Scott Hladun — Software & Instructional Designer

This intellectual production was created in collaboration with Google's Gemini LLM, specifically using the 2.5 Pro (preview) model. Gemini is a multimodal Large Language Model (LLM), meaning it was trained on a range of data such as text, images and audio. During the pre-training, these models ingest vast amounts of publicly available data to create sophisticated prediction algorithms that pick up on patterns within the data. Afterwards, these models are fine tuned by human feedback using a reward system that ranks responses based on human preferences (Manyika & Hsiao, 2024). For the sake of clarity in determining each of our contributions to this work, all Gemini responses will be displayed in a format that appears like this. Gemini and I have developed responses to the same prompts, and you can find my analysis that compares and contrasts them at the end of this post. The exact transcript of my interaction with Gemini can be found here: Gemini Conversation.

Alan Turing

Widely regarded as the founder of the field of computer science, he was a mathematician and cryptographer who worked on the code breaking efforts of German ciphers during WWII. He initially proposed the idea of the "Turing Test", which is a widely regarded artificial intelligence design standard. (Biography.com, 2020).

Alan Turing was a British mathematician and a foundational figure in computer science. He conceptualized the "Turing machine," a theoretical model of a general-purpose computer, crucial for AI's development. Turing believed intelligence could be identified through the "Imitation Game" (now Turing Test): if a machine could converse indistinguishably from a human, it exhibited intelligence.

John McCarthy

John McCarthy was an American born mathematician who was credited with coining the term "Artificial Intelligence". During his professorship at Stanford University, he founded SAIL, the Stanford Artificial Intelligence Lab. His many contributions to the artificial intelligence research community included developing programming languages LISP and Elephant 2000 (Encyclopedia Britannica, 2022).

John McCarthy, a founding father of AI, coined the term "artificial intelligence."He organized the 1956 Dartmouth workshop, launching AI as a field, and developed Lisp, a key AI programming language. McCarthy defined intelligence as "the computational part of the ability to achieve goals in the world."

Herb Simon

Herb Simon was a Nobel Prize winner in economics for his research into decision-making processes in large organizations. He proposed the idea of “bounded rationality”, suggesting that people can only make as rational decisions as the information they have, which leads to satisfactory outcomes rather than optimal ones (UBS, n.d.).

Herb Simon, a Nobel laureate and AI pioneer, co-created the first AI program, Logic Theorist. He argued that intelligence involves problem-solving using information processing and heuristics (mental shortcuts).Simon believed intelligence could be identified by a system's ability to exhibit complex, goal-directed behavior and decision-making, similar to human "bounded rationality"—making "good enough" decisions with limited information.

Marvin Minsky

Marvin Minsky was a mathematician and computer scientist. He co-founded MIT's AI Lab in 1959. He believed that the human brain was a machine that could be replicated by a computer and spent much of his research exploring how intelligence and thought are produced within the brain (BBC, 2016).

Marvin Minsky, an AI pioneer, co-founded MIT's AI Lab. He contributed significantly to neural networks and robotics, and developed "frames" for knowledge representation. Minsky viewed intelligence not as a single entity, but as an emergent property from the complex interactions of many simpler, non-intelligent "agents" – a concept detailed in his "Society of Mind" theory.

Timnit Gebru

Timnit Gebru is an AI ethics researcher who worked at Google. She challenges many of the mainstream ideas surrounding AI development to encourage diversity and environmentalism in AI practices. She believes that intelligence comes from an understanding of content and context, rather than just the ability to manipulate large swaths of content for the illusion of comprehension (Hao, 2020).

Timnit Gebru is a computer scientist known for her work on algorithmic bias and ethical AI.She co-founded Black in AI and founded the Distributed AI Research Institute (DAIR).Gebru's work focuses on the societal impact of AI, advocating for fairness, accountability, and transparency, particularly concerning marginalized communities. She emphasizes identifying intelligence through its real-world effects and potential harms, rather than abstract benchmarks.

How do “machine (programming) languages” differ from human (natural) ones?

