固定されたツイート
Orbital Dev
136 posts

Orbital Dev
@sguestsett
Building websites with AI 🌑 I respect your time
Poland 参加日 Mart 2024
124 フォロー中21 フォロワー

@ModerateIndian1 @sebkrier China only open-sources its models because otherwise no one would care about them
English

@sebkrier Either call it closed source vs open source or call it China vs US. It's a shame US doesn't have a single open source model on that list. And it's a badge of honour for china that all of the open source models are from China
English

DeepSeek V4’s capability lags behind leading U.S. models by about 8 months. nist.gov/news-events/ne…

English

@bridgemindai Opus is 5T according to Musk and GPT might be around 3-6T so it's even better for Grok
English

Grok 4.3 scored 53 on the Intelligence Index at just 500B parameters.
For context: GPT-5.5 and Opus 4.7 are 1T+ parameter models scoring 57-60.
That means Grok 4.3 has roughly 1.5-2x better intelligence per parameter than every other frontier lab.
xAI is the most efficient AI shop on the planet.
The Colossus compute is paying off.

English

So why is mythos gatekept now?
AI Security Institute@AISecurityInst
OpenAI’s GPT-5.5 is the second model to complete one of our multi-step cyber-attack simulations end-to-end 🧵
English
Orbital Dev がリツイート

Here's a general summary of Elon Musk’s 3.5 hour long testimony today during day 2 of the OpenAI trial:
• Elon on Sam Altman & Greg Brockman: “If they wanna get rich, they should go do so as a for-profit. They should not get rich off a nonprofit. I gave them $38 million of essentially free funding to create what would become an $800 billion company.”
• Elon says he could’ve made OpenAI for-profit but chose nonprofit for AI safety
• Wanted initial ~51% control due to funding, but expected dilution over time
• Donated $10M+ and trusted assurances OpenAI would remain nonprofit
• Elon says his view of OpenAI evolved in three stages: first, he was “enthusiastically supportive.” Then, he became “a little uncertain” about whether the founders would keep it nonprofit. Now, he says he’s convinced “they’re looting the nonprofit.”
• Elon says he “I would have filed the lawsuit sooner if I thought they had stolen the charity sooner.”
• Left board in 2018 (Tesla/SpaceX commitments), still believing mission wouldn’t change
• Was not concerned by 2019 capped-profit shift at the time
• Elon feels the Microsoft partnership was a “bait and switch” and risked corporate control of AGI. When Microsoft invested $10B in OpenAI at the end of 2022, Elon sent a text to Sam asking, “What the hell is going on?” Elon became very concerned about OpenAI’s direction. Sam Altman responded to Elon's text with: “I agree this feels bad. We offered you equity when we established the cap profit, which you didn't want at the time. We are still very happy to do any time you’d like.”
Elon says "I didn’t understand how you could have stock in a nonprofit. It just didn't seem to make sense to me. Frankly, it felt like a bribe."
• Rejected equity offer from OpenAI, saying it didn’t make sense for a nonprofit, later called it a “bribe”
• Says he sued only after the “charity was actually violated,” not earlier concerns
• Denies seeking permanent control; says goal was ensuring AI goes in the “right direction”
• Early OpenAI relationships were positive, later deteriorated
• Elon criticized OpenAI as “not open” and effectively controlled by Microsoft
• OpenAI attorney asked Elon about the potential upcoming SpaceX IPO. Elon said that it would be illegal for him to comment on it before it actually goes public. When asked if he will control SpaceX after the IPO, Elon said: “Yes" and said he will still be CEO.
• Elon’s last $5M quarterly donation to OpenAI was in May 2017; he stopped covering office rent in 2020.
• OpenAI attorney on suggesting that xAI’s Grok lags far behind ChatGPT. Elon said, “Well, not anymore.” The OpenAI attorney asked, “It’s catching up, you think?” to which Elon replied, “Yeah.”
Elon will continue his testimony tomorrow.

English

@nicdunz but there are many types of benchmarks, text is quite useless at this level
English

designers who used to charge $5,000 for this right now:
Jacob Rodri@jacobrodri_
bro graphic designers are so cooked just generated these app store screenshots with appkittie.com in literally 5 minutes WTF
English

@StarkXInc @iamgingertrash yh atleast I can get off the internet and live good
seeing your profile picture internet is your only escape
English

@sguestsett @iamgingertrash “Animal weirdo”
Get off the fuckin internet you Polish man 🤣
How the fuck do they even still have internet in Poland LMAOOO
English

A Chinese company facing chip restrictions can train this
But xAI can’t even get SOTA
With a million H100 equivalents
Xiuyu Li@sheriyuo
Welcome DeepSeek V4 Pro Max huggingface.co/deepseek-ai/De…
English

1. stop talking you animal weirdo
2. they currently train 7 models simultaneously including a 10T one
3. Don’t expect shit to happen in seconds xAI is four times younger than OpenAI
4. They are #1 in hallucination rate likely meaning they spend lots of time on post training but you have prolly no clue what I’m saying anyway
English

@sguestsett @iamgingertrash No one gives a shit, should have trained a larger model then?!! Lmao What kind of fuckin excuse is that 🤣 nigga you got all the compute in the world, if you’re training a 3x smaller model and that’s why ur shit then that’s COMPLETE COPE
English

@Linahuaa What if that’s only Chinese propaganda?, not sure it is, but it sounds reasonable to me
English

Number of 150+ IQ people in China is more than 4 times the USA because Chinese average iq is 107 vs 98.
It's closer to 30 times more.
And the ratio of 150+ IQ people getting whipped into studying hard and going into STEM is also significantly higher.
Holtern@Skull_Kid_87
@Linahuaa Its simply a number game in the end. China has 4 times as many people with 150 iQ than the U.S.
English

@sguestsett @iamgingertrash if they turn it completely off,
they will reach 0 hallucination rate
English

@jerzydejm @iamgingertrash The lowest hallucination rate and real time data from X and 20 times bigger model is yet to come
xAI is also 4 times younger than OpenAI
Not mentioning terafab project and the biggest claster
English







