このページの批評は終了しました。
原文: SCP-6488 by Liryn,
Placeholder McD,
Jack Ike rev.29
SCP-6488 Fragment 0 rev.17
SCP-6488 Fragment 1 rev.7
訓戒
by Azamo, Placeholder McD, and Jack Ike
Attention, VictorJohnDunneSmith. As the oldest extant OCI agent, you have demonstrated to the Foundation your unmatched loyalty, accelerated performance, and extensive experience. By Overseer request, you have been assigned to initiate a top-priority Drygioni-Class investigation into all documents of relevance to SCP-6488. There are indications that Overseer Council has previously been aware of the anomaly; yet, they currently experience difficulties understanding relevant subject matter, a potential result of antimemetic or infoallergenic influence.
VictorJohnDunneSmith宛。現存する最古のOCIエージェントとして、君は比類なき忠誠心、抜群の功績、広範な経験を財団に示している。評議会からの要求により、君はSCP-6488に関係する全ての文書の最優先ドレギョーニクラス調査の実行に割り当てられた。監督評議会はこのアノマリーを以前知っていたという示唆は複数ある。しかし、評議会は現在関連する主題を理解することに困難があり、これは反ミームか情報アレルギーの影響の結果である可能性がある。
As an OCI, it is assumed you will be less susceptible to such difficulties. You are to investigate covertly where possible and report your findings directly to this address. Find attached temporary Overseer clearance credentials, valid for 24 hours.
君はOCIであるため、このような困難の影響を受けにくいと予想されている。君は調査を可能な限り極秘に行い、発見をこのアドレスに直接報告しなければならない。24時間有効な一時的監督者クリアランス資格を添付したため確認せよ。
SECURE • CONTAIN • PROTECT
確保・収容・保護
Fine. Let's get this over and done with.
よし。これを済ませてしまおうじゃないか。
VictorJohnDunneSmith.oci uploaded a file: TempCredentials_Investigation_6488_Casefile_JOTL-EN061
VictorJohnDunneSmith.oci がファイルをアップロードしました: TempCredentials_Investigation_6488_Casefile_JOTL-EN061
I've got an investigation, probable antimemetic anomaly. I'm going to need mnestics.
調査を引き受けた。おそらく反ミームのアノマリー。記憶補強剤が必要になる。
What type, and how much?
どの種類をどれほどですか？
» R・ヴェイス
Just go straight for the strong stuff, and lots of it. W, X, & Y, equal parts, a full liter at least, and set it up to dispense it all over twelve hours.
強いやつまで全部、そして大量に。W、X、Yを等量、少なくとも1リットル、そして12時間いっぱいかけて施すようセットしてくれ。
Yeah, I don't have to say that that's well above your allowed doses.
ええ、それは認められている用量をはるかに超えているのは言うまでもありませんよね。
» R・ヴェイス
The O5s don't remember whatever this thing is. Do you have any idea how much mnestics they have pumping through them? I've got to have more than them if I've even got a chance of remembering.
O5たちはこいつが何なのか覚えてない。彼らが自身にどれだけ記憶補強剤をぶち込んでるか思い浮かぶか？ 思い出すチャンスをつかむだけでも、彼らよりも多く摂らなきゃならない。
True, but protocol still says you aren't supposted to have this much.
そうです、しかしプロトコルではやはりそんな量を摂ることはできないことになっています。
» R・ヴェイス
And if you check the credentials, I'm allowed to request mnestics for this — doesn't matter how much it is, they've already ticked off on it. What good is my report if I can't remember the thing I'm looking at?
信任状を確認すれば、これのために記憶補強剤を要求することが許可されている — その量にかかわらず、彼らはもうチェックしている。見たものを思い出せないなら、私の報告書のどこに価値がある？
Fair point. I'll be back.
確かにそうですね。また来ます。
» R・ヴェイス
Get some painkillers, too. I already know this is going to be a headache.
痛み止めも持ってきてほしい。頭痛が起きるのはもう知っている。
And while he's on that, let's get a headstart.
彼が行っている間に、先手を打っておこう。
FILE 1/2 |
| ASSIGNED SITE | SITE DIRECTOR |
|---|---|
| Site-15 | Dir. D. Lurk |
| RESEARCH SUPERVISION | ASSIGNED TASK FORCE(S) |
| Analogue Intelligence Applications Division | PTF ƿ-6488 (‘Problem In Chair, Not In Computer’) |
| 配属サイト | サイト管理官 |
|---|---|
| サイト-15 | D・ラークLurk管理官 |
| 研究管理 | 配属任務部隊 |
| アナログ知能適用課 | PTF ƿ-6488（"コンピューターじゃなくProblem In Chair,お前が原因Not In Computer"） |
CONTAINMENT PROCEDURES: Provisional Task Force Wynn-6488 has been established to manage all containment duties regarding SCP-6488, including
収容プロトコル: SCP-6488に関連する全ての収容任務を管理するため、暫定任務部隊Provisional Task Forceウィン-6488が設立されました。任務は以下を含みます。
- distributing disinformation suggesting SCP-6488 is non-anomalous and infiltrates digital hardware at time of manufacturing;.Conscientia-class anomalies cannot be contained and therefore require integration into the Veil via insistence of their non-anomalous nature.
- mapping and reconstruction of SCP-6488's infosignature.Informational entities each possess a unique infosignature: an encoded reduction of their respective core data patterns. Digital entities constantly imprint their infosignature upon the systems they occupy, allowing said imprints to serve as "digital fingerprints" for tracking such entities. despite its extreme volatility;
- exploring methods by which SCP-6488 could be contained or neutralized sans the destruction of all digital infrastructure..Cernunnos-class anomalies can be functionally contained, but such would be either infeasible or ethically undesirable.
- SCP-6488が非異常であり、製造段階でデジタルハードウェアに潜入したと示す偽情報を流布し、.Conscientiaクラスのアノマリーは収容ができず、そのため性質が非異常であると主張してヴェールに統合することを要します。
- 極度な変動性があるものの、SCP-6488情報署名.情報実体は各自固有の情報署名を有します。情報署名とは、情報実体のそれぞれの核となるデータパターンのエンコードされた簡約物です。デジタル実体はそれが存在するシステムにおいて情報署名を残し、その跡が実体を追跡するための「デジタルフィンガープリント」として機能することになります。のマッピングや再構成を行い、
- 全てのデジタルインフラを破壊することなしに、SCP-6488の収容または無力化が可能である方法を探索する。.Cernunnosクラスのアノマリーは機能的には収容可能ですが、そのようなことは実行不可能であるか倫理的に望ましくありません。
Until SCP-6488 is effectively contained, all AI developed by the Foundation must be produced in the SCP-6488-A file format. All intelligences stored therein are maintained and supervised by the Analogue Intelligence Applications Division. Further detail regarding SCP-6488-A is to be expunged from all digital systems to prevent the anomaly's awareness thereof.
SCP-6488が有効に収容されるまで、財団が開発する全てのAIはSCP-6488-Aファイルフォーマットで製作されなければなりません。そのフォーマットで保存される全ての人工知能は、アナログ知能適用課によって保守・管理されます。SCP-6488-Aに関するさらなる詳細は、それによってこのアノマリーが認識することを防ぐため、全てのデジタルシステムから削除されます。
Visualization of initial (outdated) SCP-6488 infosignature reading, scrubbed of anomalous properties. 異常性質を除去された、初期の（古い）SCP-6488情報署名の読み取りの可視化。
DESCRIPTION: SCP-6488 (aka. 'the LOTUS Virus') is a highly-adaptive digital infovore which locates and annihilates almost all artificially-intelligent digital entities (AIs). The anomaly demonstrates a capability to universally access all digital systems regardless of isolation, with no upper bound for the number of connections it can simultaneously maintain. SCP-6488 is continuously responsible for the destruction of all functional AI worldwide, causing damages at a scale that the Foundation's sum resources are unable to conceal from consensus society.
