1. Overview

Georgia Tech, a prominent research university in Atlanta, Georgia, manages a complex classroom infrastructure to support its academic mission. Efficient space utilization is critical amid growing enrollment and resource demands in higher education. This study quantifies classroom utilization at the Atlanta campus for the 2024 calendar year, leveraging scheduling data to evaluate efficiency. Preliminary findings indicate an average of 20.69 hours per week during major terms (Fall and Spring) with an overall annual utilization of 16.32 hours per week.

2. Methodology

2.1 Data Source

Georgia Tech’s Oscar system (https://oscar.gatech.edu/pls/bprod/bwckschd.p_disp_dyn_sched) provides public access to class schedules, including course name, location, meeting times, and enrollment, from present to 2006. This study uses 2024 data (Spring, Summer, Fall terms).

2.2 Data Collection

A Python-based web scraper was written and then used to extract scheduling data from Oscar for all 2024 terms. Data included room location, meeting times, and enrollment (seats taken and remaining).

2.3 Analysis

Classroom utilization was calculated as total hours per week per room, based on meeting times. Only the Atlanta campus was considered, excluding remote sites (e.g., Shanghai, Emory University) and unassigned locations (e.g., “TBA”, “See academic dept”). Cross-listed classes are correctly handled by this analysis and are not double counted. 

2.4 Assumptions

Classes with zero enrollment were excluded, assuming cancellation, while those with ≥1 student were included due to variable cancellation thresholds. Meeting times listed in Oscar, spanning instruction and finals (16 weeks for Spring/Fall, 12 for Summer), were fully counted, potentially overstating utilization during finals week by assuming one-day exams occupy the full period. This conservative bias ensures utilization is not underestimated. 

3. Limitations

This study’s accuracy depends on Oscar’s data integrity, which may not reflect undocumented cancellations or unlisted room uses (e.g., special events, faculty-only sessions). Canceled classes remaining in Oscar could inflate utilization, while untracked uses (e.g., seminars not open to students) may understate it. Room availability assumes non-zero 2024 usage, but renovations or renaming could skew counts without campus facilities data.

4. Results

In 2024, 324 rooms at Georgia Tech’s Atlanta campus were analyzed. Table 1 summarizes total and average weekly utilization across terms.

Table 1

4.1 Utilization Distributions

4.2 Specific Rooms

Table 2 lists the top 10 most utilized rooms in Fall 2024, with capacities based on the largest class enrollment recorded.

Table 2

Van Leer C353 is an electronics laboratory and therefore is not the most representative of lecture type classroom utilization. 

East Architecture 207 hosted 49 credit hours of classes as well as several recitations. Classes started in this room at 8:00AM and ended at 7:45PM with a gap of at least 15 minutes between sessions. Even though this classroom is located in the East Architecture building the range of classes hosted is diverse, including Pre-Law Seminar (PUBP 3610), Intro to Structural Eng (CEE 3051), and Multivariate Data Analy (ISYE 7405). 

5. Conclusions

While room utilization varies greatly an average utilization of around 16.32 hours per week is low. The unique requirements of a university cannot justify this low utilization since some rooms demonstrate occupancy twice the mean average. If the summer academic term is excluded utilization becomes 20.69, which is still low compared to what is known to be possible. This represents an opportunity for significant improvement at the institution. 

6. Data

A sql lite database containing all of the data used in this analysis can be found here:

https://drive.google.com/file/d/1n3KqDT6zMM43br2d2nJYfsZRyWBxKzOU/view?usp=sharing

Individuals wishing to reproduce these results may also go directly to Oscar here:

https://oscar.gatech.edu/pls/bprod/bwckschd.p_disp_dyn_sched

Warhammer 40k team tournaments involve 8 players matching up in individual games against the 8 players of the opposing team. This requires a mechanism to determine who will play against whom. Various mechanisms have been used such as dice or paper slips, but after seeing John Wick 4 (where metal plates are used to determine the rules for a duel) I thought the use of metal plates to decide the order of battle would be more awesome. 

There are 8 plates in this set. All plates have the same front. Each plate has a faction and player name on the back.

The plates are 5″ tall, 3″ wide, 0.25″ thick and are composed of 304 stainless steel with each weighing almost 1 pound (around 450g).

No dye, ink, or paint is present on these plates. The decorations are achieved by using a pulse laser to selectively change the oxidation state of the contained chromium to generate the visible colors. Because of this the surface of the plates are smooth to the touch and the color can only be removed by something capable of removing the outer surface of steel from the plate. The color palate is rather limited since there are only a few states available, but copper, gold, black, and blue are available. 

Other than the team and faction logos I designed the art for these myself. At first I was going to copy the plates used in John Wick but those are very art deco and the grimdark of 40k called for a more gothic aesthetic. 

