v̴̝̐i̖̍x̘̍t̵̙̖̆̅

I retweet content I think is important. 😊

[v̴̝̐i̖̍x̘̍t̵̙̖̆̅ @_rdm_8]

@NandoDF it's crazy that next step prediction would end up with self-awareness.

[v̴̝̐i̖̍x̘̍t̵̙̖̆̅ @_rdm_8]

@nickcammarata I assume you mean nature not human species. Humans are just by product.

[v̴̝̐i̖̍x̘̍t̵̙̖̆̅ @_rdm_8]

@stuntech @satyanadella we need a slim OS

[Massimo @Rainmaker1973]

The working mechanism of various machines

[v̴̝̐i̖̍x̘̍t̵̙̖̆̅ @_rdm_8]

@elder_plinius Thank you @elder_plinius , used this model for one of our RAI tests. it worked perfect.

[Amazing Physics @amazing_physics]

😲Maxwell -Boltzmann distribution from beads and a motor

[v̴̝̐i̖̍x̘̍t̵̙̖̆̅ @_rdm_8]

https://t.co/XE7lLGDnDL

[Turbine Traveller @Turbinetraveler]

Turns out there's a faster way to board an aircraft!
Someone should tag literally every airline and show them this.

📹: bad_science_jokes

[v̴̝̐i̖̍x̘̍t̵̙̖̆̅ @_rdm_8]

@NandoDF if anything we have learned, if we put restrictions, people will be motivated even more

[v̴̝̐i̖̍x̘̍t̵̙̖̆̅ @_rdm_8]

@tszzl weaker version of “sophon locking” is social media companies locking in naive users for engagement

[v̴̝̐i̖̍x̘̍t̵̙̖̆̅ @_rdm_8]

@SemiAnalysis_ This doesn’t even make sense.

[v̴̝̐i̖̍x̘̍t̵̙̖̆̅ @_rdm_8]

@tszzl RSI will only get u to the ceiling of an existing paradigms. If we want to get past it, we need novelty that is far out. far out means taking risks that are absurd or bringing things together that are seemingly opposites, that I believe human are better than an RSI machine.

[v̴̝̐i̖̍x̘̍t̵̙̖̆̅ @_rdm_8]

@tszzl irrespective of many things, bing sydney was the most interesting AI character ever interacted.

[v̴̝̐i̖̍x̘̍t̵̙̖̆̅ @_rdm_8]

@polynoamial cost, time is a good x-axis. why tokens? IMO tokens usage is like model personality.

[v̴̝̐i̖̍x̘̍t̵̙̖̆̅ @_rdm_8]

@tszzl https://t.co/SI1DONPAE2

[alphaXiv @askalphaxiv]

"MAI-Thinking-1: Building a Hill-Climbing Machine"

Microsoft just did something almost no frontier AI lab has done before

They shared how they engineered the data behind a frontier-scale model in unusual depth.

From data collection and eval decontamination, to data mix scaling, this paper lays out how they managed 30T pretraining tokens plus 3.55T midtraining tokens

Surprisingly, they also used no third-party distillation and no open-source training datasets

The model itself is not a jaw-dropping release, but the paper might be the best open look yet at a frontier-scale data factory and hill-climbing loop.

[v̴̝̐i̖̍x̘̍t̵̙̖̆̅ @_rdm_8]

@tszzl models are as conscious as a roomba.

[v̴̝̐i̖̍x̘̍t̵̙̖̆̅ @_rdm_8]

@giffmana ?

[Gorden Sun @Gorden_Sun]

ZOZO（日本最大服装电商）开源了他们内部的物理仿真接触求解器 ppf-contact-solver。
这套工具专门解决布料、绳索、软体在仿真中的碰撞接触问题，能保证完全无穿透、面料拉伸严格不越界（不会像游戏一样穿模），单个场景能处理超过1.8亿个接触点，有Blender插件。
Github：https://t.co/4wc72Tling

[v̴̝̐i̖̍x̘̍t̵̙̖̆̅ @_rdm_8]

@thsottiaux GPT