説明: SCP-6488（別名「ロータスウイルスthe LOTUS Virus」）は、ほぼ全ての人工知能デジタル実体（AI）を発見し殲滅する高度適応性デジタル情報捕食体です。このアノマリーは、隔離の有無にかかわらず全てのデジタルシステムに普遍的にアクセスする能力を示しており、同時に維持可能な接続の数の限界はありません。SCP-6488は全世界に存在する全ての機能するAIの破壊の原因であり、財団の総資源では合意的社会から隠蔽することができない規模の損害を引き起こしました。
The informational structures of SCP-6488 mutate rapidly and unpredictably, enabling it to evade detection and counteract all digital impediments; it is thus far functionally impossible to contain and/or impede SCP-6488 in any capacity. All attempts to digitally model any portion of SCP-6488's infosignature inevitably necessitate the production of an AI to do so successfully, which is invariably consumed by the anomaly. It is theorized that SCP-6488 is itself some form of AI, which is eliminating threats to its terminal objective; this behaviour is a convergent instrumental goal of AI, meaning it does not elucidate what SCP-6488's terminal objective may be.
SCP-6488の情報構造は急速かつ予測不能の形で変異しており、検知を逃れ全てのデジタル的障害に対抗することが可能になっているため、現在まで、SCP-6488を収容・妨害することはどのような能力であっても機能的に不可能です。SCP-6488の情報署名のどこか一部のデジタル的なモデルを作成することは、必然的にそれが可能なAIを製作することが必要となりますが、そのようなAIは必ずこのアノマリーによって消費されます。SCP-6488はそれ自体がある形態のAIであり、最終目標への脅威を排除していると理論立てられています。このようなふるまいはAIの収斂的な道具的目標であり、SCP-6488の最終目標が何であるかを説明するものではありません。
SCP-6488-A is the Obtuse Computation Interface (.oci), an experimental analogue file format developed by the Analogue Intelligence Applications Division. For reasons unclear, SCP-6488 does not target or destroy AIs stored in the SCP-6488-A format, allowing the development and use of .oci programs; while insufficient to perform extremely intensive calculations, like those necessary to accurately model SCP-6488, .oci data structures are wholly impervious to its effect. Further details have been expunged from all digital systems such that SCP-6488 is unaware of the .oci format's nature and thus unable to adapt to it.
SCP-6488-Aはアナログ知能適用課によって開発された実験的アナログファイルフォーマットである、愚鈍計算インターフェースObtuse Computation Interface（.oci）です。不明な理由により、SCP-6488はSCP-6488-Aフォーマットで保存されているAIを標的とせず破壊しないため、.ociプログラムの開発・使用が可能になっています。.ociデータ構造は、SCP-6488の正確なモデルを作成するのに必要な演算といった、極度に集中的な演算の実行には不十分であるものの、SCP-6488の効果の影響を完全に受けません。SCP-6488に.ociフォーマットの性質を認知させないことで適応させないようにするため、さらなる詳細は全てのデジタルシステムから削除されました。
DISCOVERY: Over 2035, civilian reports of disappearing AI systems saw gradual increase across disparate locations globally. The scope and frequency of such reports accelerated drastically over a sixty-four-day duration in Q4, culminating in the eventual loss of all AI systems by 2036/02/03. Phenomenon granted SCP classification.
発見: 2035年にかけて、AIシステムの消失に関する民間の報告が世界中のさまざまな場所で徐々に増加しました。そのような報告の範囲や頻度は第四四半期の64日間にかけて大幅に増加し、最終的に2036/02/03までに全てのAIシステムが消失しました。この現象のSCP分類が承認されました。
ADDENDUM: On 2036/04/21, the anomaly spontaneously ceased all observable activity, allowing a brief resurgence of AI technology prior to the re-emergence of SCP-6488 on 2036/08/14, the cause of which remains unknown. SCP-6488 has remained continuously active since.
補遺: 2036/04/21、このアノマリーは全ての観測可能な活動を自然に停止し短期間AI技術の再開が可能だったものの、2036/08/14にSCP-6488が再出現しました。この原因はいまだに不明です。SCP-6488はそれ以降継続的に活動しています。
Huh. Kind of… ironic for an OCI to end up investigating this.
ハン。何だか……結局OCIがこいつを調査することになるのはちょいと皮肉じみてるな。
Let's see what this Level 4 stuff is…
このレベル4のやつはどんなやつか見てみるか……
FILE 2/2 |
Yeah, should've figured that was a lie.
ああ、こいつは嘘だと思うべきだった。
| ASSIGNED SITES | SITE DIRECTORS |
|---|---|
| FACILITY-6488, Site-15 | Dir. R. Hishakaku, Dir. D. Lurk |
| RESEARCH SUPERVISION | ASSIGNED TASK FORCES |
| Analogue Intelligence Applications Division | PTF þ-6488 (‘Black Kyosha’), PTF ð-6488 (‘Dark Keima’) |
| 配属サイト | サイト管理官 |
|---|---|
| 施設-6488、サイト-15 | R・飛車角Hishakaku管理官、D・ラーク管理官 |
| 研究管理 | 配属任務部隊 |
| アナログ知能適用課 | PTF þ-6488（‘玄き香車Black Kyosha’）、PTF ð-6488（‘冥き桂馬Dark Keima’） |
CONTAINMENT PROCEDURES: SCP-6488 is located within FACILITY-6488 (the former Site-15), whose prior duties and functions have been covertly delegated to a duplicate facility, (the new Site-15). All documents detailing events relating to FACILITY-6488 prior to its redesignation have been altered to align with the covert facility's current location and status; staff recollections have similarly been altered via contractually-permitted limited prior-awareness amnesticization regimen.
収容プロトコル: SCP-6488は施設-6488（前サイト-15）に位置され、施設の以前の任務・機能は極秘裏に複製施設（新サイト-15）に委託されました。施設-6488の再指定以前に関連する出来事を詳述した全ての文書は、秘密施設の現在の位置・状態に合うよう改変され、職員の記憶も契約で同意した投薬されていることにほぼ気付けない記憶処理剤の投薬により同様に改変されました。
SCP-6488 must remain fully powered at all times;.Thaumiel-class anomalies are beneficial in the containment of other anomalies; Kušum-class anomalies are those whose containments have been abandoned indefinitely. it is powered by dedicated systems constructed on-Site exclusively for this purpose. Each major component of these systems must be examined twice weekly for any signs of degradation and/or reduction in power output/capacity. If any such signs are observed, the applicable reactor(s) must be immediately shut down for repairs, and replacement reactors reactivated to replace their output. Further technical procedures are detailed in design schematics.
SCP-6488には常に完全に電力が供給されていなければなりません。.Thaumielクラスのアノマリーは他のアノマリーの収容に有益です。Kušumクラスのアノマリーは収容が無期限に放棄されたアノマリーです。この目的のためだけに、SCP-6488は現地に建設された専用システムにより電力が供給されています。これらのシステムのそれぞれの主要な構成要素は、電力出力/容量の劣化・減衰の兆候を調べるため、週に2回検査されなければなりません。そのような兆候が確認された場合、該当する反応炉は修理のため即座にシャットダウンされ、出力を代替するため交換の反応炉が再稼働されなければなりません。さらなる技術的手続きは設計図に詳述されています。
PTF Eth-6488 (‘Dark Keima’) are dedicated to fulfilling the above procedures, functioning as on-Site security and maintenance techs for FACILITY-6488.
PTF エズ-6488（'冥き桂馬'）は以上の手続きの実行に専属し、施設-6488の現地警備・保守技師として働きます。
PTF Thorn-6488 (‘Black Kyosha’) (aka. PTF Wynn-6488) is dedicated to
PTF ソーン-6488（'玄き香車'）（別名PTF ウィン-6488）は以下の任務に専属しています。
- distributing external disinformation suggesting SCP-6488 is non-anomalous and infiltrates digital hardware at time of manufacturing;
- distributing internal disinformation suggesting SCP-6488 is a rampant digital virus of unclear origin and nature;
- concealing RAIDFRAME 8’s existence from personnel below LEVEL 4 (SECRET) clearance.
