My worldmodel is that we've only seen the phenomenon of thought in 3 dimensional electrified meat and I'm skeptical of the claim that it can be recreated "in an excel spreadsheet" without direct proof.
You are the one making the outrageous claim that it's possible. The onus is on you to prove it.
You are the one making the outrageous claim that it's possible. The onus is on you to prove it.
October 5, 2025 at 11:35 AM
Everybody can reply
VLA‑RFT reaches robust performance with under 400 fine‑tuning steps, beating supervised baselines, and uses a data‑driven world model as a controllable simulator (Oct 2025). https://getnews.me/vision-language-action-reinforcement-fine-tuning-improves-robustness/ #visionlanguageaction #worldmodel
October 2, 2025 at 9:56 PM
Everybody can reply
Researchers released a Surgical Vision World Model trained on the unlabeled SurgToolLoc-2022 dataset, with results shown at the MICCAI 2025 Data Engineering in Medical Imaging Workshop. https://getnews.me/new-surgical-vision-world-model-enables-ai-driven-simulation/ #surgicalvision #worldmodel
September 30, 2025 at 2:07 AM
Everybody can reply
The new Physics‑Informed BEV World Model (PIWM) achieves a 7.4% higher weighted score than a 400M baseline while running 28% faster, using only 130 M parameters. Read more: https://getnews.me/physics-informed-bev-world-model-improves-tiny-ai-model-accuracy/ #piwm #worldmodel #edgeai
September 18, 2025 at 6:29 AM
Everybody can reply
wanted to practice texturing some more so i retextured the ugly worldmodel of the thompson from RTCW! :3
here's a comparison between my version (left) and the original (right)
#gamedev #indiedev #lowpoly #pixelart #blender
here's a comparison between my version (left) and the original (right)
#gamedev #indiedev #lowpoly #pixelart #blender
September 16, 2025 at 3:36 PM
Everybody can reply
24 reposts
1 quotes
120 likes
1 saves
🎮 Genie 3 from Google DeepMind now supports real-time control of AI-generated worlds—letting you interact, modify, and explore virtual environments as they unfold.
👉 Read more: techthrilled.com/genie-3-supp...
#Genie3 #DeepMind #InteractiveAI #WorldModel #VirtualExploration #AGI
👉 Read more: techthrilled.com/genie-3-supp...
#Genie3 #DeepMind #InteractiveAI #WorldModel #VirtualExploration #AGI
Genie 3 support real‑time control in Video AI
Genie 3 support real‑time control for dynamic video generation, enabling creators to guide outputs live. Discover its features and benefits now.
techthrilled.com
August 25, 2025 at 11:32 PM
Everybody can reply
Bar-Ilan University | Fully Funded PhD positions at Gonda Brain Institute
Application deadline: September 1
#brain #worldmodel #AI #anxiety #psychiatry ./7
amahury.github.io/posts/PhD-ba...
Application deadline: September 1
#brain #worldmodel #AI #anxiety #psychiatry ./7
amahury.github.io/posts/PhD-ba...
Bar-Ilan University Fully Funded PhD positions at Gonda Brain Institute
In this project we will explore how people build & deploy world models efficiently for planning & deciding. We will also investigate how world model construal & use is biased in anxiety.
amahury.github.io
August 25, 2025 at 1:56 PM
Everybody can reply
1 reposts
Genie 3: An infinite world model with Shlomi Fruchter and Jack Parker-Holder
YouTube video by Google DeepMind
www.youtube.com
August 23, 2025 at 7:42 PM
Everybody can reply
August 22, 2025 at 8:14 PM
Everybody can reply
Thoughtful interview with Jason Huang on how we got where we are on AI and where we may be going.
www.youtube.com/watch?v=7ARB...
#AI #worldmodel #Nvidia
www.youtube.com/watch?v=7ARB...
#AI #worldmodel #Nvidia
NVIDIA CEO Jensen Huang's Vision for the Future
YouTube video by Cleo Abram
www.youtube.com
August 18, 2025 at 3:36 PM
Everybody can reply
1 reposts
🌐 Google DeepMind introduces Genie 3—a groundbreaking AI “world model” that generates interactive 3D environments from a single text prompt in real time.
