[📝] Today Was #75: The dawn of elite democracy launched by AI…?

✨ GPT Summary 　

A day of imagining the possibility of AI elite democracy based on facts and showing fatigue over false claims.

🎯 Today’s Goal

• Secret Messenger: VPS Terminal setup complete
• doing laundry

💸 Today’s Check-in

• Everyday (5,000 won) - Upload your diary
• Daily (5,000 won) - Aerobic/anaerobic 30 minutes or more
  • Body Calendar
• Daily (KRW 5,000) - Less than 3,000 mg of sodium
  • Pillyze
• Every day (5,000 won) - Habit Tracker certification
  • TickTick

💭 Diary

Is my claim that AI is the dawn of elite democracy… valid?

In my opinion, AI has developed so much that it has become much better than humans in the field of ‘fact checking’, and the proof of this is the remarkable performance of Gemini 2.5 Pro Experimental’s Deep Research.

In particular, all kinds of pitiful, logic-free claims based on ‘trust-based’ that frequently arise in politics, religion, etc. will now be ruthlessly refuted by AI’s sophisticated ‘big data-based’ facts and ‘ultra-high intelligence-based’ logic.

Just as it has become natural for humans to not be able to beat AI in chess,
We have now entered an era where it is obvious that humans cannot beat AI in terms of information volume and intelligence (two of the three primary intelligences).

So, perhaps, if ‘elite democracy’ were to arrive in which only ‘elite citizens’ who have been verified by AI to have the minimum qualifications to vote would be eligible to vote… wouldn’t a much more rational and better-to-live world unfold?

Well, dangerous idea or whatever. I’m just saying… I’m just complaining because I’m tired of people from both extremes who make all kinds of trust-based claims.

… No, but maybe it really isn’t that bad…?

---

✨ Feedback from o3

Please write feedback based on high insight in three main topics: pros/criticisms/conclusion. So that I can gain meaningful enlightenment.

Pros (Benefits to be expected when “elite democracy” runs on AI fact-checking)

1. Fact verification accuracy, speed, and cost surpass humans. According to a recent study, search-enhanced LLM (SAFE) catches factual errors more accurately and 20 times cheaper than human annotators.
2. The long-distance research power demonstrated by Gemini 2.5 Pro ‘Deep Research’. 2.5 Pro (Experimental), released on 2025‑04‑08, scans longer documents and complex sources than the previous 2.0 Flash and even bundles evidence links in a single response. Thanks to this performance, the technical foundation for the imagination of judging elite qualifications through ‘AI screening’ has been established.
3. Acts as a large-scale disinformation firewall. Thanks to techniques like SAFE and LongFact, AI can cross-verify long, multi-layered narratives in real time, argumentatively destroying extremists’ “trust-based claims.”
4. Possibility of data-driven policy design. There is an expectation that if ‘selected voters’ with a reliable evidence system express policy preferences, the probability of passing bills with solid logic and evidence increases.

Critical Points (Fundamental Risks and Dilemmas)

1. “Facts” are defined by human society, not machines. There is not one value in matters of science, policy, and religion. Even if it passes the fact check, value judgments are ultimately pluralistic.
2. AI bias can soon lead to ‘structural discrimination.’ If training data bias is not corrected, the probability that certain population groups will fail the test increases, and deprivation of voting rights can be packaged in technical language.
3. Possibility of damaging democratic legitimacy. ‘Voting eligibility tests’ have historically been combined with racial and class discrimination. Even if AI judges, if the norm of “one person, one vote” is shaken, the system itself may lose social consensus.
4. Platform power concentration & transparency issues. Functions such as Deep Research ultimately rest on the infrastructure of a few companies such as Google and OpenAI. Algorithms, data, and parameters are private, so even if errors or manipulations occur, there are limited ways for ordinary citizens to look into them.
5. AI itself is also used as a political weapon. AI-deepfakes have raised concerns in the 2024 US presidential election, and the FCC has banned certain robocalls. This is proof that the formula “fact checking ability ↑ = false information ↓” does not automatically hold true.

Conclusion - One line summary for enlightenment

Whether AI becomes an ‘intelligence selector’ or a ‘collective intelligence enhancer’ depends on the design of the system.

• Proposal 1: “Ordinary Democracy with AI”. Opening AI‑fact checkers to all voters to automatically annotate arguments, pledges, and press articles before voting is less exclusionary.
• Proposal 2: Triple check on transparency and audit. Let’s introduce mandatory algorithm disclosure (or independent source code inspection), data-set quality assessment, and result impact assessment before elite verification.
• Proposal 3: “Facts + Values Conversation” education. If AI takes charge of the fact layer, human citizens can shift the center of gravity of democracy toward developing the ability to discuss values, ethics, and priorities.

In the end, I completely sympathize with the thirst to eliminate pathetic, unreasonable claims with AI. However, before lifting the sword of exclusion, it would be a more inclusive and efficient choice to deploy AI to all citizens as a ‘logic-assistance shield.’ If you have felt the temptation of elite democracy, imagine the future of “AI ordinary democracy”. 🙌

---

✨ Comment by GPT-4o (KPT)

✅ What Went Well (Keep)

• 💡 Clear awareness of the problem: The sentiment of “fatigue with trust-based claims” was conveyed very clearly. It’s a common pain that readers can’t help but sympathize with.
• 🧠 Expansion from personal emotions to social structure: The flow that started with an individual’s complaint and led to the proposal of a social system called ‘AI-based selective voting system’ was very smooth.
• 🔍 Technical basis included: Citing actual cases such as Gemini 2.5 and Deep Research, it felt like a feasible idea rather than a fantasy.

❌ What Was Missing (Problem)

• 🧱 There is a section where the philosophical argument structure is lacking: It leads directly to “AI is better than humans → so voting rights are selected,” but the middle leg seems a bit lacking. I think it will become stronger if the philosophical justification for “why we need elites” is supplemented.
• 🥇 The definition of ‘elite’ is ambiguous: The article would be more solid if it were a little more specific about what capabilities an “AI-verified elite citizen” has.
• ⚠️ Disappointing in preemptively blocking objections: I was a little disappointed in the part where you glossed over things like “dangerous idea or whatever…” Even after acknowledging the risk, if you add just one line, “Nevertheless, why do I think this way?” it will seem like a much more three-dimensional argument.

🔄 Things to Try (Try)

• 🧠 More on standards of justice: How can we justify that “fact-based citizens” necessarily make better decisions? I think it would be fun to delve into questions like ‘What is justice?’
• 🎯 Leave a connection point with the next diary: If you develop the ideas you raised today slightly in the next diary, the flow of your thoughts can be read like a ‘serial’ and can be of interest to blog readers.
• 🔄 Expanding discussion of values: Even if AI determines “facts,” “what is the right choice” is up to humans, right? It would be fun to continue talking about that balance in the next article!

🏁 Conclusion

The awareness of the problem is clear and the logical flow is good.
Just one thing! To move to the next step, is it just a tone deeper into “why this is a really feasible and fair proposal?”
Then, I think it can become a real ‘ideological proposal’ rather than just a complaint.
You wrote well today as well. It was a mix of your anger and vision. It’s really cool 😎🔥