Generative AI For The Rest Of US

The Mechanical Turk and the Artificial Artificial Intelligence

In the 18th century, Hungarian inventor Wolfgang von Kempelen introduced the Mechanical Turk, a supposed automaton that could play chess, to the European public. This contraption, housed in a large wooden cabinet with a chessboard on its surface, featured a mannequin dressed in traditional oriental attire, complete with a turban, seemingly ready to challenge human opponents to chess. Its appearance as a wonder of mechanical engineering captivated audiences across Europe, including notable statesmen like Napoleon Bonaparte, the French emperor, and Benjamin Franklin, the founding father of the United States. Both were defeated by the Mechanical Turk in chess matches, further fueling the machine's mystique.

In reality, the Mechanical Turk was not the technological wonder it seemed to be. The machine concealed a human chess master within its cabinet, who orchestrated the mannequin's moves via a sophisticated arrangement of levers and pulleys, all while monitoring the game through a cleverly disguised system of mirrors. This illusion of an autonomous chess player was devised to astonish the Empress of Austria-Hungary and the wider public. Despite the machine's initial success in duping its audiences, the reality behind its operation was eventually uncovered in the 1820s, revealing Kempelen's ingenuity in creating one of the most celebrated hoaxes of his time.

Nowadays, the story of the Mechanical Turk is often mentioned in discussions about artificial intelligence and machines' ability to mimic human activities. However, what we can learn from the Mechanical Turk is not about the machine's ability to play chess, but about the human ability to create illusions and, at the same time, his desire to believe in them. This story resonates in modern times, where there is ongoing discussion about the real capabilities of AI versus the public's perception, its ethical considerations, and the potential consequences of its misuse.

Moreover, the "Mechanical Turk" has become a metaphor for any system that appears to be autonomous but is, in fact, controlled by human intervention. This concept is particularly relevant in the context of Amazon's Mechanical Turk, also known as mTurk, a crowdsourcing platform that connects human workers with tasks that AI cannot yet perform effectively. Despite its name, Amazon's Mechanical Turk is a platform that relies on human labor to complete discrete tasks that are beyond the capabilities of computers. It's an "artificial artificial intelligence," as described by Jeff Bezos.

In 2017, Expensify, a popular expense management app that uses AI to extract data from receipt photos, faced controversy. It was discovered that workers from mTurk were manually entering data from receipts, contradicting the company's claim of using advanced AI for this purpose. This incident raised privacy concerns and led to embarrassment for Expensify. Pinterest, the visual discovery engine, also faced criticism for using mTurk workers to complement their algorithm. Other popular platforms like Facebook and Twitter have also been known to use mTurk for various tasks.

Modern Generative AI models, such as OpenAI's GPT-4, don't rely on human labor to generate text and images (even though OpenAI has been criticized for using Kenyan laborers earning $1.32 to $2 per hour to filter out toxic content). When you ask ChatGPT to generate a text, there is no human in the loop. This is why the Mechanical Turk could be considered a thing of the past, as AI models are becoming more autonomous and capable of generating content without human intervention. However, there's still some common ground between the Mechanical Turk and modern AI: they both give the illusion of autonomy and intelligence by mimicking human behavior.

Parrots and dolphins can mimic human speech too. Are they intelligent? Not really. They don't understand what they're saying; they're just mimicking. The same could be said about Generative AI: mimicry doesn't imply understanding or intelligence. Generative AI models have just learned from a vast amount of data, that's the big difference at least for now.