I-AI iye yandisa iimveliso zokusetyenziswa kwiminyaka emininzi ezidlulileyo, ezisetyenziswa kumazwe afana ne-healthcare, i-finance, kunye nezixhobo zokuvelisa. Enye izixeko ezininzi ezininzi ezininzi zihlanganisa kwi-interactive storytelling, apho inokukwazi ukukhuthaza ama-characters ezinokufunda, zihlanganisa, kunye nokuphendula ngokufanelekileyo: ukwakha iintlobo, ukuvelisa iimpazamo, kunye nokufumana imibuzo ngaphandle kwe-scripted prompts. I-Cris Lenta yaziwa le ngempumelelo kunye nokwakha , i-AI-native-life simulation platform eyenziwe ngexabiso engaphezu kwama-2 billion tokens kwiintlobo zeemigodi ezigidi ziye ziye ziye ziye ziye ziye ziye ziye ziye ziye ziye ziye ziye ziye ziye ziye ziye ziye ziye ziye ziye ziye ziye ziye ziye ziye ziye ziye ziye ziye ziye ziye ziye ziye ziye ziye ziye ziyi-consumer product. Ukucinga Ukucinga The Theory Phantsi LifeSim Xa iingcebiso like " "Wagqibeleleyo, bafumane isibuyekezo esibonakalayo yekhono: Iimpawu ezisetyenzisiweyo ze-LLM ezibonakalayo kunye nenkqubo yokukhusela iziganeko kunye nokuxhomekeka kwakhona kunokukwazi ukuvelisa izivumelwano zamazwe ngokufanelekileyo ngaphandle kwe-scripting. apho ama-agents baboniswa kwi-1,000 ama-individualities kwaye bakwazi ukufumana kubo ngexesha elikhulu. I-Generative Agents: I-Interactive Simulacra of Human Behavior Ukuhlolwa kweStanford ngo-2025 I-Generative Agents: I-Interactive Simulacra of Human Behavior Ukuhlolwa kweStanford ngo-2025 Kodwa i-experiment yenzelwe phantsi kweengxaki ezizodwa: ama-agents ezingu-25 ezisebenza kwiintsuku ezimbini ze-simulated kwi-sandbox ebonakalayo, kunye neengxaki ze-computing nje zithunyelwa kumadoda ezininzi kwi-scope ebonakalayo. Yintoni iinkcukacha ziye ziye ziye ziye ziye ziye ziye ziye ziye ziye ziye Kwimeko yokubhalisa ukuba i-architecture ye-character-driven AI agents kunokwenzeka ngokutsho kunye nenkqubo kwi-consumer-scale, kunye nemizuzu emininzi yama-characters kunokuba emininzi emininzi. I-Cris ye-Lenta I-Cris ye-Lenta Zonke i-subsystem ebalulekileyo, njenge-memory retrieval, i-reflection synthesis, kunye ne-real-time coherence, ibonise iingxaki ze-engineering ezininzi xa amaxabiso zithunyelwe ngamaxabiso. I-Latency isixazululo, i-token cost isixazululo, ukulawulwa kwizixazululo ze-asynchronous isixazululo, kunye ne-temporal coherence inokufumana iingxaki ze-architectural. Kodwa uCris waya le ngqongqo phakathi kwezobugcisa kunye ne-practical kwaye ilawula ukuba yinkqubo yobugcisa yayo yokusombulula. Ukufundwa e-University Ludwig Maximilian eMunich, uqhagamshelane namaphepha ezimbini ezidlulileyo ezidlulileyo kwi-MIT Press. I-journal kunye ne-IEEE Xplore, zombini zihlanganisa kwi-self-replicating networks ne-neuronal. "Izifundo ezidlulileyo zibonise le nkqubo kunye ne-25 ukuya kwi-50 ama-agents kwiinkalo zophando eziholile," uqhagamshelane. "Ndiye ndithanda ukuvelisa kwimveliso ye-consumer apho abadlali zibonise amayunithi ezininzi." Ubomi obungapheliyo Ukusebenza njenge-memory Kwi-architecture ye-LifeSim ibekwe kwinkqubo ye-memory efakwe kwi-recovery ye-size-weighted kunye ne-temporal decay. Kwi-system, imemori ye-high-intensity ibekwe ngelixa iintlobo zokusebenza zokusebenza zokusetyenziswa. Yonke imemori eyenziwa yi-character ibekwe kwiiyunithi ezimbini: i-intensity ye-emotional ngexesha lokufunda kunye ne-recent relative to the time simulated. Imemori ye-high-salience ibekwe ukuxhaswa; i-interactions ze-mundane (njenge-greetings ze-routine kunye ne-exchanges ze-trivial) ibekwe ngokugqithisileyo. I-effect esebenzayo i-character ibonisa iingxowa zokuqala ze-RAG-based NPC kwiinyanga ezidlulileyo emva kwexesha elidlulileyo, kodwa ibandakanya iingxowa ze-routine evela kwiiveki edlulileyo. Oku ikhiqiza uhlobo lwe-recall ye-selective eyenza iingxowa ze-authentic, ngokuthe ngqo isixazululo esebenzayo kwi-naive RAG-based NPC systems, apho ama-characters zithembisa yonke into nge-equal weight (okukwazi ukuvelisa i-uncanny effect), okanye ukufumana i-interactions ebalulekileyo kwi-contextual ngokupheleleyo, ukunciphisa