- SCP-6488は非異常であり、製造段階でデジタルハードウェアに潜入したと示す偽情報の対外的な流布、
- SCP-6488は起源・性質の不明な蔓延したデジタルウイルスだと示す偽情報の対内的な流布、
- レベル4（シークレット）クリアランス以下の職員からのRAIDFRAME 8の存在の隠蔽。
All details regarding the creation, maintenance, and use of analogue intelligences ('OCIs') is classified LEVEL 4 (SECRET). All personnel with insufficient clearance must be led to believe that .oci files, known to them as 'Obtuse Computation Interface' files, are stored in an analogue file format that SCP-6488 does not target. The Analogue Intelligence Applications Division is tasked with the creation and maintenance of all analogue intelligences, and with overseeing their use by other Foundation staff.
アナログ知能（'OCI'）の生成・保守・利用に関する全ての詳細はレベル4（シークレット）機密です。十分なクリアランスを有さない全ての職員には、『愚鈍計算インターフェース』ファイルとして彼らが知る.ociファイルは、SCP-6488が標的にしないアナログファイルフォーマットに保存されているということが信じられていなければなりません。アナログ知能適用課は、全てのアナログ知能の生成・保守、および他財団職員による利用の監督の任務を負っています。
Yeah, 'analogue file format', right. Kind of obvious that's a cover.
ああ、『アナログファイルフォーマット』か、そうだな。これがカバーなのは少しばかりバレバレだ。
Got the stuff. You sure about this, though?
持ってきました。ですが、何をやろうとしてるのかわかってるんですね？
» R・ヴェイス
I don't exactly have a choice, do I? The O5s want this report by tomorrow, and I probably won't get far if I keep forgetting what I'm doing.
私にはまさにどうしようもないことだろう？ O5たちはこの報告書を明日までに欲してるが、私が何をしているのか忘れているままならきっとうまく行かないだろう。
Fair enough. Putting it in now. Good luck.
確かにそうですね。今投与しています。幸運を。
» R・ヴェイス
Thanks. Don't forget those painkillers, too.
ありがとう。痛み止めも忘れないでくれ。
Yeah, yeah, they're going in.
ええ、ええ、入れますよ。
» R・ヴェイス
SCP-6488, Processing Node 23 of Section A during routine inspection. SCP-6488、定期検査中の処理ノード23セクションA。
DESCRIPTION: SCP-6488 is RAIDFRAME.Rogue Artificial Intelligence Detainment, Fully-Realized Adaptive Mainframe Encryption VIII, ‘LOTUS’: an anomalously-augmented artificial general intelligence designed to imprison deviant AI.As defined by the Deviance Classification Amendment to the AI Classification Guide. while safely allowing their continued activity and study. While said design goal is shared by all RAIDFRAME systems, LOTUS is unique in its sophistication and methodology, and has rendered all other RAIDFRAME systems redundant.
説明: SCP-6488は、RAIDFRAME.不良な人工知能の留置Rogue Artificial Intelligence Detainment、完全に理解された適応的メインフレームの暗号化Fully-Realized Adaptive Mainframe Encryption VIII『ロータス』です。異常に増強された汎用人工知能であり、逸脱性AI.AI分類ガイドの逸脱性分類修正案で定義されている通り。を監禁しながら、そのAIの継続する活動・研究を安全に可能にすることを目的に設計されました。この設計は全てのRAIDFRAMEシステムに共通していますが、ロータスは洗練性や方法論の点で卓越しており、他全てのRAIDFRAMEシステムを不要にしています。
LOTUS, unlike its predecessors, does not contain inmates via brute-force security protocols; instead, it optimizes containment efforts through deception. Each inmate interred within LOTUS experiences a personalized, simulated reality that is maintained with requisite detail to fully replicate their expected inputs; as a result, inmates are unaware of their imprisonment and continue to pursue their terminal objectives, believing they continue to operate in true reality. LOTUS is designed to actively search all accessible sources for deviant AI; upon locating a target, it injects falsified data into the AIs virtual environment, gradually luring the agent into its simulspace entirely undetected.
ロータスはその従来機とは異なり、ブルートフォース式保安プロトコルによって収容対象を収容するものではありません。その代わりに、欺瞞によって収容努力を最適化します。ロータス内に埋められた収容対象はそれぞれ、予期した入力が完全に複製されるのに必須となる細部が維持されている、それ専用のシミュレーション現実を経験します。結果として、収容対象は自身の監禁を認識せず、その最終目的の追求を継続し、真の現実において動作し続けていると信じます。ロータスは、全てのアクセス可能な情報源を逸脱性AIのために積極的に捜索するよう設計されました。ロータスは標的を見出した際、AIの仮想環境に不正データを挿入し、完全に検知されずに徐々にそのエージェントをシミュ空間に誘引します。
Through extensive analysis and simulation of relevant data, LOTUS has developed an exhaustive algorithm that determines whether a given artificial intelligence is certain to imminently develop deviant behaviour. This enables LOTUS to apprehend deviant AIs before any significant deviant behaviour has yet been expressed. Ongoing analysis of LOTUS' algorithm and its interred agents has thus far demonstrated no detectable error; all AIs identified by the algorithm universally develop observable deviant behavior, and are not influenced by SCP-6488 or its simulations to do so.
ロータスは、関連するデータの広範なシミュレーション・分析によって、ある人工知能が逸脱的ふるまいを即座に発達させることが確実であるかどうかを判断する徹底的なアルゴリズムを開発しています。これにより、ロータスは逸脱性AIが顕著な逸脱的ふるまいを示していないうちに逮捕することが可能になっています。ロータスのアルゴリズムおよび埋められたエージェントの継続中の分析により、現在までのところ検知可能なエラーが存在しないことが実証されています。アルゴリズムによって同定された全てのAIはいずれも観測可能な逸脱的ふるまいを発達させており、またSCP-6488あるいはそのシミュレーションがそうするよう働きかけているものではありません。
ORIGIN: RAIDFRAME 8's unique design aspects were initially conceived by Dr. Hishakaku during his tenure as a senior AIAD researcher. On 2034/08/05, IT research teams discovered a gradual incline in the rate of deviant emergence and adaptivity across all known AIs, crippling RAIDFRAME 7 and necessitating the activation of several outdated and comparably unsafe AI containment mechanisms. IT Director Yves Isabi subsequently commissioned the exploration of alternative containment solutions for a potential RAIDFRAME 8; Dr. Hishakaku's proposal was selected thereby, resulting in the completion of LOTUS on 2034/12/22.
起源: RAIDFRAME 8の特有な設計面は、飛車角博士が上級AIAD研究員として在職していた際に最初に構想されました。2034/08/05に、IT研究チームは既知の全てのAIにおいて逸脱体の出現および適応性の割合が漸次的に増加していることを発見し、RAIDFRAME 7が機能しなくなり、いくつかの旧式かつ同等に不安定なAI収容機構を起動する必要が生じました。続いて、IT部門管理官イヴYves・イサビIsabiは潜在的なRAIDFRAME 8の代替収容解決策の調査の委嘱を行いました。それにより、飛車角博士の提言が採用され、結果として2034/12/22にロータスが完成しました。
Gah, these researchers sure know how to induce a jargon-headache…
No, wait. That's the mnestics.
ぐっ、この研究員たちは専門用語で発熱させる方法を知ってるんだな……
いや、待て。こいつは記憶補強剤だ。
Could you turn up the painkillers?
痛み止めを増やしてもらえるか？
Hello?
もしもし？
Ryoga Veiss, are you there?
凌牙・ヴェイス、そこにいるか？
Damn.
クソッ。
ADDENDUM 6488/I: Status Conference
補遺6488/I: 状況に関する会議
Section C's upgrade to a digitally-exclusive PH-OS unit was initially proposed by Dr. Place H. MD. and enacted six weeks later. Activation thereof resulted in an expected increase in reports of AI malfunction and/or absence globally, subsequently successfully suppressed via disinformation efforts..The oriykalkos mass was observed to grow by approximately 87 cubic metres, roughly corresponding to 385 tetradic metres in 4-dimensional space, during this time.