Programming languages are scoped to the context of the program the developer is building, while natural language can have various meanings depending on the communication context, such as tone, body language, or even the location in which a conversation is taking place. Natural language evolves and adapts, with nuances between communities and individuals, where slang and jargon are still comprehended by the message recipient. Programming language is rigid in this sense, it is unable to understand jargon or slang as computers cannot interpret the meaning afforded to it. Lastly, programming languages are unable to accommodate imperfections with syntax or vocabulary, leading to errors, while human language communication can still be understood with limited vocabulary (Harris, 2018).

Machine (programming) languages are formal, designed for computers. They demand precise syntax and are unambiguous, directly instructing hardware or being compiled into machine code. Their vocabulary is limited and their structure rigid.

Human (natural) languages, conversely, are evolved, complex, and inherently ambiguous. They possess rich vocabularies, flexible grammar, and rely heavily on context and unspoken understanding. Natural languages facilitate nuanced communication, emotional expression, and abstract thought, purposes far broader than the direct command-and-control function of machine languages. They are learned intuitively, while programming languages require deliberate study.

How does “machine (artificial) intelligence” differ from the human version?

Using François Chollet’s framework, we can consider that human intelligence is defined by our efficiency to acquire skills across a scope of tasks, and our ability to generalize these skills to novel situations. Humans are capable of combining past experiences, and knowledge to entirely new and abstract tasks, even if those experiences are limited.

Machine intelligence, on the other hand, focuses on specific sets of tasks that are often known beforehand and trained for by the developer. This creates the illusion of understanding, yet when provided with modified tasks, the AI fails to generalize the acquired skills to the new situation. (Chollet, 2019)

Machine (artificial) intelligence excels at processing vast data, pattern recognition, and learning specific tasks with high speed and accuracy, but typically lacks true understanding, consciousness, and generalizability across diverse, novel situations.

Human intelligence, while slower in raw computation, possesses consciousness, emotional depth, common sense, and a superior ability for abstract reasoning, creativity, and adapting to entirely new environments. AI currently mimics cognitive functions; humans possess genuine understanding and self-awareness. AI is a tool, whereas human intelligence is an intrinsic, holistic capability.

How does “machine learning” differ from human learning?

Machine learning relies on vast datasets and algorithms to identify patterns and make predictions for specific tasks. It learns by optimizing mathematical functions, often without genuine comprehension or consciousness.

Human learning, conversely, is more versatile and context-aware. It integrates experience, abstract reasoning, intuition, and social interaction. Humans can learn from sparse data, transfer knowledge broadly, understand causal relationships, and possess self-awareness and intention, which current machine learning lacks. Humans learn holistically; machines learn statistically for defined objectives. Machine learning relies on algorithms and vast datasets to identify patterns and make predictions, often requiring explicit programming for different tasks. Its "learning" is statistical, improving accuracy on specific problems with more data, but typically lacks genuine understanding or common sense.

Human learning is more holistic, driven by curiosity, experience, and social interaction. It involves conceptual understanding, abstract reasoning, and the ability to generalize from few examples. Humans learn with contextual awareness, develop intuition, and can transfer knowledge fluidly across diverse, unrelated domains, a capability current machine learning largely lacks.

References

BBC. (2016, January 26). Ai pioneer Marvin Minsky dies aged 88. BBC News. https://www.bbc.com/news/technology-35409119

Biography.com Editors (2020, July 22). Alan Turing. Biography.com. https://www.biography.com/scientists/alan-turing

Britannica, T. Editors of Encyclopaedia (2022, October 20). John McCarthy. Encyclopedia Britannica. https://www.britannica.com/biography/John-McCarthy

Chollet, F. (2019, November 5). On the measure of intelligence.

Harris, A. (2018, November 1). Human languages vs. programming languages. Medium. https://medium.com/@anaharris/human-languages-vs-programming-languages-c89410f13252

Hao, K. (2020, December 4). We read the paper that forced Timnit Gebru out of google. here’s what it says. MIT Technology Review. https://www.technologyreview.com/2020/12/04/1013294/google-ai-ethics-research-paper-forced-out-timnit-gebru

Manyika, J., & Hsiao, S. (2024, July 25). What is Gemini and how it works. Gemini. https://gemini.google/overview/

UBS. (n.d.). Herbert Simon. https://www.ubs.com/microsites/nobel/en/laureates/herbert-simon.html. https://www.ubs.com/microsites/nobel-perspectives/en/laureates/herbert-simon.html