Dan's tech blog - this and that

Sunday, December 7, 2025

Programming the Impossible:

Why the StarTrek Tanuga IV High Orbit Research Station Holodeck-Simulation Required 392.76 TB - Assuming Terabyte Scale Optimization

We've all marveled at the Star Trek Holodeck – a wondrous technology capable of creating any environment with seemingly infinite detail. But what does it take to program such a marvel? And what kind of digital footprint does a truly complex Holodeck simulation leave behind?

Let's dive into a hypothetical scenario, combining canonical Star Trek timelines with real-world computing concepts, to explore the sheer amount of data compiled for a specific, complex program: the Tanuga IV research station from The Next Generation, Season 3, Episode 14.

Table of contents

1. Knowledge Storage Density Revolution

with Offline LLMs

2. Windows Recall: Security Flaws

through RAG AI Agents

3. Local LLM-Inference on Gaming-GPU —

Environmental & Privacy Benefits

4. Potential SkyNet-Level Threat: Exascale AI

Supercomputer NVIDIA GB200

5. Own witnessed incident: Bluetooth 5.2 Audio

Routing Vulnerability

The Setup: Enterprise Power and Programming Time

Our journey begins with a detail from Star Trek: The Next Generation: programming the detailed Tanuga IV High Orbit Research Station simulation — a Regula I-type facility administered by the Tanugans in the 24th century — took a considerable 18 hours and 11 minutes.

For this thought experiment, we assumed this process utilized the Enterprise's main computer operating at approximately 60 PFLOPS (Peta Floating-point Operations Per Second) for the duration.

In that 18 hours and 11 minutes, the Enterprise's computer would perform an astounding number of calculations:

The Challenge: A High Orbit, High-Fidelity Simulation

If every single one of those FLOPs equated to a byte of stored data, the compiled file would be over 3,927 Petabytes. That’s an astronomical figure. For a Holodeck program to be efficiently managed by the computer and remain within a reasonable, though massive, storage range (the terabyte scale), it must be incredibly optimized and compressed.

This specific program is even more demanding because of the environment it simulates: The Tanuga IV research station was situated in high orbit at a minimum distance of 5,000 kilometers above the planet. Accurately simulating this high-orbit environment requires synthesizing complex factors: low-gravity physics, the detailed view of the planet and stars from that specific altitude, and the precise layout and interactive systems of a Regula I-type science facility.

The "Highly Optimized" Assumption

To achieve a realistic file size that keeps the result in the terabyte scale (i.e., less than 1 Petabyte or 1000 TB), we must assume that Starfleet's technology employs a degree of extreme optimization. We'll use a 10,000:1 compression ratio, meaning that for every 10,000 floating-point operations the computer performs during the programming phase, only 1 byte of resulting data is stored in the final, compiled program. This is the definition of a "highly optimized" file.

The Result: A "Lean" Holodeck Program at 392.76 Terabytes

Under this assumption of a 10,000:1 optimization ratio, our massive 3.93 × 10²¹ FLOPs translate into a compiled Holodeck program size of approximately:

Even with monumental optimization, the fully simulated Tanuga IV High Orbit Research Station program requires nearly 400 Terabytes of the Enterprises data storage ("Isolinear Chips").

Why So Much Data?

Despite the extreme compression, 392.76 TB makes sense when you consider what must be encoded:

Specific Environmental Physics: Accurate gravity and motion modeling for the 5,000 km orbital altitude.
Regula I-Type Schematics: Detailed blueprints, interactive systems, and operational procedures for this specific class of station.
Complex AI & Scripting: The behaviors and administrative functions of the simulated Tanugan personnel.
High-Fidelity Textures & Models: Every object, window view, and instrument panel must be rendered in stunning detail.

This result confirms that while Starfleet can compress information to an almost unbelievable degree, the creation of a fully immersive, interactive simulation in the 24th century still requires a staggering dedication of computational power and storage.

source: Google Gemini 2.5 Flash

Wednesday, November 6, 2024

YouTube's Algorithmic Trump Bias

According to a Guardian investigation, YouTube's recommendation algorithm may have disproportionately promoted pro-Trump and anti-Clinton videos during the 2016 U.S. presidential election, raising concerns about the platform's role in shaping political discourse and voter perceptions.

Pro-Trump Video Surge

The Guardian's investigation revealed a stark imbalance in YouTube's video recommendations during the 2016 election, with 95% of recommended candidate speeches favoring Trump. This algorithmic bias extended beyond official campaign content, promoting conspiracy theories and unsubstantiated claims about Clinton's health and personal life. Notably, the most recommended channel belonged to Alex Jones, known for propagating extreme right-wing views. The analysis of over 8,000 videos in a previously unseen database exposed the algorithm's tendency to amplify pro-Trump narratives, potentially influencing voter perceptions and contributing to the spread of misinformation. This phenomenon underscores the significant impact of recommendation algorithms on political discourse and highlights the need for greater transparency and accountability in social media platforms' content curation mechanisms.

Algorithmic Bias in Recommendations

YouTube's recommendation algorithm exhibits a left-leaning bias, according to a controlled experiment that constructed archetypal users across the US political spectrum and analyzed over 120,000 recommended videos. The study found that the algorithm pulls users away from political extremes asymmetrically, with a stronger pull away from far-right content compared to far-left content. Furthermore, the recommendations skew left even when the user lacks a watch history.