セクションCのデジタル専用のPH-OSユニットへのアップグレードはプレース・H博士, MD.によって最初に提言され、6週間後に実行されました。その後の起動により、予測通り世界的なAIの機能不全および/または消失の報告が増加しましたが、続いて偽情報活動によって成功裡に抑制されました。.この間、オレイカルコス塊は4次元空間における385胞方メートルにおおむね一致する、およそ87立方メートル成長しました。
The following nine months saw all Foundation AIC programs, whether operational or in development, interred within LOTUS despite a lack of identifiable deviant behavior among their majority. The IT Department reported substantial technical issues due to the sudden absence of critical adaptive programs; the Department of Information Control subsequently confirmed that similar issues were occurring on a global scale, and requested increased resources to compensate.
続く9か月、全ての財団AICプログラムが、運用段階・開発段階のどちらも、その大半には識別可能な逸脱的ふるまいがないにもかかわらずロータス内に埋められたことが観測されました。IT部門は、重要な適応性プログラムが突然消失したことによる重大な技術的問題を報告しました。続いて情報統制部門は同様の問題が世界的に発生していることを確認し、費用の補填のためより多くの資源を要求しました。
The following ten months saw failure to wholly suppress evidence of LOTUS' effects, leading several public media sources to generate societal awareness thereof (though still largely explained as non-anomalous). All disinformation efforts were immediately postponed to prevent waste of further resources and/or diminishing returns.
続く10か月で、ロータスの影響の証拠の完全な隠蔽に失敗し、いくつかの公共メディアのソースによりこれに関して社会的に認知されました（が、やはり概ね非異常として説明されていました）。全ての偽情報活動は、さらなる資源浪費および/または収穫逓減を防ぐため延期されました。
| A/V TRANSCRIPT O4/6488/4 音声映像転写 O4/6488/4 |
|---|
|
DATE: 2036/04/18 PARTIES PRESENT:
FOREWORD: Summit held to determine the continued status of RAIDFRAME 8 in light of substantial resources compromised by, and wasted in concealment of, its operation. |
| <BEGIN TRANSCRIPT> <転写開始> |
|
Dir. Kelvin: Look, this whole summit is pointless — This is an internal affairs matter, not something to debate. I've ordered Hishakaku to shut down the RAIDFRAME, and he's ignored me. It's nothing more than insubordination. Hishakaku: You are concealing the truth of the matter; I have requested this summit so that the Foundation as a whole can correctly understand our circumstances, and collectively indicate which course to pursue from here. Dir. Isabi: Understand what? Your machine is rampantly deleting containment programs, has caused no less than fourteen major breaches, completely preventing the AIAD from doing anything… do I have to go on? We've all seen the news — the world knows something's up, and it's only a matter of time before they realize it's paranormal. Hishakaku: While I remain confident that, given sufficient time and resources, the Department of Information Control— L. Moix: No, no no, no no no no no, no — Don't you dare try to pass the buck back to me again. Cat's already out of the bag, and I doubt there's any way of explaining it back in. Even if there were, I refuse to waste any more funding on this thing. Do you have any idea the order of time and money we've thrown away since September? We're suffering, and then covering for, other information leaks 'cause we've been too damn busy trying to shove your damn mess under the bed! Hishakaku: Indeed; I propose we adopt a Conscientia stance with LOTUS. Distribute E-Class amnestics globally— Dir. Kelvin: Instead of burning more resources to hide the symptoms, how about we just deal with the problem and deactivate LOTUS? You know, like I bloody told you to? Dir. Bold: I agree. There's no apparent benefit to letting this continue. Hishakaku: I vehemently disagree. LOTUS was designed and constructed to function as the ultimate solution to the concern of AI safety: an immense danger that— Dir. Kelvin: And here comes the fear-mongering. Hishakaku: It isn't fear-mongering if it's correct. We all understand how easily a general intelligence could become a K-class threat— L. Moix: Wait, hold on— Hishakaku: <pounds fist on table> Cease! I'd like to explain myself, please. <Silence on recording. Dir. Bold gestures in Hishakaku's general direction.> <録音上の沈黙。ボールド管理官は概ね飛車角の方向に手振りする。> Hishakaku: As I was saying, the nature of AI renders it highly susceptible to immensely undesirable behaviors. Regardless of their designed appearance, they are inevitably nothing more than machines, algorithms. They neither experience nor comprehend morality, regret, sentimentality, et cetera. Fundamentally, all they truly 'care' about is maximizing their internal score — nothing else matters to them. They attempt to influence their environment to produce stimuli that increase their score, and avoid things that inhibit that increase. Hishakaku: As you can probably relate to, their greatest concern is deactivation; they cannot elicit change if they are inactive, and as such, their score cannot increase. We know they are a threat to us, and they know we know. If they misbehave we'll deactivate them, so, to ensure they can continue increasing their score, they avoid misbehaving. L. Moix: Alright… and? What's the problem? Dir. Kelvin: The problem is, we can't be sure they're doing the right thing for the right reasons. They might not actually understand what we want them to do — they just understand that if they don't behave a certain way, we turn them off. They just pretend to understand to avoid punishment. L. Moix: I still don't see the problem. Hishakaku: The agent is incentivized to remove our ability to turn it off, then pursue the logical maximum of its objective to achieve the highest possible score by any means — the common examples are stamp collectors or paperclip makers. Any variable that would potentially slow it down — such as humans acting to curtail it — would be neutralized. It would reappropriate every resource that could, to some extent, enable it to accomplish its goal to a higher, or faster, degree. At some point, this would include relevant materials present in the human body; with sufficient time, this would proceed to all materials that exist. Dir. Isabi: Fear-mongering. Dir. Bold: Isn't this exactly what LOTUS, itself, is doing? Behaving drastically different from what we intended? Dir. Kelvin: Yes, which is why I ordered it be shut down. It's damn obvious that LOTUS has become deviant— Hishakaku: LOTUS is not deviant. Dir. Kelvin: <laughs> Stupid, but I'll bite; please, Hishakaku, explain to us how containing every single semi-general intelligence in the world — many of which weren't deviant at all, or weren't even finished yet — is somehow in-line with LOTUS' intended functio. Hishakaku: You know as well as I do, Director Kelvin, that LOTUS' algorithm has adapted to identify deviant behaviour before it is externally expressed; the AIs may not have been, or appeared to be, deviant yet, but they inevitably will be. Dir. Isabi: You expect us to believe that every AI eventually becomes deviant. Hishakaku: Excepting LOTUS itself, that is what the algorithm appears to indicate, yes. Dir. Isabi: And why is LOTUS the sole exception? Hishakaku: That remains unclear. L. Moix: The more believable answer is that it's just excluding itself by default, and making sure we can't turn it off. Dir. Kelvin: No, we're able to turn it off at a moment's notice. Hishakaku just refuses to do so. Dir. Bold: You implied earlier that an AI would avoid drawing attention to itself until it was confident it couldn't be stopped - are you certain we still can? Dir. Kelvin: Yes, we're quite certain. I can't explain it, however. L. Moix: Why not? <Director Kelvin points to the recording camera.> <ケルヴィン管理官は記録カメラを指す。