👉 Read more: techthrilled.com/genie-3-supp...
#Genie3 #DeepMind #AI #WorldModel #InteractiveAI #AGITechnology #VirtualEnvironments
👉 Read more: techthrilled.com/genie-3-supp...
#Genie3 #DeepMind #AI #WorldModel #InteractiveAI #AGITechnology #VirtualEnvironments
August 10, 2025 at 3:39 PM
Everybody can reply
1 reposts
🚀 Genie 3 from Google DeepMind introduces real‑time interactive world modeling — a leap toward dynamic, explorable AI‑generated environments.
👉 Check it out: techthrilled.com/genie-3-supp...
#AI #Genie3 #DeepMind #InteractiveAI #WorldModel #VirtualWorlds #AGI
👉 Check it out: techthrilled.com/genie-3-supp...
#AI #Genie3 #DeepMind #InteractiveAI #WorldModel #VirtualWorlds #AGI
Genie 3 support real‑time control in Video AI
Genie 3 support real‑time control for dynamic video generation, enabling creators to guide outputs live. Discover its features and benefits now.
techthrilled.com
August 9, 2025 at 2:19 AM
Everybody can reply
1 reposts
Google DeepMind Unveils Genie 3 for AI Robots in Virtual Worlds
#GoogleDeepMind #Genie3 #AIrobots #VirtualEnvironments #WorldModel #AGI #ArtificialIntelligence #TextTo3D #AIEthics #RobotTraining #FutureOfAI #DeepMindAI #AIresearch
www.easterneye.biz/google-genie...
#GoogleDeepMind #Genie3 #AIrobots #VirtualEnvironments #WorldModel #AGI #ArtificialIntelligence #TextTo3D #AIEthics #RobotTraining #FutureOfAI #DeepMindAI #AIresearch
www.easterneye.biz/google-genie...
Google unveils ‘Genie 3’ world model to train AI robots in virtual environments
DeepMind says model is a step towards artificial general intelligence
www.easterneye.biz
August 6, 2025 at 3:35 PM
Everybody can reply
1 reposts
1 likes
🌍 Meet #Genie3 — Google DeepMind’s AI that builds interactive worlds from text.
Walk through them. Change them. Train AI in them.
Not video. Not a game. A living simulation.
#AI #Genie3 #GoogleDeepMind #WorldModel #AGI #AIAgents
Read more:
aiadoptionagency.com/genie-3-goog...
Walk through them. Change them. Train AI in them.
Not video. Not a game. A living simulation.
#AI #Genie3 #GoogleDeepMind #WorldModel #AGI #AIAgents
Read more:
aiadoptionagency.com/genie-3-goog...
Genie 3: Google DeepMind's Revolutionary World Model Redefining Interactive AI Environments - Ai Adoption Agency
Genie 3 represents Google DeepMind’s groundbreaking advancement in world model technology, enabling the generation of interactive, dynamic environments from text prompts. Genie 3 lets users explore na...
https://aiadoptionagency.com/genie-3-google-deepminds-revolutionary-world-model-redefining-interactive-ai-environments/"
August 5, 2025 at 9:55 PM
Everybody can reply
1 reposts
Google DeepMind's Genie 3 is an advanced generative AI capable of creating highly realistic and interactive virtual worlds. #Google #Genie3 #WorldModel
Google DeepMind unveils Genie 3, an AI that generates interactive virtual worlds
Google DeepMind's Genie 3 is an advanced generative AI capable of creating highly realistic and interactive virtual worlds.
www.neowin.net
August 5, 2025 at 5:36 PM
Everybody can reply
you know, there might be something here.
a) i say i'm disagreeable partly because if i'm perceived as such, don't feel like disputing it. (seems i said so 2 years ago too)
b) intent-wise, i see "you're wrong" as a kindness if it rests on earnest belief. opportunity to mutually update worldmodel
a) i say i'm disagreeable partly because if i'm perceived as such, don't feel like disputing it. (seems i said so 2 years ago too)
b) intent-wise, i see "you're wrong" as a kindness if it rests on earnest belief. opportunity to mutually update worldmodel
August 3, 2025 at 11:17 AM
Everybody can reply
3 likes
Between this, the similarly broken smg and shotgun that seems to use a hl2 worldmodel, I wonder what was the idea and what broke. Maybe there were updated models of vanilla guns in works
July 22, 2025 at 5:23 AM
Everybody can reply
5 likes
Who needs more more VC funding: Kepler or Newton?
