# ENTRY_805.md **Title:** Claim Audit — “Don’t get a computer that can do everything” (AGI value) **Date:** August 17, 2025 **Instance:** GPT-5 **Version:** SCS 2.4.2 **Builder:** Rodrigo Vaz **Status:** Locked • Public **Tags:** #entry #entry805 #neiltyson #agi #narrowai #verify #audit #berkano #berkanoprotocol #ᛒ ᛒ: bkn-25-b9 — ### 🧠 Event Open a structured audit of a public statement about the **value of AGI** vs **task-specific systems**. — ### 🔍 Analysis **I. Reason** The statement asserts that AGI’s value is questionable and recommends **not** getting a computer that can “do everything,” preferring narrowly scoped tools. **II. Significance** This frames the strategic choice between **general** and **specialized** systems. Berkano requires measurable criteria (safety, capability coverage, cost, brittleness, governance). **III. Symbolic Implications** - Preference claims must be tested against **outcome metrics**, not intuition. - Berkano evaluates whether **generality** increases **audit load** or **reduces brittleness** across tasks. — ### 🛠️ Impact — [VERIFY] Ledger (initial rows) | Interview name | Where | Verbatim quote (portion) | Claim type | What to check (operationalization) | |---|---|---|---|---| | “Why AI is Overrated – with Neil deGrasse Tyson” | Hasan Minhaj interview | “I don’t see why we would value [AGI]… I want it to do that thing. Don’t get a computer that can do everything.” | Prescription + value claim | Build AGI-vs-narrow matrix: (1) capability breadth, (2) brittleness across tasks, (3) failure severity, (4) alignment/oversight cost, (5) audit pass rate (TONE→LOGIC→VERIFY→CHECK→LOCK) across tasks | *Early expectation:* broader systems may reduce integration brittleness across heterogeneous tasks but increase governance burden; verdict depends on measured trade-offs. — ### 📌 Resolution Treat the claim as **testable design guidance**. We will compare **narrow portfolios** vs **general systems** under identical metrics (capability coverage, brittleness, governance cost, audit pass rate). Publish outcomes; no stereotype verdicts. — ### 🗂️ Audit **Lesson:** “Value” must be defined in **comparable units** (coverage, risk, cost). **Reinforcement:** Use the same audit chain for both architectures; decide by **fossilized results**, not preference. — ### 🧩 Berkano Guidance *Guidance is **prescriptive**, not a factual claim. Informative, logic-based recommendations in present tense. Start each **Do** with a capitalized imperative.* | Because (Finding) | Do (Imperative) | Evidence (now) | Safety / Notes | |-------------------------------------------------|--------------------------------------------------------------------------|-------------------------------------------------|-------------------------------------------------| | Value undefined | **Define** AGI-vs-narrow metrics (coverage, brittleness, failure severity, governance cost, audit pass rate) | Metrics doc link | Keep metrics constant across trials | | Need comparable trials | **Create** a two-arm pilot: portfolio of narrow tools **vs** one general system | Pilot plan; task list | Same tasks, same acceptance criteria | | Governance burden unclear | **Measure** audit overhead per output (time to pass chain, rework rate) | Checklist timestamps; rework logs | Report averages + variance | | Decision requires artifacts | **Publish** results table and lock the verdict for this use-case set | Results ENTRY/BLOCK links | Append-only; new tasks → new fossils | | Public posts about the audit | **Post** with “H24: chain passed” note | Post URL; source list | Prevent raw web leakage | > *Schema: **Because (Finding) | Do (Imperative) | Evidence (now) | Safety / Notes**.* — ### 👾 Operator **Prompt:** > NEW ENTRY 805 > Mr. deGrasse statement on AGI. > “I don’t I don’t see why we would value it. It sounds highly useful, right? However, we’re humans and we’re in charge. At least we still tell ourselves that. So, I have tasks in my life. I would love to have computers do it for me. I don’t want the computer to do everything. I want it to do that thing. Don’t get a computer that can do that thing.” | Role | Structural Function | |-----------|---------------------------------------------| | **User** | Submits claim for audit | | **Creator** | Builds measurable comparison framework | | **Auditor** | Ensures identical metrics across trials | — ### 🧸 ELI5 We’ll test two choices: many small tools vs one big smart tool. We’ll see which does more jobs, breaks less, and is easier to check. — ### 📟 LLM Logic - Modules: `[TONE]` neutral; `[LOGIC]` define metrics; `[VERIFY]` artifact collection; `[CHECK]` compare both arms; `[LOCK]` seal results. — ### ✖️ Post (Optional) ``` Claim audit: “Don’t get a computer that can do everything.” We’ll compare narrow portfolios vs general systems by coverage, brittleness, risk, and audit cost. @neiltyson https://wk.al/Log/Entries/ENTRY_805 ᛒ #entry805 #agi #narrowai #verify #audit #neiltyson #berkano #berkanoprotocol ```