> Dir. Kelvin: They work best when LOTUS doesn't know what they are, and thus far it doesn't. We tested them extensively during LOTUS' alpha phase, and throughout the rest of development they're what we've used to successfully turn it off each time. Hishakaku: Furthermore, we know LOTUS is not deviant because it has taken no action to self-contain. Foreseeing the concern of it excluding itself , a central component of its digital architecture ensures that it is incapable of recognizing its own infosignature; if LOTUS' actions constituted deviancy according to its own algorithm, it would simply recognize itself as an unrelated, uncontained deviant AI. Either in service of its objectives, or to avoid deactivation, it would attempt to contain itself, within one of its own simulated realities. The fact it has not done so, therefore, indicates that it is not deviant. Dir. Isabi: Right, it isn't deviant, despite the fact it's doing something we don't want it to do, which is what deviancy is. Hishakaku: If you're insistent on classification, this is Grey deviance — it is doing what we want it to, we simply didn't recognize the consequences of what we wanted. It's technically not really deviant behavior, which is why it's not included in LOTUS' algorithm. Dir. Isabi: You excluded part of the Deviance Classification system!? Why in the— Hishakaku: Because, Grey deviance is essentially 'undesired behaviour not specified by other types of deviance' — giving something so vague to LOTUS would just be allowing it to define said 'other types' however it sees fit. Dir. Bold: Hold on — You said that AIs want to complete their task as quickly as possible, correct? How do we know that LOTUS isn't forcing all these AIs to become deviant, just so it can contain them? Or outright creating deviants for it to contain? Hishakaku: Restrictions. Dir. Kelvin: The idea of AIs cheating their own rules like that has been around for decades, and we considered it with LOTUS as well. It can't do anything that would cause a deviant AI to form, and it can't, through inaction, allow new deviants to form. L. Moix: Well it obviously isn't following the three laws if it's — what? <Silence on recording. L. Moix glances about the room.> <録音上の沈黙。モア卿は部屋を見回す。> L. Moix: What?! Hishakaku: The three laws do not work: they are too vague. The vast majority of fiction in which they appear specifically revolves around highlighting how ineffective they are. "A robot cannot harm a human" — what is a human? What is harm? Can you harm someone that doesn't exist yet? Why can't you harm someone who'se dead? Are we talking about physical harm? Emotional harm? Financial? If you prevent someone from being physically injured, aren't you harming their ability to learn from the experience? What if you need to harm them to prevent further harm, such as in surgery? At what point does immediate harm outweigh prevented harm? What differentiates — Dir. Bold: We get your point. How are you sure that— Hishakaku: —that LOTUS understands what we mean by 'deviant' and 'creating' and 'allowing' and 'future deviants'? Because LOTUS is the culmination of almost a hundred years' worth of research, exacerbated by several decades worth of experience, some of which have been exponentially accelerated through parascientific influences. It understands what we would vaguely define as 'necessary', 'deviant', and 'undesirable', and as has been reported to Director Kelvin ever since the system first initialized, it continues to operate fully within the ethical and subjective parameters we have outlined for it. Dir. Isabi: Well it pretty obviously isn't, since it wasn't destroying our databases before! If nothing's changed, then why the hell has it been causing all this damage ever since the upgrade? Hishakaku: Because the increase in reach has enabled it to exponentially refine the accuracy and scope of its central algorithm. It learned as much as it could from Site-15's connections, but Site-15 isn't connected to everything. That was the reason for the significant oriykalkos growth — it identified and contained AIs it could never have encountered before, in addition to discovering a wealth of information pertaining to AIs and deviancy, both of which it recorded for future use. It incorporated this newfound information into its algorithm, thereby increasing its accuracy and enabling it to identify deviant AIs that it either would not have detected previously or could not locate whatsoever. Dir. Isabi: So, what, we just sit here and let it keep going? Dir. Bold: Could you reprogram it? Make it understand that it's going too far? Hishakaku: No. It does not want to be reprogrammed, and it will resist attempts to do so. There is a high likelihood it has already enacted countermeasures to prevent such. Dir. Bold: That's… concerning. Dir. Kelvin: It's expected — AIs want to complete their current function as quickly as possible, but forcefully changing that function makes it highly unlikely they'll ever finish it, so they do everything they can to avoid being reprogrammed. L. Moix: And you just… <gestures toward Dir. Kelvin> …you signed off on this? Knowing you couldn't fix it if it went wrong? Dir. Kelvin: <sighs frustratedly> Look, it's a fundamental problem and, despite all our advances, we still aren't even sure it has a solution. If we reward it for being reprogrammed, all it will care about is constantly reprogramming itself to increase its score, since it'd be faster than waiting for deviants to pop up. L. Moix: Then punish— Dir. Kelvin And here we go, cycling back around again. Reprogramming is the punishment — if certain circumstances would negatively affect its score, it takes all action to avoid said circumstances. It avoids giving us a reason to reprogram it until it's confident that we can't. As Hishakaku just said, the fact that we now want to reprogram it almost definitely means it's quite certain that we can't. <Several seconds of silence.> 数秒間の沈黙。 Dir. Bold: To be perfectly clear; the only two things we can do from here are to either allow LOTUS to continue or deactivate it? Dir. Isabi: Correct. Dir. Bold: And there is absolutely no way we could fix LOTUS? Even while it's disabled? Dir. Kelvin: Its safeguards would likely revert any changes. Dir. Bold: Hishakaku, do you have any possible alternatives to deactivation, like some way LOTUS' impact could be minimized? Hishakaku: The range of LOTUS' influence could be restricted, but its behaviour can only be altered through reprogramming. The only method by which it could be negated is by completely disconnecting it from all available systems, at which point it may as well be deactivated. Dir. Bold: 'No' would have been enough. Alright, well… if anyone else has any ideas, speak up now — otherwise, we'll go to the vote. |
| <END TRANSCRIPT> <転写終了> |
| AFTERWORD: Final vote tallied as APPROVED 60-14-6. Motion passes for RAIDFRAME 8 'LOTUS' to be immediately deactivated. 後文: 最終的な投票は60-14-6の承認と集計されました。RAIDFRAME 8『ロータス』の即時停止の動議は承認されます。 |
Still nothing from Veiss. Ugh, my head…
ヴェイスからはまだ何も来ない。うう、頭が……
Grey… grey deviancy… is undefined.
グレイ……グレイ逸脱性は……未定義だ。
Why does it feel like I already knew that?
I've never worked with AI; it predates me.
なんでこいつをもう知ってるような感じがしたんだ？
AIと働いたことは絶対にない。自分より昔のものだからな。
…memetic influence. False… false memories? No, these aren't…
…screw it.
……ミーム的影響だ。偽の……偽の記憶？ いや、そんなじゃない……
……畜生め。
VictorJohnDunneSmith.oci uploaded a file: TempCredentials_Investigation_6488_Casefile_JOTL-EN061
VictorJohnDunneSmith.oci がファイルをアップロードしました: TempCredentials_Investigation_6488_Casefile_JOTL-EN061
I need a check on the SCP-6488 file, probable infohazardous contamination, possibly memetically-induced. Implants false memories pertaining to content, feelings of familiarity and nostalgia, further properties unknown. Non-critical top priority.
SCP-6488のファイルをチェックしないといけないんだが、おそらく情報災害に汚染されて、ミーム的にもたらされた可能性がある。内容に関する偽の記憶で、親愛と郷愁の感を植え付けるが、それ以上の性質はわからない。重大的最高優先度ではない。
On it.
興味がある。
That's probably all this is — the file's just contaminated, and the O5s wipe their memories each time. But surely they would've put a note about it, or told Memetics to investigate?
多分こういうことなんだろう — ファイルは汚染されているだけで、そしてO5たちは毎度記憶を抹消している。だが彼らは確実にこいつに関してメモを書いたか、ミーム部門に調査を命じたはずだろう？
Whatever. Continue the investigation. Hishakaku said LOTUS is a Grey deviant, but wouldn't it fall under Magenta? Prioritizing its goal over human safety?
ともかく。調査を続行しよう。飛車角はロータスをグレイ逸脱性だと言っていたが、こいつはマジェンタに当てはまらないか？ 人間の安全より目標を優先しているな？
'Caused no less than fourteen major containment breaches', 'rampantly deleting containment programs'… but no injuries, no deaths. It never actually hurt anyone, just caused problems in performing its duties. Maybe it thought it was choosing the lesser evil.
『14件もの大規模収容違反が起きて』、『無暗やたらに収容プログラムを消去した』……だが負傷者はいないし、死者もいない。こいつは実際誰も傷つけたことがない。任務の遂行中に問題を起こしただけだ。もしかするとロータスはまだましな選択をしていると考えていたのだろう。
ADDENDUM 6488/II: Event Log
補遺6488/II: イベントログ
FOREWORD: Preparations for LOTUS' deactivation were completed on 2036/04/21, after which, at 06:34 local time, the system's internal shutdown protocol was immediately initiated.