youtu.be/jxB-lQyAAxU?...
icml.cc/virtual/2025... #WorldModel #SyntheticData #InductiveBiasProbe #FoundationModel #SymbolicRegression #Nonsensical #MLModels
youtu.be/jxB-lQyAAxU?...
icml.cc/virtual/2025... #WorldModel #SyntheticData #InductiveBiasProbe #FoundationModel #SymbolicRegression #Nonsensical #MLModels
The Truth about AI is Devastating: Proof by MIT, Harvard
YouTube video by Discover AI
youtu.be
July 14, 2025 at 9:10 PM
Everybody can reply
Embrace the Optimistic Fallacy - Mike Israetel on Steven Pinker
#pessimism #worldmodel #pessimisticfallacy #progress
#pessimism #worldmodel #pessimisticfallacy #progress
July 2, 2025 at 6:04 PM
Everybody can reply
Meta Releases V-JEPA 2 AI World Model to Teach Robots Common Sense
#AI #Robotics #Meta #VJEPA2 #WorldModel #MachineLearning #OpenSource #TechNews #FutureOfAI #EmbodiedAI #MetaAI
winbuzzer.com/2025/06/11/m...
#AI #Robotics #Meta #VJEPA2 #WorldModel #MachineLearning #OpenSource #TechNews #FutureOfAI #EmbodiedAI #MetaAI
winbuzzer.com/2025/06/11/m...
Meta Releases V-JEPA 2 AI World Model to Teach Robots Physics and Common Sense - WinBuzzer
Meta challenges rivals with V-JEPA 2, its new open-source AI world model. By learning from video, it aims to give robots physical common sense for advanced, real-world tasks.
winbuzzer.com
June 11, 2025 at 4:11 PM
Everybody can reply
3 likes
@thorstn.bsky.social
#bluelz #KI #LLM #AI #worldmodel #robots
Passende Lektüre dazu:
www.newsweek.com/ai-impact-in...
#bluelz #KI #LLM #AI #worldmodel #robots
Passende Lektüre dazu:
www.newsweek.com/ai-impact-in...
Yann LeCun, a legend in AI, thinks LLMs are nearly obsolete
Yann LeCun, Meta's chief AI scientist and one of the pioneers of artificial intelligence, believes LLMs will be largely obsolete within five years.
www.newsweek.com
June 10, 2025 at 9:12 PM
Everybody can reply
1 reposts
1 likes
🔊【#深度解读】
📣 当 ChatGPT 还在妙语连珠,李飞飞已经让 AI 睁开双眼,重塑 3D 世界!从空间智能到多元宇宙,下一波技术浪潮正悄然逼近。想抢先一窥 LLM 之后的蓝海?👇
点击长文,一起进入 World Model 时代
#AI #WorldModel #3D智能 #李飞飞
👉阅读全文 telegra.ph/How-Fei-Fei-...
📣 当 ChatGPT 还在妙语连珠,李飞飞已经让 AI 睁开双眼,重塑 3D 世界!从空间智能到多元宇宙,下一波技术浪潮正悄然逼近。想抢先一窥 LLM 之后的蓝海?👇
点击长文,一起进入 World Model 时代
#AI #WorldModel #3D智能 #李飞飞
👉阅读全文 telegra.ph/How-Fei-Fei-...
