If you thought that Artificial Intelligence was just a modern twenty first century venture, are you in for a surprise. While the present level of sophistication has largely happened in the last decade or two, the roots trace back to almost 100 years ago. Here is a list of some things that would qualify as the earliest ventures into AI - when computers were basic. 

1. Logic Theorist (1956)

Logic Theorist is widely regarded as the first true artificial intelligence program. Developed by Allen Newell, Herbert Simon, and Cliff Shaw, it was designed to replicate the way humans solve logical and mathematical proofs. The program successfully proved 38 theorems from Principia Mathematica and even found a proof that was shorter and more elegant than the human-written version.

Its importance lies in demonstratingthat computers could perform symbolic reasoning—one of the major goals of early AI. For the first time, a machine wasn't just calculating numbers; it was manipulating ideas.

2. General Problem Solver – GPS (1957)

General Problem Solver, created by Newelll and Simon after Logic Theorist, attempted to take AI a step further by solving problems across many different domains. It used heuristic search—strategies that mimic human reasoning—to find solutions to puzzles, logic tasks, and planning problems.

Although GPS worked only in some tightly structured scenarios, it helped lay the basic foundation for modern AI planning and cognitive models. Its creators believed it could one day in the future model any human thought process, making it one of the most ambitious early AI systems.

3. IBM Shoebox (1961)

IBM’s Shoebox was one of the earliest attempts at speech recognition patterns. The device could recognise 16 spoken words and the digits 0 to 9, translating voice input into electrical signals to perform simple calculations. For its time, this was groundbreaking—computers were finally beginning to “listen” to humans!

While this sounds very basic by today’s standards, Shoebox showed that machines could bridge the gap between human speech and computer commands. It planted the earliest seeds for the voice assistants which we use today.

4. ELIZA (1964–1966)

Written by Joseph Weizenbaum at MIT, ELIZA was one of the first natural-language conversation programs. Its most famous script, “DOCTOR,” mimicked a Rogerian psychotherapist by turning the user’s statements back into questions. This made many users feel like the computer genuinely understood them!

ELIZA’s impact went far beyond its technical complexity. It revealed how easily people could form emotional connections with computers—a theme that continues to shape AI ethics, chatbot design, and human–computer interaction research.

5. DENDRAL (1965)

Developed in Stanford by Edward Feigenbaum, Joshua Lederberg, and colleagues, DENDRAL was the world’s first successful expert system. It analysed chemical mass-spectrometry data and helped chemists determine the molecular structure of organic compounds. In many cases, DENDRAL’s conclusions were more accurate and faster than those produced by human experts.

Its success proved that AI could make meaningful, practical and logical contributions to scientific research. DENDRAL also marked the beginning of knowledge-based systems, a major branch of AI throughout the 1970s and 1980s.

6. SIR – Semantic Information Retrieval (1966)

SIR, created by Bertram Raphael, was an early experiment in machine understanding of English. The program could interpret simple sentences, retrieve relevant information, and answer basic questions posed in natural language. Although limited in vocabulary, it was one of the first systems to go beyond pattern matching and attempt semantic interpretation.

Its ideas became foundational for modern search engines and question-answering systems. SIR demonstrated that computers could begin to extract meaning from language rather than simply matching text to text. This paved the way for a higher level of interaction. 

7. MACSYMA (1968)

MACSYMA, developed by MIT, was the first large and powerful computer algebra system. It could integrate and differentiate symbolic equations, solve algebraic problems, manipulate expressions, and tackle complex mathematical operations that previously required human mathematicians.

It became one of the most influential mathematical software systems ever created and directly inspired later tools such as Mathematica and Maple. MACSYMA showed that AI could handle sophisticated mathematical reasoning, not just simple logic or language tasks.

8. Shakey the Robot AI System (1966–1972)

Shakey, developed by SRI International, was the first robot controlled by an integrated AI software system. It combined perception (via cameras and sensors), reasoning, and physical movement. Shakey could map its environment, plan a sequence of actions, and execute those actions autonomously without further human input. 

One of its key innovations was the STRIPS planner, which became the basis for much of modern AI planning. Shakey proved that AI could be embodied in physical machines—not just live inside software—and laid the groundwork for modern robotics.

9. SHRDLU (1968–1970)

SHRDLU, built by Terry Winograd at MIT, was a major leap forward in natural-language understanding. It allowed users to type conversational commands into a computer, and the system would manipulate virtual blocks inside a simulated world. It could answer questions, perform tasks, reason about objects, and even explain why it made certain decisions.

Because SHRDLU operated in a constrained “blocks world,” it achieved a level of language understanding decades ahead of its time. It remains a landmark example of how AI can combine language, logic, and simulated action.

10. MYCIN (1972)

MYCIN, developed at Stanford University, was one of the most influential expert systems ever created. It diagnosed bacterial blood infections and recommended specific antibiotic treatments based on rule-based reasoning. In controlled tests, MYCIN performed at or above the level of expert physicians.

Although never deployed in hospitals due to legal and ethical concerns, MYCIN shaped the design of countless expert and decision-support systems. Its approach to handling uncertainty (the “certainty factor” model) influenced AI research for decades.

So there you are - proof that AI is not just some modern mumbo-jumbo, but has a journey that stretches to almost a century ago. What are your thoughts?