i-immersion. I-memory system inikeza kwimeko yeengxaki ze-autonomous relationship formation. I-NPC isebenze iinkqubo ze-planning eyahlukileyo kunye ne-theory-of-mind modeling, ukwakha iingxaki kunye nokuphuhlisa iingxaki ngaphandle kweengxaki ze-player, kunye neengxaki zeengxaki zeengxaki ze-compatibility ezisetyenziswe ngokusebenzisa i-compatibility scoring kwiiveki ze-personality kunye neengxaki zeengxaki zeengxaki zeengxaki zeengxaki zeengxaki zeengxaki zeengxaki zeengxaki zeengxaki zeengxaki zeengxaki zeengxaki zeengxaki zeengxaki zeengxaki zeengxaki zeengxaki zeengxaki zeengx Ukuphendula i-time problem Enye ingxaki olungabonakaliyo yokuNkqubo yeCris olungasetyenziswa yi-temporal coherence: nto leyo, indlela i-agents ibekwe emangalisayo kwi-time scales ezahlukeneyo xa isivinini se-simulation ibekwe. Isisombululo wayo yaba inkqubo ebizwa ngokuba yi-attention-based time mechanics, apho i-simulation granularity ibekwe kunye ne-player attention. Ixesha elandelayo ngokukhawuleza ngexesha le-routine, iimizuzu eziqhelekileyo kunye nokunciphisa ngexesha le-interactions ezininzi ezibonakalayo ezidlulileyo. I-system ibhagamshelana ne-density ye-interaction, i-valence ye-emotional, kunye ne-narrative tension ngexesha elifanayo. Kwiimeko ze-high-signal, i-tick rate ikunciphisa, ukwandisa ixesha le-subjective, xa ngexesha le-low-signal, i-time compresses, kwaye iiyure zingadlulanga kwiiveki. Phantsi ke ibekwe i-double-tick architecture enza i-simulation yokufumana umdlali ukusuka kwihlabathi yokuphuma kwihlabathi. Xa umdlali akuyona ngokugqithisileyo, i-NPC ibekwe ekugqithisileyo kwi-cycle ye-tick ye-low-cost. I-system ibekwe i-serialized world-state snapshots yokufumana i-entity positions, i-relationships, kunye ne-in-progressive NPC goals. Xa umdlali ibekwe kwihlabathi, i-simulation ibekwe kwihlabathi ukusuka kwihlabathi lokugqithisileyo. Le nqakraza ezimbini yenza iimveliso ezihlabathi ezihlabathi ngokufanelekileyo ngaphandle kwezimali ze-computing. Iimveliso ze-background zithunyelwe ngexabiso ngenxa ye-granularity yayo aphansi kwaye ngenxa yokuba zithunyelwe kwi-rendering-related overhead. I-AI-Driven Living World Iinombolo ezidlulileyo kwi-LifeSim zihlanganisa ukuxhaswa okuqhubekayo kunokuba ngexesha elifanelekileyo. Abadlala wamkelekile ngaphezu kwama-60,000 kwi-platform, ngamnye kwenza i-memory graph kunye ne-network ye-relationship yayo. Le nkqubo ilawula yonke into kwiiyure ze-asynchronous ngaphandle kokuchaswa kwimifanekiso. Iiyure ze-peak ze-user zithunyelwe kwiiyure ezisixhenxe, oku kubonisa izinga elipheleleyo lwe-immersion eyenza kwi-chatbot-style kuphela. Ukwakha i-platform ukusuka kwangaphambili njengomfundisi omnye, uCris uye wabelane i-$350,000 evela kumazwe abavela kumdlalo, kuquka i-San Francisco-based Founders Inc, ebandakanya i-exits kwi-Twitch, kunye nePaul Braghiel nge-SMOK VC, i-investor yokuqala kwi-Unity kunye ne-Roblox. I-Cris isakhelwe ngoku i-seed round ukwandisa i-team kunye ne-infrastructure, kunye neengcingo yokufunda i-creator program ebonelela kumazwe omnye ukwakha kunye nokuthintela iinkonzo ze-AI-driven narrative kwi-platform. Izixhobo ze-generative zithembisa kuphela iinkqubo zayo zangaphakathi kunye neengcali ezisetyenziswa, kwaye umsebenzi ka-Cris Lenta kunye ne-LifeSim ibonise ukuba zinokufumana ukufika kwiphepha le-research kwi-product ye-consumer. 
 
 Le nqaku lithunyelwe njenge-release ye-Jon Stojan phantsi kwe-HackerNoon's Business Blogging Program. Le nqaku lithunyelwe njenge-release ye-Jon Stojan phantsi kwe-HackerNoon's Business Blogging Program.

This writer has a vested interest be it monetary, business, or otherwise, with 1 or more of the products or companies mentioned within.

Le audio iveliswe ngolwimi lokuqala lwebali!

I-AI Life Simulator Isebenza I-2 Billion I-Token - Kwaye Umbhali Wabelani I-It Is Just Getting Started

About Author

IZIMVO

ZIJONGE IIMPAWU

ELI NQAKU LINIKEZELWE KU

Related Stories

Light-Mode

Classic

Newspaper

Minty

Dark-Mode

Neon Noir

Minty

HN StartUps