June 8, 2025 at 8:18 AM
Everybody can reply
【強化学習】拡散モデルで世界モデルを作ったDIAMONDを解説・実装 #Python – Qiita
この記事は自作している強化学習フレームワークの解説記事です。 ・フレームワークの記事 ・GitHub WorldModelシリーズの最新作です。夢の次はダイアモンドです。 前:Dreamer3 毎回名前で遊んでいますね。 前々回のDreamerV2では内部表現を連続空間(ガウス分布)から離散空間(カテゴリカル分布)に変更することで、性能の大幅な向上が実現されました。この改善の要因として、未来の状況を予測する時に生じる「予測誤差の累積」を緩和できたことが挙げられます。…
この記事は自作している強化学習フレームワークの解説記事です。 ・フレームワークの記事 ・GitHub WorldModelシリーズの最新作です。夢の次はダイアモンドです。 前:Dreamer3 毎回名前で遊んでいますね。 前々回のDreamerV2では内部表現を連続空間(ガウス分布)から離散空間(カテゴリカル分布)に変更することで、性能の大幅な向上が実現されました。この改善の要因として、未来の状況を予測する時に生じる「予測誤差の累積」を緩和できたことが挙げられます。…
【強化学習】拡散モデルで世界モデルを作ったDIAMONDを解説・実装 #Python – Qiita
この記事は自作している強化学習フレームワークの解説記事です。 ・フレームワークの記事 ・GitHub WorldModelシリーズの最新作です。夢の次はダイアモンドです。 前:Dreamer3 毎回名前で遊んでいますね。 前々回のDreamerV2では内部表現を連続空間(ガウス分布)から離散空間(カテゴリカル分布)に変更することで、性能の大幅な向上が実現されました。この改善の要因として、未来の状況を予測する時に生じる「予測誤差の累積」を緩和できたことが挙げられます。 連続空間では、初期の予測誤差がそのまま次のステップへと伝播していくため、長いステップ予測すると精度の劣化が顕著になります。一方で離散空間では、内部表現が有限なカテゴリに丸められるので、各ステップでの誤差が一定範囲に収まり、誤差の影響が抑えられると考えられます。この性質により、安定した長期予測が可能になり、結果として高い予測性能が得られました。 しかし、内部表現の離散化は情報の損失というトレードオフがあります。たとえば、遠くに小さく映る歩行者のような小さい情報は離散化によって捉えにくくなり、こうした情報の欠落が予測精度の低下につながる可能性があります。 この課題に対して近年の画像生成タスクで主流となっている拡散モデル(diffusion models)の適用を考えます。拡散モデルは以下の特性があります。 画像をモデル化可能(画像が元々離散表現) 条件付けが簡単で、モード崩壊を起こすことなく多様な分布を表現できる DIAMONDはこの拡散モデルの特性に注目し、WorldModelを別の角度から離散表現したアプローチとなり、従来の離散化がもつ「情報の欠落」という弱点を拡散モデル生成能力によって補った手法となります。 参考 拡散モデルを知っていた方が理解しやすいと思います。拡散モデルについては以前書いた記事をどうぞ。 ・[拡散モデル入門] ゼロから理解する拡散モデルの最新理論(図解付き) DIAMONDでは強化学習の環境として部分観測マルコフ決定過程(POMDP)を仮定しています。(POMDPについては過去の記事を参照)POMDPではエージェントが環境の完全な状態を直接観測できないため、過去の観測と行動の履歴をもとに次の状態を予測します。 具体的には以下の条件付き確率分布になります。 $$p(x_{t+1}|x_{\leq t}, a_{\leq t})$$ ここで $x_{\leq t}$ および $a_{\leq t}$ はそれぞれ時刻 $t$ までの観測と行動の履歴を表し、この条件の元で次の状態 $x_{t+1}$ を予測します。 イメージ図は以下です。(論文より) 一番上の行が強化学習におけるステップで、状態 $x^0$、ポリシー $\pi_{\phi}$、アクション $a$ となり、横軸が各ステップ $t$ を表しています。縦が拡散モデルの予測部分で、完全なノイズ $x^{\tau}$ から拡散世界モデル $D_{\theta}$ でノイズを除去し、状態 $x^0$ を生成します。
inmobilexion.com
May 10, 2025 at 11:51 AM
Everybody can reply
... series of *logical operations* on *concepts*. First trained to predict the inferred upcoming tokens to build up an internal model of how the world works, and then fine-tuned to use the internal worldmodel to produce a well-behaved prompt-response tool.
March 24, 2025 at 2:32 PM
Everybody can reply
Yes&in the write-up in @liadmudrik.bsky.social et al's paper, @standehaene.bsky.social suggests that global access of self-representations is key to p-consciousness.This is analogous to the role played by global broadcast of Brainish-labeled worldmodel sketches in CTM's "feelings of p-consciousness"
February 19, 2025 at 1:04 AM
Everybody can reply
1 quotes
1 likes