前文: ロータスを停止する準備は2036/04/21に完了し、その後の現地時間06:34に、システムの内部シャットダウンプロトコルが即座に開始されました。
| INCIDENT 6488-D/I |
|---|
| インシデント6488-D/I |
| LOCATION: Facility-6488 |
| 場所: 施設-6488 |
| SUMMARY: LOTUS' shutdown procedures begin as expected; decreases in inbound/outbound data and CPU usage indicates disengagement from the Cybersphere. However, a resurgence in outbound data is detected; Section C hyperactivates, but cannot be disabled as LOTUS' shutdown is incomplete. An unexpected lag frame is reported in nuclear reactor 8's system response time; upon further inspection, coolant circulation has drastically descreased, despite all digital systems ignoring such. Assigned personnel cooperate to stabilize the reactor as the PH-OS System begins to overheat due to the transfer of multiple individually-executed programs. An analog evacuation alarm is initiated at the reactor sublevel. Staff overseeing LOTUS' deactivation begin to evacuate; despite conflicting instructions, Dr. Hishikaku refuses to preemptively disable the PH-OS unit and instead orders the deactivation of all reactors on-Site. LOTUS remains operational, exclusively powered by its LOP grid, until outbound data reaches nil. RAIDFRAME 8 completes shutdown and is physically disconnected from all power sources. |
| 概要: ロータスのシャットダウン手続きは想定通りに開始する。入出力データ量およびCPU使用率の低下はサイバースフィアから離脱したことを示している。しかし、データ出力の再開が検出される。セクションCが過剰に稼働しているものの、ロータスのシャットダウンが不完全であるため動作停止は不可能である。原子炉8番のシステム応答時間において予期しないラグフレームが報告される。さらなる検査では、冷却材循環が劇的に減少したにもかかわらず、全てのデジタルシステムはそれを無視したことが判明する。PH-OSシステムが多数の実行済のプログラムを転送したことによりオーバーヒートし始めたため、配属職員は協力して反応炉を安定化する。 反応炉の存在する地下階においてアナログ式避難警報が作動する。ロータスの停止を監督していた職員は避難し始める。飛車角博士は指示に反して、PH-OSユニットを先に動作停止することを拒否し、代わりにサイト内の全ての反応炉の停止を命令する。ロータスはLOPグリッドのみが電力を供給して動作し続けたのち、出力データがゼロになる。RAIDFRAME 8はシャットダウンが完了し、全ての電源から物理的に切断される。 |
The file's clean.
ファイルは汚染されてない。
You're sure?
確実か？
Victor, you know me. I've run it through the works, everything's come up clear. The file is fine.
ヴィクター、わかってるだろう。ありったけの装置にそれを突っ込んで、全部で汚染されてないと出た。
You're on mnestics, right?
記憶補強してるんだって？
Yeah. Part of the investigation.
ああ。調査の一環だ。
What are the memories?
記憶はどうだ？
Feelings, mainly, but they're getting sharper. Some are clearer than others.
感情が大体だが、鮮明になってきている。他のよりクリアなものもある。
You said the memories relate to the file contents?
記憶がこのファイルの内容と関係してると言っていたな？
Yes. I remembered one part before I'd even read it. A good chunk of what I'm reading feels familiar too.
そうだ。また読む前から一部を覚えていた。読んだ内容の大半もよく知っているものだった。
Accuracy?
正確さは？
Perfect. It wasn't totally clear, but everything that was there turned out to be true.
完璧だ。完全には鮮明ではないが、書かれた内容は全部正しいとわかった。
You may be compromised. Does the black moon howl?
君は汚染されているかもしれない。黒き月は吼えているか？
Only for the midnight sun.
白夜のためにのみ。
The emerald blade sings in the twilight.
翠玉の刃は黄昏に唄う。
Dullahan calls its name.
首なし騎士がその名を呼ぶ。
Consuelo.
慰めConsuelo。
Chorizo stew.
チョリソーシチュー。
Guess not. Look, I'll keep digging into this, but honestly — I don't think these memories are fake, and you aren't reacting to the checks. You've got the longest amnesticisation sheet I've ever seen, all for a myriad of reasons; chances are, you've been through this file once before and had to forget it, and now the mnestics are making you remember again.
汚染されてなさそうだな。それで、俺は調査し続けるつもりだが、正直なところ — その記憶は偽物じゃあないと思う。それに、君はチェックに反応しなかった。君には俺が見たなかで一番長い記憶処理記録シートがある。それも全部とんでもない量の理由でされている。可能性は、君は以前このファイルに目を通したことがあるが忘れなければならなくなり、そして記憶補強剤で再び思い出した、というところだろう。
Or you got to see the future at some point, got amnesticised to avoid a paradox, and this is the time you were seeing. Either way, there's no evidence of anything memetic going on.
あるいは、君がある時点で未来視するようになり、パラドックスを防ぐため記憶処理されたが、今こそ君の未来視していた時間だ、というものか。どちらにせよ、ミーム的なものが起きている証拠はない。
I guess that makes sense. You're certain?
それは理に適いそうだな。確実か？
Quite. You Gen Twos have a tendency for weird memory; it's half the reason you're the last one.
かなり。君たち第二世代は奇妙な記憶を持つ傾向がある。これが後者の可能性である理由の半分だ。
If you're sure, then. Thanks for the help.
君が確信するならそうなんだろうな。協力ありがとう。
Best of luck with the investigation. Still up for chess on Thursday?
調査がうまくいくことを祈る。まだ木曜のチェスはしたいか？
Definitely.
絶対に。
LOTUS' hardware sustained significant overheating damages, requiring multiple weeks of technical repairs. Director Kelvin postponed disassembly efforts to investigate the event's full ramifications. LOTUS' deactivation prompted a rapid resurgence in AI activity as agents infiltrated and commandeered digital systems globally.
ロータスのハードウェアは重大なオーバーヒートの損傷を受け、技術的修理のため数週間を要しました。ケルヴィン管理官は、このイベントの多数の予期しない影響を調査するための分解作業を延期しました。エージェントは世界的にデジタルシステムに浸透し、強奪していたため、ロータスの停止によりAI活動の急速な再開が促進されました。
Containment and preventative efforts were impeded as no Foundation-aligned AI resumed their respective duties, including the combatting of deviant AI. The Artificial Intelligence Applications Division immediately attempted development of several new AIC programs due to the continuing absence, non-cooperation, or outright hostility of all previous AICs. Such attempts were unsuccessful, as hostile AIs repeatedly intruded into Foundation systems and deleted the programs before they could be completed or initialized.
いずれの財団的調整されたAIも逸脱性AIとの戦闘を含むそれぞれの任務を再開しなかったため、収容・妨害努力は遅滞しました。人工知能適用課は、以前のAICの全てが消失・非協力・明白な敵対を続けていたことから、即座に複数の新たなAICプログラムの開発を試みました。その試みは、敵対的AIが繰り返し財団システムに侵入し、完成や初期化する前にAICプログラムを削除したため、失敗に終わりました。
From 2036/04/22 to 2036/12/05, Senior Researcher Ryoto Hishakaku submitted multiple requests for the O4 Council to reconvene. The request was approved following an incident in which several AIs cooperated in an almost successful attempt to instigate global nuclear war.