However, other research suggests a right-wing bias in YouTube's recommendations, with a growing proportion of recommendations coming from problematic channels like "Alt-right," "Conspiracy," and "QAnon," especially for users categorized as "right" and "very-right." The probability of a far-right user receiving ideologically compatible recommendations is higher than for a far-left user, and cross-cutting recommendations are offered less often to far-right users. Additionally, the algorithm increases the odds of right-leaning users continuing to watch ideologically congenial videos, while not exhibiting the same effect for left-leaning users.

Conspiracy Theories Amplified

YouTube's recommendation algorithm has been implicated in the amplification of conspiracy theories, particularly in the political sphere. While the platform's role in radicalizing users remains debated, studies suggest that its recommendation system can create echo chambers and promote problematic content. The algorithm's tendency to recommend ideologically congenial videos, especially for right-leaning users, may contribute to the spread of conspiracy theories.

Conspiracy theories often emerge from fringe communities and are legitimized through social media channels.
YouTube comments sections serve as participatory spaces for conspiracy theory development, functioning similarly to anonymous message boards.
The platform's recommendation system has been observed to increase the visibility of conspiracy-related content, particularly for users with extreme political beliefs.
Discursive strategies used by conspiracy theorists in comments, such as "connecting the dots," often evade automated detection systems. These findings highlight the complex interplay between algorithmic recommendations, user behavior, and the propagation of conspiracy theories on YouTube, emphasizing the need for more nuanced approaches to content moderation and platform design.

source: free Perplexity GPT-3.5 query using the "Page"-feature @ https://www.perplexity.ai/page/youtube-s-algorithmic-trump-bi-qDtRTqwJQfSwnHwEDpR9Bw

Friday, September 27, 2024

Sensing COVID Through Odor

Recent studies suggest that some individuals may possess heightened sensitivity to the presence of COVID-19 patients, potentially detecting unique odors or respiratory changes associated with the infection. While research has demonstrated the effectiveness of trained dogs in identifying COVID-19 through scent, the mechanisms behind human sensitivity to these cues are still being explored and remain less understood.

Breathing Difficulties as Indicators

Respiratory symptoms associated with COVID-19, such as coughing, shortness of breath, and altered breathing patterns, may be noticeable to some individuals in close proximity to infected persons. These changes in breathing can be particularly pronounced in severe cases, potentially alerting highly observant people to the presence of an infected individual. However, it's important to note that breathing difficulties are not exclusive to COVID-19 and can be caused by various other conditions. Additionally, many COVID-19 cases, especially in the early stages or among asymptomatic carriers, may not present with obvious respiratory symptoms, making this method of detection unreliable for general use.

Scent-Based Identification of COVID-19

Volatile organic compounds (VOCs) produced during SARS-CoV-2 infection may create unique odor profiles detectable by some individuals.
Research has primarily focused on trained dogs' ability to identify these scents, with studies showing canine detection accuracy exceeding 99% for COVID-19 patients.
The potential for human olfactory detection remains less explored.
Some highly sensitive individuals might perceive subtle changes in body odor or respiratory emissions from infected persons.
This ability is not well-documented or understood in scientific literature.
The presence of these VOCs could theoretically serve as an olfactory cue for those with heightened sensory perception.
Further research is needed to confirm and characterize this phenomenon.

Sensory Changes in COVID-19

COVID-19 infection can significantly alter sensory perception, potentially making some individuals more sensitive to the presence of infected persons. A study published in Nature found that 20% of human sensory neurons were susceptible to SARS-CoV-2 infection, with changes in sensitivity and hedonic rating to transcutaneous electrical nerve stimulation observed in COVID-19 survivors. These sensory changes included:

Reduced threshold of sensory sensitivity
More unpleasant perception of electrical stimuli
Altered subjective sensations from tingling to pinching or other unpleasant feelings

The effects were most pronounced in individuals who had recently experienced COVID-19 symptoms, with hypersensitivity diminishing over time. These findings suggest that some people may indeed experience heightened sensitivity to sensory cues associated with COVID-19 infection in others, though the exact mechanisms and prevalence of this phenomenon require further investigation.

Comparison with Canine Detection

While some humans may exhibit heightened sensitivity to COVID-19 patients, trained dogs have demonstrated remarkable accuracy in detecting the virus.

Canines can identify SARS-CoV-2 infection with over 99% accuracy, both in vitro and in vivo.
Dogs can detect the virus in human respiratory secretions and by directly scenting patients.
Results are available in less than 5 seconds per subject.
This method is highly effective across various settings, including hospitals and low-risk environments.
It is not affected by COVID-19 severity or symptom presentation.
Dogs' superior olfactory capabilities make them more reliable and well-documented compared to human sensory perception, which remains largely anecdotal and unverified in scientific literature.

The superior olfactory capabilities of dogs make them a more reliable and well-documented means of detecting COVID-19 compared to human sensory perception, which remains largely anecdotal and unverified in scientific literature.

source: free Perplexity GPT-3.5 query using the "Page"-feature @ https://www.perplexity.ai/page/sensing-covid-through-odor-ijO19XDaRrGnqUIk3YGUPw