All posts on AI

Diving into the final one, expertise becomes cheap. What will that mean? Computing (Cheap Information Processing & Logic) vs. AI (Cheap Expertise & Decision-Making)

Computing (Revolution 7) and AI (Revolution 8) seem similar because both involve automation and data processing, but they differ in what they make cheap and what types of human labor they replace.

Revolution 7: Computing → Cheap Information Processing & Cheap Logic**

🔹 Key Idea: Computers automate calculations, data storage, and repetitive logical processes. 🔹 What Became Cheap?

Mathematical calculations (spreadsheets replacing accountants).
Data storage & retrieval (databases replacing file cabinets).
Repetitive logic-based tasks (software replacing clerks). 🔹 What It Replaced?
Bookkeepers, clerks, human “computers” (pre-electronic era math workers).
Manual data entry and processing jobs.
Logic-based human decision-making (e.g., tax software vs. accountants). 🔹 Limitations:
Computers execute strict rules—if a situation isn’t pre-programmed, they fail.
They don’t “understand” problems—only process logic as instructed.
They require explicit programming to function (if X, then Y).

Examples of Revolution 7 (Computing in Action)

✅ Excel & Spreadsheets → Automated business calculations. ✅ Databases & Search Engines → Made finding and storing information instant. ✅ Enterprise Software (ERP, CRM, etc.) → Automated workflows in businesses. ✅ Robotic Process Automation (RPA) → Automates repetitive digital tasks.

Revolution 8: AI → Cheap Expertise & Cheap Decision-Making

🔹 Key Idea: AI automates knowledge work and decision-making, reducing reliance on experts. 🔹 What Became Cheap?

Pattern recognition & prediction (AI detects fraud, diseases, and legal risks).
Creative & analytical work (AI writes code, generates content, and composes music).
Decision-making under uncertainty (AI assists in law, finance, hiring, and medicine). 🔹 What It Replaced?
Lawyers, doctors, financial analysts (partially).
Writers, artists, and software developers (for certain tasks).
Customer support agents (chatbots replacing tier-1 help desks). 🔹 Key Difference from Computing:
AI isn’t explicitly programmed with rules—it learns from data.
AI generalizes beyond its training data (unlike traditional computing).
AI can operate without human supervision in some cases (self-driving cars, trading bots).

Examples of Revolution 8 (AI in Action)

✅ GPT & Large Language Models → Replacing copywriters, summarizing complex texts. ✅ DALL·E & Midjourney → Generating art and replacing some graphic designers. ✅ AI-Powered Medical Diagnosis → Radiology AI reading scans as well as human doctors. ✅ Autonomous Systems (FSD, AI Traders, AI Lawyers) → Reducing human expertise dependency.

Category	Computing (Cheap Logic & Processing)	AI (Cheap Expertise & Decision-Making)
Key Advantage	Automates structured, rule-based logic	Automates expertise & decision-making
Requires Explicit Rules?	✅ Yes	❌ No (learns from data)
Handles Uncertainty?	❌ No	✅ Yes (probabilistic reasoning)
Data Type	Structured (numbers, databases)	Unstructured (text, images, voice)
Replaces	Clerks, bookkeepers, programmers (to an extent)	Lawyers, doctors, writers, artists
Examples	Excel, SQL, ERP, Payroll Software	ChatGPT, Midjourney, AlphaFold, Self-Driving Cars

How AI Goes Beyond Traditional Computing

1. From Fixed Logic to Adaptive Learning

Computing (7) → If you want a system to identify spam emails, you write rules like: “If email contains ‘Nigerian prince,’ mark as spam.”
AI (8) → AI learns by analyzing millions of spam emails, finding hidden patterns humans might miss.

2. From Simple Automation to Complex Decision-Making

Computing (7) → An e-commerce site’s software follows fixed rules: “If order > $50, apply free shipping.”
AI (8) → AI optimizes shipping costs dynamically, predicts demand, and personalizes discounts for individual customers.

3. From Human Input to Autonomous Systems

Computing (7) → A self-driving car must follow predefined rules (“Stop at red light, Go at green”).
AI (8) → AI interprets the environment using neural networks, making real-time driving decisions.

If Computing made logic cheap and AI made expertise cheap, what’s next?

AI + Robotics → Cheap Skilled Physical Labor
- AI-powered robots replacing electricians, plumbers, chefs, surgeons.
- AI-driven factories with zero human workers.
AI + Biology → Cheap Health & Longevity
- AI designing new drugs & personalized treatments.
- AI managing healthcare, eliminating most routine doctor visits.
AI + Generalized Automation → Cheap Everything?
- Fully automated businesses (AI CEO, AI employees, AI customer service).
- AI predicting and managing global supply chains.

🚀 The jump from Computing to AI was about moving from automation of logic to automation of intelligence. 🚀 The next leap might be from automation of intelligence to automation of reality.

Would you like to focus on a particular area of Revolution 8 or explore Revolution 9 in more depth? 🚀

Examples of Alchamey

Doing worst job in dec, (and July for European models) - crappy source
Doing a better job if you say you are stressed
Telling the LLM not to hallucinate (fasicnating, this only works now, not in early models as early training data didn’t talk about hallucinations)

How do LLMs think?

“Complete the next word” is super hard. Imagine a big long mystery novel, and then the next sentence is - And the killer was …. That requires a very deep understanding
LLMs currently are very “fast thinking” from the thinking fast and slow book.
AI Mental Breakdown and getting help from another LLM

LLM training efficiency

LLMs are trained very inefficiently. Imagine teaching a kid by making them read everything and see what happens most.
In the future rewards based on process , not outcome (e.g did I take a good first step, vs did I get the right answer).
We train LLMs based on hard/easy for humans, what about when we do it by stuff that is hard/easy for LLMs

It’s not good enough today

Imagine saying that of a pre-school student.
In 2024, we have gpt-4o (high school student), 2 years ago, we had a pre-schooler.
What will we have in 2 more years?

Where will we get the OOMs - Order Of Magnitude Improvements

See situational Awareness

Compute power: The essay mentions an increase of “~0.5 orders of magnitude or OOMs/year” in compute power - . This is driven by the rapid expansion of computing infrastructure, with plans for trillion-dollar compute clusters and hundreds of millions of GPUs being deployed across the United States.
Algorithmic efficiencies: Another “~0.5 OOMs/year” is expected to come from improvements in AI algorithms- . These advancements are likely to enhance the capabilities of AI systems significantly.
“Unhobbling” gains: LLMs have latent capabilities (for example when they use Chain Of Thought). By using that, we get a boost. How many more such boosts are there.
AI research automation: Once AGI (Artificial General Intelligence) is achieved, the essay suggests that “Hundreds of millions of AGIs could automate AI research, compressing a decade of algorithmic progress (5+ OOMs) into ≤1 year” - This rapid acceleration in AI capabilities could lead to a dramatic increase in overall intelligence.