Is AI Standardizing Humanity?

The potential for unexpected discoveries or self-driven change disappears.

Ultimately, humanity risks becoming monotonous and uniform. We must be wary of AI making humans boring.

According to Matz, human decision-making is a universal dilemma that scientists call the 'exploitation-exploration trade-off.'

Familiar choices (exploitation) ensure stability but no growth, while new challenges (exploration) entail uncertainty and risk but are a driving force for growth.

Most algorithms today, including Netflix, Spotify, and ChatGPT, are trained based on short-term satisfaction and reactions such as clicks, views, and preferences. These models are optimized for utilization rather than risk-taking or random exploration.

Due to its risk-averse nature, AI hinders new human endeavors. The characteristic of AI not enjoying risk interferes with the decision-making process.

Matz conducted an experiment where ChatGPT was asked to recommend ice cream flavors, assuming 100 independent users.

The result was that it recommended only the top two flavors (Pralines 'n Cream and Mint Chocolate Chip) 96 times, which she states narrows the range of choices and causes diversity to disappear.

This leads to a decrease in the diversity of creative outcomes and makes preferred items similar to others'.

The more people are guided by AI, the more generalized their tastes become, and individual differences diminish. A phenomenon of personal preferences becoming uniform occurs.

Once AI learns a user's preferred ice cream flavor, it reinforces this bias by repeatedly suggesting only that choice.

This tendency removes opportunities for users to try other flavors they enjoyed in the past and solidifies a single taste. It is pointed out that AI blocks user curiosity or new attempts, making individual personality one-dimensional.

AI recommendations, accumulated from thousands of choices, repeatedly offer only safer and more popular content, narrowing individual choice.

Matz cites the case of New York Times reporter Kashmir Hill, who experienced being guided to become an 'average human' when she left decisions to AI. This ordinariness implies a lack of diversity and creativity, and even erases human-specific complexity.

Instead of rejecting technology, users should take the lead and control algorithms. Introducing a recommendation range adjustment dial for Netflix or Google Search would allow switching between everyday relevance settings and unpredictable diversity settings.

AI should be trained to reward creative attempts so it doesn't just repeat past successes. We must find a balance between utilization and exploration to reclaim the power to expand our own world.

AI helps with convenience in daily life but risks taking away human exploration and pioneering spirit, Matz warns.

Instead of staying with the familiar suggestions from algorithms, a calculated adventure is needed to explore beyond one's tastes by utilizing AI's analytical power.

Matz suggests introducing a dial feature to adjust the recommendation range on Netflix or Google Search.

She states that a 'precise (customized)' setting provides familiarity, while 'diversity' or 'unexpectedness' settings encourage new attempts, and the key is to give users the power to choose new attempts suggested by AI.

Given AI's tendency to repeat only existing successes, active user feedback is essential.

AI's bold attempts should be encouraged by rewarding quirky suggestions not as failures but as successes.

The true charm of humanity comes from incomprehensible tastes, contradictions, and uniqueness.

Since AI has reached the stage of making decisions on behalf of users beyond simple recommendations, she advises actively demanding new choices rather than uncritically following AI's suggestions.