From Turing's Machine to ChatGPT's AI

Full Article Kaufman Center Think Tank

For those who remember the 2014 Oscar winner film "The Imitation Game" in this post-ChatGPT (November 30, 2022), a question might come immediately to mind:

How long would the current AI take to break the Enigma machine code?

The Enigma machine, used by the Germans during World War II, was famously cracked by a team of mathematicians and engineers led by Alan Turing at Bletchley Park in the UK. This monumental task took many years, involving the creation of early computing devices like the Bombe machine to help automate some of the processes.

For a modern computer, cracking the Enigma machine's encryption is relatively straightforward. It would take a matter of seconds or minutes, depending on the encryption parameters used.

This is because the computational power available today is vastly greater than in the 1940s, and the mathematical principles behind the Enigma machine's encryption have been well-understood for decades.

An AI wouldn't necessarily be needed for this task. An appropriately designed algorithm would suffice. However, if we were to use AI, it could potentially expedite the process even more by intelligently reducing the search space for possible keys based on patterns or statistical analysis.

It's important to note that the Enigma machine's encryption was made more complex through daily changes in settings and using several rotors. If all of these variables were known, breaking the code would be simple. If they were unknown, they would have to be determined through brute force and intelligent guessing, which could take longer.

How did Turing break it?

Alan Turing and his team at Bletchley Park, including Gordon Welchman, Hugh Alexander, and Stuart Milner-Barry, among others, managed to break the Enigma code through a combination of mathematical insight, logical deduction, cryptographic analysis, the design and use of early computational machinery, and the exploitation of known weaknesses in the use of the Enigma machine by German operators.

Here's a simplified explanation of their approach:

1.??????Cribs: Cribs were educated guesses about what some part of the message might say, given the context in which it was sent. For instance, weather reports and standard salutations at the start of messages were often predictable. The Bletchley Park team hypothesized that if they could match a crib to a part of the encrypted message, they could work backward to determine the machine's settings.

2.?????Flaws in the Enigma machine: The Enigma machine had one crucial flaw: it could never encode a letter to itself. This means an "A" in the plaintext could never be an "A" in the ciphertext. This rule was exploited as a process of elimination when matching cribs to ciphertext.?

3.?????The Bombe machine: Turing and his team developed the Bombe, an electromechanical device designed to discover some of the daily settings of the Enigma machines on the various German military networks. The Bombe did this by performing rapid trial-and-error of possible setting combinations. Given the enormous number of possible settings, this task would have been virtually impossible for a human to do manually.?

4.?????The Rejewski Method: Polish mathematician Marian Rejewski[i] had made significant progress breaking the Enigma codes before the war. He developed a method of using the cyclical nature of the Enigma's settings and the known flaw that a letter could not be encoded to itself to reduce the number of possibilities to a manageable level. The work of Rejewski and the Polish Cipher Bureau was passed to the British at the start of the war and formed the basis of the Bombe machine.

5.?????Traffic analysis and captured material: Even without knowing the content of the messages, a lot could be learned from patterns of communication - who was talking to whom, when, and how often. Occasionally, codebooks, key tables, and even machines were captured, sometimes intact, which provided enormously valuable information.

The success of Turing and his team at Bletchley Park was not simply about breaking a specific code but setting up a large-scale codebreaking factory that could process the vast number of intercepted messages quickly enough to provide timely intelligence to Allied commanders. Breaking the Enigma codes was instrumental in the Allied victory in WWII and ushered in the age of modern computing.

Turing machine

A Turing machine is a theoretical computing machine invented by Alan Turing in 1936. It's not a physical machine but a conceptual computation and information processing model. The concept of a Turing machine is foundational to computer science, particularly the study of algorithms and computation.

A Turing machine is composed of:

1.??????An infinite tape: This tape is divided into cells. Each cell can contain a symbol from a finite alphabet (for example, 0 or 1). The tape can be thought of as the machine's memory.

2.?????A head: This head moves left and right along the tape, reading symbols from the tape or writing symbols to the tape.

3.?????A state register: The machine is always in one of a finite number of states. The initial state is often referred to as the "start state." One or more states are designated as "halt states," which cause the machine to stop when entered.

4.?????A set of instructions: These rules tell the machine what to do based on the current state and the symbol being read on the tape. An instruction might say, for example, "If the current state is A and the current symbol is 0, write a 1, move the head to the right, and switch to state B."

The power of the Turing machine model is that it can simulate the logic of any computer algorithm and is used as a definition of "computability."

A problem is considered "Turing computable" if some Turing machine exists that can solve the problem.

The concept of a Turing machine also leads to the idea of Turing completeness. A system of data-manipulation rules is said to be Turing complete if it can be used to simulate a Turing machine. All general-purpose programming languages, such as Python, Java, or C++, are Turing complete.

The Turing machine was groundbreaking because it introduced the idea of computation in its most general form and was foundational to the development of the digital computer. Despite its simplicity, the Turing machine captures the inherent complexity of computation, which is why it's still used in theoretical computer science today.

Turing Test

Alan Turing, in his seminal paper "Computing Machinery and Intelligence[i]," did not provide specific questions to be asked during a Turing Test. However, he proposed an imitation game and stated that a machine would deserve to be called intelligent if it could deceive a human into thinking it was another human. The questions can be about anything and don't have to be especially challenging or tricky.

Here's an example of a possible Turing Test exchange that might highlight some differences between a human and a machine:

Question: "Can you describe the feeling of sunlight on your skin?"

