WORK IN PROGRESS

TL;DR

Computers are boxes of densely packed, microscopic light switches. They use 1’s & 0’s (called “bits”) to store & manipulate data (think of nucleic acid bases in DNA [G,A,T,C]). These machines can connect to each other & communicate at light speed$^4$. They can also be scaled to use 1 Trillion bits in their calculations$^5$. This greatly surpasses the estimated neuron count in the human brain (61billion – 99billion)$^3$.

While neurons are much more complex than light switches, new computer architecture which simulates neural activity is becoming much more prevalent in data science & logic / reasoning tasks$^6$. These leaps in technology & architecture have taken computers from simple calculation machines to predictive oracles. Add to this the innovations in linguistics and you have a device which can communicate eloquently with the facts to back up each statement.

More than just pure power, these machines are able to communicate faster & more accurately than human brains. Humans are limited to a handful of modalities: vocal, writing, body language. Each of these must translate neural activity into a different method of communication, greatly reducing the message. Computers can share exact copies of information, at lightspeed, over vast distances.

The question then changes from “how much do we trust AI?” to “how much do we trust humans?”

---

Introduction

The reason to understand transistor count & Moore’s law is not simply for factual knowledge. It is important to see how rapidly technology has advanced & what the foundational mechanism is.

In a single lifetime, we went from machines that couldn’t store 3 digit numbers to pocket sized super computers; from arithmetic calculators to differential equation solvers; from a pure math machine to a predictor of human nature.

To address the Ethics of Artificial Intelligence, we need to understand what AI really is & how it came to be.

The Evolution of Computer Use

Any sufficiently advanced technology is indistinguishable from magic

Arthur C. Clarke

TL;DR

We took a fancy calculator, made it more powerful, added in a feedback loop, plugged in some fancy statistics & wrapped the whole thing in a web of linguistics.

Origin

Alan Turing is credited with the first modern implementation of a computer (see Imitation Game)$^7$. His thought experiment was very simple:

• Consider you have a long tape divided into cells (with each cell having a symbol), you have a “tape head” which can read & write the symbols, and you have a finite set of states which determine which action to take
• Given enough time & tape, this simple system can calculate any complex function

Turing Machine Example – Simple

Turing Machine Example – Complex

💡Fun Fact: The concept of a Turing machine has played an important role in the recent philosophy of mind. The suggestion has been made that mental states just are functional states of a probabilistic automaton, in which binary inputs and outputs have been replaced by sensory inputs and motor outputs. This idea underlies the theory of mind known as “machine functionalism”$^7$.

---

We took this simple system and expanding its capabilities, making magic boxes which can calculate anything we ask it. The computer in your microwave is just as powerful as the computer used to land astronauts on the moon (America 🦅 circa 1969).

Now we plug in more and more data to solve more and more complex problems: protein folding $^8$, weather forcasting $^9$, predicting Human Nature $^{10}$, and so much more.

The Internet

While each individual machine is strong in its own right, the power is not what each median person uses it for. The regular person is most concerned with connecting to global society via these machines.

By connecting computers together across the entire globe, a network is formed which connects each individual user to every other user on the entire planet. This degree of connection was unfathomable a century ago when the best we had were phone lines which connected major towns & cities to each other, but still faced geographic constraints (see The Victorian Internet). Communication by speech was incredible in its own right, but transferring data was limited to Morse code or postal mail.

The Internet changed life as we knew it: suddenly photos could be shared, text and speech could occur around the globe, games could be played with other people, news from Africa could reach Alaska within seconds, ideas could be shared on bulletin boards like Reddit for any human on the planet to read, critique, and contribute. Globalization was catalyzed.

And again…. this all occurred in a single lifetime.

Technocrats

Unlike phone calls which are ephemeral by nature, this new world order had object permanence.

The data you shared on Facebook stayed on Facebook. Your name, information, photos, posts, thoughts & beliefs, likes & dislikes, habits, behavior, attention span… this all exists and is saved on the platform.