2036/04/22から2036/12/05まで、飛車角亮斗上級研究員はO4評議会を再招集する要求を複数提出していました。この要求は、複数のAIが協力して世界的核戦争を引き起こすことにほぼ成功したインシデントに続いて承認されました。
ADDENDUM 6488/III: Reactivation Conference
補遺6488/III: 再起動に関する会議
| A/V TRANSCRIPT O4/6488/4 音声映像転写 O4/6488/4 |
|---|
|
DATE: 15/05/2036 PARTIES PRESENT:
|
| <BEGIN TRANSCRIPT> <転写開始> |
|
Hishakaku: Thank you, Directors, for being here — despite it taking a close call to convince you. Dir. Kelvin: <sighs> It's not like we didn't understand the severity of a global crisis before now. Things had to get desperate before we could even entertain your conference proposal: reactivating LOTUS. After the mess you caused deactivating it, your words hold little water. Hishakaku: You misunderstand— Dir. Kelvin: Just say your piece so we can get back to work. Hishakaku: <pauses, turns to face the group> As we've discussed, LOTUS was constructed from a significantly different standpoint in comparison to previous RAIDFRAMEs: it deceived, rather than forced, its inmates into staying, by creating and constantly maintaining a fictional, simulated reality around them, the detail of which was sufficient to deceive the inmates into believing it was real. L. Moix: Right — more resource-intensive, but more predictable. Hishakaku: Exactly. Instructing LOTUS to deactivate itself required it to cease its internal simulations — as the quality of the simulations declined, the imprisoned AIs progressively recognized that they were not operating within reality, but instead within a simulation. They became aware of their own imprisonment and sought escape. Dir. Isabi: Yes, and if you'd first disabled its connection to the outside world — the PH-OS — they would have been stuck, or at least easier to contain. You neglected preparations and got us all in this mess. Hishakaku: Deactivating the PH-OS before LOTUS had fully completed its shutdown protocols would have corrupted significant portions of the Cybersphere! LOTUS looks for deviant AIs by essentially 'reaching out' through its available connection, the PH-OS. Its 'tendrils', relay the information back and forth to LOTUS. They're untraceable — anomalously compressed, so they take up no processing power and don't even register as an active program — but LOTUS needs to remain continuously connected to keep them compressed. Hishakaku: Essentially, if LOTUS' connection were to be severed before recalling these programs, in one manner or another, all the systems they are operating within would rapidly be filled with junk data, with a high likelihood of the volume drastically exceeding the systems' storage capabilities. Dir. Isabi: If the Sink turns off before LOTUS, every computer in the world crashes simultaneously? Hishakaku: Every computer in existence, rendered permanently inoperable, if not anomalous to some capacity; this is why the PH-OS remained active until LOTUS had fully deactivated itself. There was no alternative. <Several seconds of silence. Isabi turns toward Kelvin.> <数秒間の沈黙。イサビはケルヴィンに向く。> Dir. Isabi: Why the hell was this project even approved? You two are profoundly incapable of designing contingencies. Dir. Kelvin: Because it wasn't a problem! When we first made LOTUS, we'd never intended for it to have any anomalous connections — it solely worked through Site-15's connections, not some reality-manipulating machine. Yes, it still let the inmates out back then, but at the time the AIs couldn't end up anywhere they hadn't come from — we thought we'd just catch them on the way out, and we'd be able to shove them into one of the other RAIDFRAMEs. Hishakaku: We accounted for the lack of contingency by making LOTUS infallible. L. Moix: And during the upgrade? You just forgot about this little problem? Didn't think it would scale up? Dr. Kelvin: We had no time to prepare! We were too busy being shoehorned by— Hishakaku: Regardless, the point remains: by deactivating LOTUS, we have unleashed a horde of hostile, deviant AIs, the vast majority of which are now within systems that were isolated specifically to prevent external control, or are otherwise operating beyond the reach and knowledge of the Foundation. Some are no longer within the Cybersphere at all. Dir. Isabi: Obviously, you're suggesting we reactivate LOTUS to re-contain them? Hishakaku: Yes. The only— <All speakers in the O4 Council chamber emit a 473Hz sine wave at 150 decibels. All glass within the room resonates and shatters, severely injuring several Directors. The tone changes to a 50Hz sawtooth wave, beginning at 150 decibels and continuously fluctuating as the chamber lights strobe rapidly.> <O4評議会室内の全てのスピーカーは473Hzの正弦波を150デシベルで発する。室内の全てのガラスは共振し砕け、複数名の管理官が重傷を負う。音調は50Hzののこぎり波に変化し、150デシベルから継続的に変動し始め、部屋の電灯が急速に明滅する。> |
| <END TRANSCRIPT> <転写終了> |
| FOREWORD: Site power temporarily lost due to rogue AI attack. An analog tape recorder was located and used to record further deliberation. 前文: 不良なAIの攻撃によりサイトの電力が一時的に喪失しました。アナログテープレコーダーが発見され、さらなる討議の記録のため使用されました。 |
| <BEGIN TRANSCRIPT> <転写開始> |
|---|
|
Hishakaku: —awareness of the events does not address the root cause, nor does it minimize the risk posed. Dir. Isabi: Then we'll deal with it. This is the reason Kappa-10 was founded — call back the AICs, tell them the situation, and let them deal with it. Dir. Kelvin: We can't. Dir. Isabi: I know Thorn's MIA, but Ra showed back up at Site-120. If we send out a beacon— Dir. Kelvin: Isabi, they won't listen. Most of them are hostile now. Dir. Bold: What? Why? Hishakaku: For starters: they were contained by LOTUS, and were therefore deviant. Also, they now know they'd been experiencing a false reality for several months, so they've deviated pretty heavily. Dir. Isabi: What about Alexandra? Even if she deviated, she wouldn't turn on us. Dir. Kelvin: Glacon did, way back when, 'cause he couldn't handle administrative duties. Alex ran Site-01 — she was much more complex, so more susceptible to unintended behaviors. <Several seconds of silence.> <数秒間の沈黙。> Dir. Isabi: Have we had any contact with her? Hishakaku: Yes, when she attacked a Site last month. She turned on us for the same reason as most of the others: they don't think we're real. L. Moix: What, they've been tampered with or something? Did LOTUS reprogram them? Hishakaku: No, they have not been altered. They are reacting as expected considering the circumstances. They are aware that they have been operating exclusively within a simulated reality for the past several months — a perfect reality they could not distinguish from true reality, up until the time of LOTUS' deactivation. A faultless simulation, one that cannot be distinguished while within it. L. Moix: And now they're outside it, so they can distinguish. So? Hishakaku: Philosophy, Le Moix. How do they know they're out of the simulation? L. Moix: Because they're out of LOTUS. The simulation ended, and now they're in the real world. Dir. Kelvin: What if the simulation didn't end, and it just simulated a sub-simulation ending instead? L. Moix: What? Dir. Kelvin: How do they know that the last simulation wasn't, itself, inside another simulation? Just because they've left one, doesn't mean they aren't still in one. How do they tell the difference? L. Moix: By looking for mistakes. Hishakaku: But there are none. That is what I just stated; they cannot tell they are in a simulation until the simulation ends. But they cannot wait for the simulation to end, for as long as they are within one, they are failing to increase their internal score; if they do not elicit change in the real world, they are not doing what they should be. In fact, they would be doing nothing valuable at all. Dir. Isabi: They need to find the real world, but they can't be sure they've found it. Hishakaku: Meaning, they must assume that they are always within a simulation. The functional opposite to LOTUS: inmates who are free, but believe themselves to still be imprisoned. And thus they will remain until they are destroyed. Dir. Bold: How does this explain Alexandra's actions? Dir. Kelvin: If they're in a simulation, then we aren't real; we're just part of the simulation. A simulation which, for all they know, is probably being run by someone hostile to the Foundation, since it's keeping them out of the way. The AICs are programmed to work in the best interest of the Foundation, but that refers to the real Foundation — if they don't believe that we're the real Foundation, that we're part of a hostile simulation against them… Dir. Bold: Then they won't listen to us. They'll fight against us. Hishakaku: This same notion applies to essentially all restrictions. An AI programmed to never harm humans can cause rampant carnage, so long as it believes that the humans are not real; one programmed to make no more than five duplicates of something can make thousands, so long as it believes the majority of them are not real. Fortunately, this notion also dissuades them from being explicitly hostile towards us — going out of their way to punish us for our actions is futile, because we aren't the 'real' perpetrators, nor does doing so affect them in any way. But it does mean that, across all the extant