ノードから物語へ:思考の道具としてのフィクション
Andrej Karpathyは、基本的な入力が安価になることで予測不可能な新たな活動カテゴリーが爆発的に増加する現象について考察している。
キーポイント
基本入力のコスト低下
AI分野において、基本的な入力(例:計算リソース、データ)のコストが劇的に低下しているという前提を示している。
予測不可能な活動の爆発
コスト低下が、誰も予測できなかった新たなカテゴリーの活動や応用を爆発的に生み出す可能性について言及している。
思考ツールとしてのフィクション
記事タイトルから、フィクション(物語)を思考のツールとして活用する視点が示唆されているが、本文では直接言及されていない。
影響分析・編集コメントを表示
影響分析
この記事は具体的な技術的進展ではなく、AI分野の長期的なトレンドに関する概念的考察を提供している。業界全体への直接的な影響は限定的だが、リソースコストの変化がもたらす予測不可能なイノベーションについて考えるきっかけとなる。
編集コメント
Karpathyの洞察は示唆に富むが、本文が非常に短く具体的な内容に乏しいため、分析の深さに限界がある。より詳細な展開があれば評価は高まっただろう。
土曜日、私は、ある基礎的なインプットが低コスト化したとき、誰も予想しなかった形で新たな活動のカテゴリーが爆発的に拡大すると、何が起こるのかについて論じました。
原文を表示
On Sunday I wrote about what happens when a fundamental input gets cheap and new categories of activity explode in ways nobody predicts. I teased, at the end, that I’d been writing science fiction about what life looks like on the other side of that explosion.
The same day, Citrini Capital published The 2028 Global Intelligence Crisis. A fictional macro memo from June 2028 describing the economic fallout of AI displacing white-collar workers. S&P down 38%. Unemployment at 10.2%. Mortgage market cracking. It’s been everywhere these last few days.
I bring it up because it’s the same starting conditions as what I’ve been writing about and exploring. Same premise. Very different place to stand while you look at it.
The Citrini piece counts the jobs lost. It models the displacement spiral: AI gets good, companies cut headcount, spending collapses, the feedback loop accelerates. If you only count the jobs that disappear, the math is brutal.
I’ve been more focused on the new jobs that will be created.
Basically since I created it with Claude, I’ve been clicking around the Traffic Jam Explorer. Sometimes I would get stuck on some of the leaps from node -> second order effect -> traffic jam -> new role. The abstractions too abstract. So I started asking Claude to just write me a short story about a day in the life of someone doing one of these jobs. Show me what it actually *means* for a person to be a “Computational Experience Reviewer” on a random afternoon.
What came back was surprisingly useful, fiction as a *thinking tool*. The stories made the abstractions concrete in a way the tree in the Traffic Jam Explorer couldn’t. I started using them to reason through all types of AI-related problems I’d been kicking around with friends, and the narratives kept unlocking angles my analysis alone missed. Protocolized has been doing something like this with their protocol fiction work for a while now, and this exercise finally made it click for me that fiction can let you stress-test ideas or get a better sense of them by making them live somewhere you can imagine.
So I built it a home to share the best of it with you all.
Near Zero: weekly short stories, set in a world where the cost of building software has collapsed. New professions, new problems, new people navigating what comes after the disruption everyone keeps modeling. If all goes according to plan, I’ll also collect the stories into an ebook on Kindle at the end of each month.
The first piece is called *The Executable Muse* and it looks at a day in the life of a “Senior Computational Experience Reviewer” and the rise of software as content. I think you’ll all really enjoy it.
Works On My Machine continues as usual with open source and other projects, me building and thinking in public. You can think of Near Zero as a companion to this newsletter and expect to see some of the inventions in that world get built and launched here.
Subscribe to Near Zero → to catch the first post going out in about a week!
関連記事
今日のまとめ
AI日報で今日の重要ニュースをまとめ読み