As Facebook, Google, Myspace, and Twitter scaled to new heights, they needed to generate income to keep up demand. Storage space isn’t free and they had to keep building servers to account for the added users. Their solution was selling adspace. But typical adspace is random, generic, at the whim of the user (think of all the random billboards that you couldn’t care less about).

Because these platforms had so much data on each individual user, and because computers were so computationally capable, they designed algorithms that could predict which ads were most relevant to each user and sold that space to the highest bidder:

• If you were a 20 year old gym-bro, you’d get supplement ads
• if you were a 15 year old girl, you’d get makeup ads
• if you were a 35 year old parent of two, you’d get diaper & lotion ads

💡Watch The Social Dilemma for the impact this has had on society & the next generation
💡Also see Facebook / Cambridge Analytica Scandal for a glimpse of the effect this has on government & society$^{11}$. Facebook had “5,000” data points on each of its users.

---

While their motivations were capitalistic, these Tech Juggernauts were the driving force behind technical innovation. The mathematics used for particle physics in NASA were being applied for behavioral analysis. Suddenly Human Nature wasn’t a black box which only priests & clairvoyants could comprehend, it was a system of psychological states which could be defined & manipulated.

As these platforms became smarter, the way they interfaced with us became smarter too. Google no longer gave just the most popular results, it gave the most accredited results & the most relevant results for you specifically. Facebook didn’t just let you find people, it started suggesting people you may know at incredible accuracy. No longer were you searching content, now the content finds you.

Based on your viewing activity platforms like TikTok & Instagram created profiles on their users (see The Social Dilemma).

We (as a whole) typed so many thoughts, ideas, & questions into these platforms that they began to see patterns in language & thought processes. Linguists were able to see structure in each language (English uses ‘subject-verb-object’, whereas Gujarati may use ‘subject-object-verb’). They started to understand context in a broad sense (“Let’s eat, grandpa” is verrrrry different from “Let’s eat grandpa”) and figured out how to implement this in machines. Machines could analyze text and derive the general meaning.

All this came to a head with Google’s breakthrough paper on conversation technology$^{12}$. Users could now write a statement and the machine would derive the intent from the message. The bridge between human language & machine language was shrinking fast.

The State of Chatbots – CRESCENDO

The Google paper on conversation technology is the catalyst of modern “AI”. Suddenly machines can talk back to us with seemingly clear, concise language relevant to our questions & context. No longer are you searching twenty different queries to help plan your trip to Aspen, now just tell the Chatbot “Make me an itinerary for Aspen on X dates and account for the weather & travel times” and within seconds you are given a list of activities each timeboxed & adjusted according to the weather at that time and place. The Internet’s wealth of knowledge is no longer at just your fingertips, it’s a conversation away.

This new brand of interaction is under the umbrella of “generative AI”, which not only searches the web and points you in the direction to look, it digests the data for you then feeds you the data in a way that is more conversational (like a momma bird feeds its baby birds). In some cases, the “generation” goes too far and misses the mark on accuracy. In these situations it is considered to “Hallucinate” (i.e. make up inaccurate information for the sake of conversational pleasantry).

This is a crucial point of friction for the ethics of AI. If bots are going to make up information then they cannot be 100% relied upon, but given the vast knowledge base they pull from (the internet) and their statistical reasoning skills, they are critical tools which have strengths beyond that of humans.

• Examples of chatbots ChatGPT – well rounded Anthropic Claude – Shakespearean writer Grok – iPad baby Perplexity – the grad student Minstrel – Je ne sais pas Gemini – Google on steroids DeepSeek – Chinese (See Cathedral & Bazaar sections “China”, “GPU”, )

Questions

• Will AI replace lawyers?
  • Chatbots digest text & make inferences. Is this better than what humans can do?
    • How do we define “better”? Volume of content? Accuracy? Objective analysis can leave out the human story
      • “22yo male, brown, charged with theft, assault, evading arrest. “
      • “Young man was shoplifting food and ran when the cops came, bumping into people along the way”