default) search layout will reappear after clicking on the opportunity to view its details forcing the user to readjust the filter options in subsequent filters. Changing the filter options and then saving your filter through the Save Search function doesn’t save the filter preferences as well. Thus, to me, makes the Change Filter utterly useless.

I also ran into an issue when you resize the window when wanting to multi-task. When the text size has been increased from the default size, when one resizes the browser window, to half the screen, the viewable information will be cut off. For example, making the window half the screen will cause the website footer to cover the opportunity information. This makes it very difficult to read and to utilize a second window when the text size has been increased. Please note that this activity doesn’t occur if you leave the text at the default size and may depend on the browser settings.

Out of the three main browsers (Microsoft Edge, Google Chrome, Mozilla Firefox) it seems to be best to use the Contract Opportunities website with either Microsoft Edge or Google Chrome. When downloading documents, these browsers correctly recognize the file extension. This means, the browser will download the document and recognize the program, for example, file ending with .docx will be saved or opened with Microsoft Word. Mozilla Firefox will not recognize the file but the source to where the file was posted. For example, the location https://iae-fbo-attachments.s3.amazonaws.com appears in the dialog box after clicking on the download link. The file extension isn’t recognized. I only am including this since it is what I experienced with the Firefox browser and knowing that it could be a local setting or that I don’t have a Firefox extension that allows the browser to recognize the correct file through this download function. Your experience may be different if you are using Firefox.

The Contract Opportunities website has not remained static since its launch. I have noticed an early gripe of mine has been changed. At first, when unchecking the box next to Active Only, the search criteria would disappear when I removed the said checkmark. This was irritating since I had to re-enter the search criteria in the fields to duplicate the prior search while hoping I did not forget the exact criteria. At some point, after the launch of the website to the time of this writing, this behavior has been changed/corrected. Now the search criteria that I have entered remains after removing the check mark in subsequent search. The behavior is also the same when you add the check mark into the Active Only field as well. Another function that I noticed that has changed/fixed is the disappearing Title Field after adding it through the Change Filter Function. At the start, this field would disappear after either clicking on an opportunity title or conducting a different filter/search. As of this writing, I was able to view the Title filter field after moving to read a opportunity or do a second search, however, the Inactive Date filter fields remain an issue as the filter fields do not remain removed through the Change Filter function as previously described in this post.

Of my usage for the past month, the clear best possible search method is to enter a Solicitation number (70RFP419RE7000002) within the Keyword field. This will bring the desired results but the downside is that the user must first know the solicitation number to perform this type of search. The prior filter processes listed in the article are attempts to discover the solicitation number and at this point in time it is difficult to get to this information due to the current search/filter function. With access to data being the most important part of any data driven website, in my opinion, the shortcomings listed in the article should have been included. I do say this not knowing what the road map is for the Contract Opportunities search/filter function or if any of these functions will appear shortly or in the near future if at all.

The positives that I have experienced with the move to Contract Opportunities is that the website feels quicker than FBO.gov. I also like the general idea of going to one point of access website for Federal Government information. I just fear that there are too many concessions to get to this one access point that it makes the actual data identification process needlessly harder. It has taken some time to adjust to the new visual layout when compared to the fbo.gov website but I have come to prefer the clean look of the Contract Opportunities website.

This posting concentrates on the aspects of the Contract Opportunities website within the first month of use that I have not liked. I focused on this aspect since each use case and user preferences are different within the Federal Government Business Development community. I wanted to bring some of my insights in order to assist other users with their use of the new website. I do hope that there will be changes soon to address the listed issues that I have described. This website shows a lot of promise in being superior to the prior fbo.gov website but falls short in the execution to this point. This posting revolves my experiences and may not mean that the reader experiences the same issues. There is a provide feedback button to report any issue that you may experience.