What is it like to be a machine? | Tamas Vinko | TEDxUniversityCollegeRoosevelt

Key Concepts:

• Consciousness (defined as subjective experience)
• AI (Artificial Intelligence)
• AI Winter
• Maximizing Expected Utility
• Turing Machine
• Algorithm (finite sequence of mathematically rigorous instructions)
• Turing Test
• Generative AI

1. Introduction: Personal Anecdote and the Problem of Defining Consciousness

The speaker begins with a personal anecdote about attending a Depeche Mode concert in Budapest in 1988. He reflects on the shared experience with friends and the difficulty of knowing if they felt the same emotions. This leads to the central problem of defining consciousness, highlighting the subjective nature of experience. He mentions a paper that listed around 200 different definitions of consciousness, illustrating the diversity and confusion among scientists. For the purpose of the talk, he adopts a simple definition: consciousness as "subjective experience."

2. AI Winter and the Definition of AI

The talk transitions to the state of AI in 1988, which was during an "AI winter," a period of reduced interest and funding due to a lack of significant progress. The speaker explains that AI winters are often followed by periods of renewed interest and advancement, such as the current "AI summer." He acknowledges the difficulty of defining AI, noting that it's a topic that can be studied for years without a definitive answer. He presents a mantra often taught in AI courses: "maximize your expected utility." This means achieving the best solution in a probabilistic or noisy environment, where utility is a measure of performance and goodness.

3. The Applicability of AI and the Nobel Prize in Physics

The speaker argues that the value of theoretical AI lies in its applicability. He points to the 2022 Nobel Prize in Physics being awarded to computer scientists for their contributions to AI, particularly artificial neural networks, which form the basis of deep learning. He suggests that this recognition wouldn't have happened without the real-world applications of AI.

4. Computers and the Turing Machine

The discussion moves to computers, which are defined as machines that perform calculations on stored data. The speaker highlights Alan Turing's seminal 1936 paper, which introduced the Turing machine. He describes the Turing machine as a simple device with an infinite tape that can be moved left or right, and a head that can read or write symbols on the tape. The key point is that anything a real computer can compute, a Turing machine can also compute, making it a fundamental concept in computer science.

5. Algorithms and the Question of Programmed Consciousness

The speaker defines an algorithm as "a finite sequence of mathematically rigorous instructions." He contrasts algorithms with recipes, noting that recipes often contain vague instructions (e.g., "add just enough salt"). He then addresses the question of whether it's possible to create an algorithm that programs consciousness. His answer is no, arguing that if the same program is run on two different computers (both Turing machines), the result will be the same, lacking the subjective experience associated with consciousness.

6. The Turing Test and Consciousness Testing

The speaker discusses Alan Turing's other significant contribution: the Turing test, designed to measure intelligence. The test involves a human interacting with either a computer or another human and trying to determine which is which. The speaker notes that AI research is not primarily focused on creating programs that pass the Turing test. He also points out the similarity between the Turing test's reliance on verbal interaction and the way we assess consciousness in humans (asking questions like "Are you alright?").

7. "What Is It Like to Be a Machine?" and Generative AI

The speaker addresses the title of the talk, "What Is It Like to Be a Machine?" He acknowledges that it's a play on Thomas Nagel's famous paper, "What Is It Like to Be a Bat?" He explains that he researched the phrase to ensure some originality and found that while there were a few hits, most were related to the experience of being a machine learning engineer. He then discusses generative AI models like ChatGPT, noting that they learn everything from the internet. He argues that a key difference between humans and generative AI is that humans are creative and know things without the internet, a trait shared with animals, some of whom are believed to be conscious.

8. Conclusion: Current AI and Ethical Considerations

The speaker concludes that current AI systems are not conscious and are unlikely to become so with existing algorithms. However, he cautions against being too certain, drawing a parallel to the skepticism surrounding the possibility of flight before the Wright brothers. He suggests that a technological shift could change things. He raises ethical considerations, such as the potential implications of conscious AI for how we treat computers (e.g., the ethics of turning off a conscious computer). He ends by dedicating the talk to his mother and aunt.

Key Takeaways:

• Consciousness is difficult to define, but subjective experience is a useful working definition.
• AI has gone through periods of boom and bust (AI winters and summers).
• The Turing machine is a fundamental concept in computer science, demonstrating the capabilities of computers.
• Algorithms are precise sequences of instructions, but they may not be able to replicate consciousness.
• The Turing test is a measure of intelligence, but it's not the primary focus of AI research.
• Current AI systems are not conscious, but future technological advancements could change this.
• The possibility of conscious AI raises significant ethical questions.