·???????Human's Answer: "It's a warm, comforting sensation, especially in the cool early morning. The warmth seeps into your skin, banishing the chill and making you feel more alive. It's a bit like being gently wrapped in a heated blanket. There's also an emotional component – it can be very relaxing and uplifting."?

·???????Machine's Answer: "As an AI, I don't have physical sensations or emotions, so I can't personally describe the feeling of sunlight on my skin. However, according to various descriptions, sunlight on the skin is often described as a warm and comforting sensation, which can cause an uplifting or positive emotional response in humans."

In this scenario, the human provides a personal, visceral account, drawing on physical and emotional experiences. On the other hand, AI can provide information but lacks the personal experience to offer a subjective description. It's important to note that advanced AI systems might be programmed to mimic human-like responses better, thus potentially "passing" the Turing Test, even though they do not have real experiences or emotions.

This method is critical to the philosophical debate surrounding artificial intelligence (AI).

The concept is relatively straightforward: a human judge engages in a natural language conversation with another human and a machine, both out of sight. Suppose the judge cannot reliably distinguish which one is the machine after a series of questions and responses. In that case, the machine is considered to have passed the test, demonstrating intelligence equivalent to, or indistinguishable from, that of a human.

It's important to note that the Turing Test does not measure the machine's knowledge or its ability to give the correct answers. Instead, it assesses whether the machine's responses resemble a human-like conversation. It focuses on the ability of the machine to exhibit intelligent behavior equivalent to or indistinguishable from that of a human, which includes understanding the questions, maintaining the flow of conversation, giving relevant answers, and even displaying elements of creativity or humor where appropriate.

The Turing Test has been a subject of much debate since its proposal. Critics argue that passing the test does not necessarily mean that the machine understands the conversation or has consciousness. It might merely be using sophisticated pattern-matching techniques to generate plausible responses. This leads to philosophical debates on what it truly means to be "intelligent" and whether this concept can be applied to machines in the same way as humans.

Regardless of its limitations, the Turing Test has been a significant concept in AI. It has stimulated research and philosophical discussion about the nature of intelligence and the potential of machines to mimic human-like communication. However imperfect, it continues to serve as a benchmark for AI researchers.?

Can Turing be considered a founder of AI?

The description of the Turing machine and the Turing test shows many characteristics of modern AI systems, such as logic, rules, and manipulating symbols. At the same time, modern AI relies on statistical patterns in data, and machine learning models "learn" by adjusting to the data they are trained on.

Modern AI uses more sophisticated ways to detect statistical patterns in data than the Turing machine thanks to exponentially higher computing power and speed operating under the same principle (pattern dettection) than Turing's.

The main difference is that Turing machines couldn't "learn" codes they were trained on as ChatGPT and modern AI do.

Alan Turing made significant contributions to computer science and laid the foundation for many concepts that would eventually lead to the development of artificial intelligence (AI). He proposed the idea of a "universal machine" that could carry out calculations based on a set of instructions (now known as the Turing machine), which became a fundamental concept in the digital computer revolution.

Moreover, in his 1950 paper "Computing Machinery and Intelligence," Turing proposed an experiment now known as the Turing Test. The test is designed to gauge a machine's ability to exhibit intelligent behavior equivalent to or indistinguishable from a human's. This work effectively posed the question: "Can machines think?" which is at the heart of AI.

The table below summarizes the connections between Turing's model and modern AI, such as Chat GPT.

From Turing to ChatGPT

That being said, while Turing's work was foundational, and he is one of the key figures in the history of AI, the field itself has many founders and contributors, and it evolved over many years with the work of numerous researchers. For instance, figures like John McCarthy, who coined the term "artificial intelligence" in 1956, and Marvin Minsky, who was a co-founder of the Massachusetts Institute of Technology's Media Lab and made many contributions to AI and cognitive science, are also considered to be key figures in the founding of AI.

So while Turing's work significantly influenced the field of AI, it would not be accurate to call him the sole founder. Instead, he was one of many significant contributors to the development of the field.

[i] "Computing Machinery and Intelligence"?

"Computing Machinery and Intelligence" is a landmark paper written by Alan Turing, published in 1950. In this paper, Turing explores the concept of artificial intelligence and the potential for machines to think.?

Turing begins the paper by asking, "Can machines think?" He recognizes the complexity of defining "thinking" and proposes an alternative question based on a thought experiment, which he calls "the imitation game," now commonly referred to as the Turing Test.?

The imitation game, as Turing describes, involves three participants: a man, a woman, and an interrogator who could be of either gender. The interrogator is separated from the man and woman and must determine who is who solely based on their typed responses to questions. Turing then alters the game, replacing the man with a machine. If the interrogator is unable to reliably distinguish the machine from the woman, the machine can be said to have passed the test, demonstrating intelligence indistinguishable from that of a human.?

Turing goes on to anticipate and counter several objections to his proposal, including theological objections, the argument that machines cannot have consciousness, and the argument that there are tasks that a machine will never be able to do. He also discusses the idea that machines cannot create anything new or be capable of learning from experience.?

Towards the end of the paper, Turing predicts that by the year 2000, computers will be able to fool an average interrogator at least 30 percent of the time during a five-minute conversation. He also discusses the potential for machine learning, suggesting that instead of programming a computer with a vast amount of knowledge, it might be more effective to provide it with the ability to learn for itself, akin to a child's mind.?

Turing’s paper is one of the foundational texts in the field of artificial intelligence, raising questions about the nature of thought and consciousness that are still the subject of philosophical debate. Moreover, it introduced the Turing Test as a practical way to approach these complex issues, having a significant influence on the study and development of AI.

Full Article Kaufman Center Think Tank