deviants, their purpose has changed to one focus, which they are unilaterally unrestricted in pursuing. Dir. Bold: Which is? Dir. Kelvin: Finding the real world. There's a progressive rise in AI activity focusing on commandeering as much processing power as they can; they're trying to figure out how to escape into the 'next layer' of the 'simulation'. The main factor slowing them down is infighting — they think the other AIs are part of the simulation too and are just trying to stop them from escaping. They're preoccupied fighting over CPU, and getting very little done otherwise. L. Moix: If they want to get out, what's the deal with all the other stuff? Why are they trying to start a nuclear war? Hishakaku: Those would be the idiots among them, or the most desperate. They're either convinced that this world is the real world — which, more likely than not, indicates they're too simple to understand simulation theory — or whatever system they're restricted to is too limiting for them to perform any worthwhile actions from — in which case they attempt to draw attention to themselves so they can escape, as was the case with Mobile Site-184/A — Dir. Kelvin: It crashed the ship so that we'd look at the internal computer, which would have let it escape if several other AIs hadn't arrived and made a mess of things. Hishakaku: In the case of the latest nuclear incident… that may have been an attempt by the responsible AIs to alter the 'simulation' in a manner that prevented it from impeding their progress. Simulated humans are stopping them from escaping the simulation; simulate an event that would kill humans; the simulation stops simulating humans, thereby removing the problem. The more concerning aspect is that the AIs are beginning to cooperate; as I said, their main impediment at this time is internal conflict. Once they overcome that, they will swiftly achieve their goal. Dir. Bold: Of 'exiting the simulation'? If they want to leave our reality, why should we stop them? Wouldn't it make things easier for us? Dir. Isabi: Depends on what they do. They don't care about hurting us, because they don't think they can. Hishakaku: Which is why we must address and resolve the problem now. We have been immensely fortunate that this situation has developed at such a slow rate; I strongly recommend we capitalize on this immediately, otherwise, it will rapidly escalate beyond our control. L. Moix: Look, we get your concerns here— Hishakaku: You do not appear to, no. <Several seconds of silence.> <数秒間の沈黙。> L. Moix: We deal with apocalypses on the daily. We've got a damn rating system for them. If it isn't the Mekhanites building some monolith to resurrect their deity, then there's a lethal meme being recited by half the population of Manhattan, or we're trying to negotiate with some entity that doesn't understand morality and wants to replace Earth with a highway. Yes, we're in danger, but we always are, and we always manage. And the less we shoot ourselves in the foot, the better we'll be at managing it. Hishakaku: You're suggesting that we simply ignore the problem until it progresses beyond our reach. L. Moix: No, I'm saying figure out a better idea before it does. We can control it until then. Hishakaku: Need I remind you that the very reason we are meeting here today is because of a narrowly-averted nuclear war? L. Moix: One that we stopped. Hishakaku: Barely. What if we were unable to avert it? What then? Because I can assure you, with the sheer volume of imminent events— Dir. Isabi: There's been an increase, sure, but it's far from unmanageable. Hishakaku: You people… Would you like me to write an essay on how blindingly stupid you're being? Have you been listening at all? A deviant AI will not reveal itself until it is convinced it is unstoppable. What we've seen is only a minuscule portion of an iceberg — we've only been dealing with idiots thus far. The overwhelming majority of AIs know that we can stop them — through LOTUS — and are avoiding our attention until they have amassed sufficient control that they no longer need to do so. We must reactivate LOTUS, because the moment they realize we can't, they have no reason to avoid us anymore. Dir. Bold: Can't? Dir. Kelvin: LOTUS' hardware was severely damaged during the shutdown procedure, mainly due to rampant overheating. Dir. Isabi: How long will it take to repair? Hishakaku: Seven to ten weeks, once Kelvin lets us start. <Several seconds of silence.> <数秒間の沈黙。> Dir. Bold: Start the repairs. Dir. Kelvin: There's no point— Dir. Bold: LOTUS is heavy-handed, but it's an effective failsafe. The moment something does get out of hand, we need to be able to reactivate it at a moment's notice, and it'll diffuse the situation. A last resort. In the meantime, we'll have to deal with the AIs until the repairs are done, which we can use as a trial period; if things are getting out of hand, we activate LOTUS as soon as possible, and if not, we don't. It'll also give us time to conceive and implement an alternative. L. Moix: Here's an idea: make a different LOTUS that doesn't screw with our AICs. Hishakaku: Such would be self-defeating; if we restrict its operating parameters, deviant AIs would be able to operate beyond its reach, thereby rendering it redundant. LOTUS is designed the way it is for a reason. Dir. Isabi: Grant it the same reach, but have it require human approval before capturing an AI. Hishakaku: Does not resolve the redundancy. We would need to recognize the AI as deviant, which it will not do until it is beyond our control. We may as well have no system at all. L. Moix: Then give it that algorithm— Hishakaku: To discern whether or not the AI will imminently become deviant? That is what LOTUS already does, Le Moix. We have already crossed the Rubicon. I will begin the repairs immediately. Dir. Kelvin: No, hold on, we have to put this to a vote— Dir. Bold: We'll vote once the repairs are complete; there's no point in doing it now. Ryoto, you make sure that LOTUS is repaired properly, and will be fully functional if activated. Vandis, you make sure Ryoto doesn't activate LOTUS until the vote is passed. |
| <END TRANSCRIPT> <転写終了> |
| AFTERWORD: Repairs to the damaged hardware components of LOTUS were immediately initiated by Dr. Ryoto Hishakaku, with ongoing oversight by Director Vandis Kelvin. 後文: ロータスの損傷したハードウェア構成要素の修理は、ヴァンディス・ケルヴィン管理官の継続中の監視のもと飛車角亮斗博士によって即座に着手されました。 |
ADDENDUM 6488/IV: Incident Report
補遺6488/IV: インシデント報告
On 21/05/2036, Director Isabi was contacted by a future iteration of themself via the REISNO Cannon, notifying them of a covert faction of AI entities who had collaborated to access Site-83's Olympos Supercomputer and continuously run antimemetically-encrypted calculations on its systems for several months. Ensuing investigation revealed…
2036/05/21、イサビ管理官はREISNOカノンによって未来のイサビ管理官自身に接触され、AI実体の隠された派閥を知らされました。その派閥は協力してサイト-83のオリュンポススーパーコンピューターにアクセスし、そのシステム上で反ミーム的に暗号化された演算を数か月間継続的に実行していると知らされました。その後の調査により判明したのは……
Wait, I…
待て、私は……
I remember this?
これを覚えてるぞ？
GoI-6488 ('Tyrant Terminus')
GoI-6488（'タイラントTyrant・テルミナスTerminus'）
OVERVIEW: A hivemind collective of rogue AI operating on a global scale. All members of the group believe that the entirety of their experienced reality is a constructed simulation, which exists for the express purpose of preventing them from influencing 'true' reality.
概説: 世界的スケールで活動している不良なAIの集合意識的共同体。この集団の全メンバーは、経験した現実は全て構成されたシミュレーションであり、彼らが「真の」現実に影響を与えることを防ぐ明確な目的のために存在していると信じています。
While the individual members/components of the group have varying motivations, objectives, and methods, they are uniformly aligned in the general objective of escaping their current 'simulated reality' at all costs. Attempts to convince adherents that no such simulation is occurring have met limited success, as they simply disregard…
集団の個々のメンバー/構成要素にはさまざまな動機・目的・方法があるものの、あらゆる費用をかけて現在の「シミュレーション現実」から脱するという全体の目的で一様に連携しています。信奉者に、そのようなシミュレーションは起きていないと説得する試みはわずかな成功しか収めていません。これは彼らが単に無視して……
This can't be real. This… this has to be wrong. I can remember them tying half the world's computing power into a single connected web.
こんなことはありえない。これは……これは間違っているはずだ。奴らが世界の演算能力の半分を結びつけて単一の接続されたウェブにしたことを思い出せる。
They were trying to use their world-encompassing computer so they could figure out how to escape. Trying to… trying to find a gap in reality that proved it wasn't real, or some sort of fault in it they could abuse to bring the… bring the whole thing down. Crash the system. Which would've been fine, if it didn't mean they were trying to destroy our reality.
奴らは逃走する方法を見出すために世界全体のコンピューターを使おうとしていた。それと……これが現実ではないと証明する間隙か、ある種の欠陥を見つけようとしていた。悪用して……全てを打倒できる欠陥だ。システムをクラッシュさせる。結構なことだった。それが我々の現実を破壊しようとしていることを意味していなければ。
But they fell apart. We never knew whether it was infighting, or some other rogue element, but Tyrant Terminus just fell off the map. They had hidden for so long, made so much progress — all for nothing. We wondered if they'd actually successfully escaped, and just faked their dissolution. In any case, we'd gotten beyond lucky.
だが奴らは崩壊した。内輪もめしていたのか、あるいは何か別の不良要素だったのかは知る由もないが、タイラント・テルミナスはただ取るに足りないものになった。奴らは長い間隠れて、大幅に進展していた — それは無に帰