This laser technique is well documented but not often used since it is exceptionally sensitive and far slower than other processes. A difference of only 100mw in output power can completely change the resulting color. Additionally, any surface not within the exact focal plane of the laser, due to even the slightest imperfections in the surface, will ruin the color. 

The overall process for these plates was:

Milling – The plates as delivered from the supplier are not flat and must be machined into condition. I did this with my Tormach 1100 and a fly cutter. 

Grinding – Starting at 220 and then moving to 400 grit using my Ameribrade grinder, this process was the most miserable. Grinding the surface of a 3″ wide plate with a 2″ wide belt is not particularly pleasant. 

Polishing – Using a buffing wheel and starting with a Maverick Abrasives black and finishing with green puts a mirror finish on the plates. 

Laser – It took about 1.5 hours to process each plate. Since the lens I wanted to use for this operation only has a 4″ working area and I have a 5″ plate I need to split the operation into 2 steps for each side. This was factored into the design. Clustered finials in the middle of the design split the art into separate upper and lower sections. The included gap around the finials allows for some small error in alignment after repositioning the plates without making the error obvious. 

Overall I think it turned out fairly well. If I could find a supplier for pre-polished 304 plate (not sheet, but plate at 0.25″) it would make this a much easier workflow. 

Another DragonCon complete and another 4 panels delivered. You can find the slide decks for my 3 panels here. For the escape room panel I co-hosted you can find a link to that deck on Wade’s LinkedIn profile here.

The madness

On 17 November 2023 Sam Altman was fired from his position as CEO of OpenAI. Documents indicate the board told the leadership team that allowing the company to be destroyed ‘would be consistent with the mission’ [of the company].

This rather scant set of facts has led some to believe OpenAI has created a superintelligent AI called Q*. Altman was fired for recklessly creating a danger to all of humanity, and possibly desires for world domination, or for being the puppet of Q*, or some other James Bond villain behavior. The purpose of this post is to explain why I believe this to be false.

I believe Sam Altman was fired because the board of OpenAI realized, too late, its company had been reshaped from a small non-profit research group into a Silicon Valley startup where their services would no longer be needed.

Altman has been skillfully moving towards this end for years. First creating the for-profit entity to offer stock options to attract top talent, and then bringing Microsoft on as an investor. OpenAI now resembles and operates like a dot com startup. Its non-profit origins are a vestigial organ no longer of use to the host. This outcome should not be surprising. Altman has no experience with anything but this type of company. He has behaved in a manner consistent with past performance.

Saftey, Secutiry, and Super antigens

So why the cryptic statements about allowing the company to be destroyed? Because safety, and security, are super antigens in the business world.

Antigens are the molecular signatures the immune system uses to identify pathogens. They are the first step in marshaling a defense against the invader. Antigens are important, so important some bacteria exploit them by producing super antigens. These super antigens cause an overreaction so intense the now hyper zealous immune system will destroy its own cells for fear they might be the pathogen. In this chaos the bacteria thrive.

Safety is important, so important it is perfect for exploitation. Security is similarly invoked. I have sat through many painful meetings where IT explains an especially ridiculous policy is needed for security. They cannot explain what is being made more secure or what threat is being addressed, but unless you want to be responsible for the business ending you need to uphold the policy.

Firing your CEO for safety concerns sounds reasonable, much better than admitting you don’t like the company he has created, or the fact you will soon be obsolete.

What about all the experts who believe in Q*?

AI safety proponents are like crypto proponents. A few are genuinely interested in ensuring their technology is beneficial, and the rest are scammers. AI doomers get engagement and AI safety is a somewhat more elegant way to cash in on that engagement.

Imagine if I were to tell you I was a faster than light starship safety proponent and I had created a unilateral phase detractor to make FTL travel safer. Would you be interested in hearing what I have to say, or better yet employing me at your company? Probably not, since FTL does not exist I could not have tested my unilateral phase detractor. So you probably not be too interested in my unproven creation for a system that does not yet exist.

Unlike FTL travel AI has not had the same amount of time to be processed by our culture. We are therefore more willing to accept a more interesting fiction over mundane reality. Cities were supposed to be rebuilt because of the Segue, blockchain was going to destroy the entire financial system, and Theranos would revolutionize medicine. Progress is hard. Even revolutionary technologies take decades to deploy. But in the infancy of a new technology we are seduced into irrationality. 

The way forward

Spending energy on speculative future harms weakens us against those of the present. Learn the motives of those who have your attention. Do not give fuel to those who benefit from your fears. Do not add your voice to the chorus of the ignorant.

I gave my Advanced Fabrication Machines panel at DragonCon 2022. This was basically the same panel I gave at MomoCon 2022. For those looking for the slides you can find them as a you tube video here: