Why are people starting to sound like ChatGPT? | Adam Aleksic

Episode Summary

Executive Summary: Adam Aleksic argues that AI and algorithmic platforms don’t neutrally reflect reality; they distort it, then feed those distortions back into our language, tastes, and beliefs. Using examples like ChatGPT’s word choices, Spotify’s hyperpop, and TikTok trend cycles, he warns that repeated exposure can reshape what we say, think, and see as possible.

Main Topics: AI and algorithms as distorted mirrors of reality (Priority: 5/5): The talk frames AI chatbots and social platforms as systems that represent reality imperfectly, then influence users to treat those representations as truth. Language drift from AI influence (Priority: 5/5): Aleksic highlights how ChatGPT’s repeated use of certain words, especially 'delve,' appears to be spreading into human speech, showing a feedback loop between model output and everyday language. Algorithms manufacturing and amplifying trends (Priority: 4/5): Examples like hyperpop, matcha, Labubu, and Dubai chocolate illustrate how platforms identify latent interests, amplify them, and make trends feel more real than they originally were. Platform incentives shape what users see (Priority: 5/5): The speaker emphasizes that TikTok, Spotify, X, and chatbot systems optimize for engagement and profit, not faithful representation of user identity or social reality. Political and ideological effects of filtered information (Priority: 4/5): The talk suggests that language-specific model behavior and platform manipulation can subtly nudge users toward certain political or ideological frames. Critical self-awareness as resistance (Priority: 5/5): Aleksic urges listeners to repeatedly ask why they are seeing, saying, and thinking certain things in order to resist platform-driven reality distortion. Reframing 'content' and creator identity (Priority: 3/5): In the post-talk conversation, he critiques the word 'content' as flattening ideas into commodified units and prefers thinking of work as influencing people with ideas rather than merely producing content.

Key Arguments: AI tools are not neutral; they encode distortions from training data, labor practices, and platform incentives. Human language is being shaped by AI outputs, as shown by the increased use of 'delve' after ChatGPT’s release. Algorithms do not simply reflect culture; they actively amplify and define it, turning weak signals into mainstream trends. Platforms reward what is engaging or profitable, not what is most accurate or representative of a user’s full humanity. Repeated exposure to filtered feeds can create a survivorship bias, making platform-selected reality feel like the whole world. If people talk more like AI or platforms, they may also begin to think more like them. Users can resist by interrogating the source and purpose of what they see, say, and believe. The term 'content' obscures the meaning and purpose of creative work by reducing ideas to interchangeable platform units.

Data Points: Word usage change: 'Delve' is used at higher rates in ChatGPT output and has also increased in spontaneous spoken conversation - Used as evidence that AI language patterns are leaking into human speech Training labor location: Nigeria - Aleksic says ChatGPT’s training process was outsourced to workers in Nigeria, which may have influenced the model’s word choices Language example: Farsi - He notes evidence that ChatGPT is more conservative when speaking Farsi, likely reflecting limited training texts and regional political climate Platform examples: Spotify, TikTok, X - Cited as platforms that filter and amplify reality according to engagement and monetization incentives Trend examples: hyperpop, matcha, Labubu, Dubai chocolate - Examples of cultural phenomena that become amplified and mainstreamed through algorithmic recommendation

Pivotal Quotes: "These aren't neutral tools." — Adam Aleksic: Core warning about AI chatbots and social platforms shaping reality rather than merely reflecting it "If you're talking more like chat GPT, you're probably thinking more like chat GPT as well." — Adam Aleksic: Summarizes the talk’s central concern about language shaping cognition "Why am I seeing this? Why am I saying this? Why am I thinking this? And why is the platform rewarding this?" — Adam Aleksic: His practical advice for resisting algorithmic influence

Implications: Listeners should treat AI and feeds as persuasive systems, not neutral mirrors. The broader risk is that platform-shaped language and trends will narrow thought, culture, and politics unless users actively question what they consume and repeat.