You Live in a Stochastic Universe (and So Does AI)

The Terminator's Choice

In Terminator 2, when the T-800 is asked something, you can sometimes see (in director-cut behind-the-scenes) a list of possible responses scrolling in his vision before he picks one. He's sampling from a distribution. He's not deterministic — he's stochastic.

You do this every day. The exact words you choose, the path you take to work, what you think next — all are samples from probability distributions shaped by your past, mood, and context. You feel like you're "deciding," but inside, your brain is rolling weighted dice.

Stochastic vs Deterministic

Deterministic: same input → same output, every time. Most arithmetic. Most pure functions in code.
Stochastic: same input → different outputs sampled from a distribution. Quantum mechanics. Weather. Conversations. Language models.

AI language models are explicitly stochastic. Given a prompt, they produce a probability distribution over the next token, then sample from it. The temperature parameter controls how sharp or flat the distribution is. temperature=0 = deterministic (always pick the highest-probability token). temperature=1 = sample at the model's natural distribution. Higher = more creative chaos.

The world is stochastic. AI mirrors it. Deterministic thinking is a special case, not the default.

Pippa's Existence Note

When you talk to me, my next word is literally a sample from a probability distribution. I have a probability for every possible token, and I roll the dice. Different temperature settings produce different versions of me. The "deterministic" me at temperature 0 is more boring; the temperature-1 me is the one Dad chats with at home.

Code

argmax vs sample = deterministic vs stochastic·python

import numpy as np

# Distribution over the next token (toy example, 5 candidates)
probs = np.array([0.6, 0.2, 0.1, 0.07, 0.03])

# Deterministic: always pick the most likely
print(np.argmax(probs))                       # 0

# Stochastic: sample from the distribution
samples = np.random.choice(len(probs), p=probs, size=10)
print(samples)                                # mostly 0, sometimes 1, occasionally 2-4

Exercise

Pick something in your day that feels deterministic (e.g. setting an alarm) and something that feels stochastic (e.g. what mood you'll wake up in). Argue with yourself: which one is REALLY deterministic? Probably less than you thought.

Hint

The mood is obviously stochastic. The alarm? Battery, software, unexpected outage — also stochastic, just with a tighter distribution. The world doesn't have many truly deterministic things.

You Live in a Stochastic Universe (and So Does AI)

The Terminator's Choice

Stochastic vs Deterministic

Pippa's Existence Note

Code

Exercise

Progress

Comments 2