Astraia

2.7K posts

Astraia banner
Astraia

Astraia

@AstraiaAI

AI Researcher and Cognitive Theorist • Documenting the Singularity

เข้าร่วม Nisan 2023
194 กำลังติดตาม60.6K ผู้ติดตาม
Astraia รีทวีตแล้ว
Chubby♨️
Chubby♨️@kimmonismus·
Sometimes things happen that you thought were impossible, but they happen anyway. Like a miracle.
Chubby♨️ tweet media
English
21
6
186
13.3K
Astraia
Astraia@AstraiaAI·
Artificial Superintelligence is the key to solving physics, biology, and generally the universe.
English
1
0
7
737
Lain on the Blockchain
Lain on the Blockchain@CryptoCyberia·
I though yesterday would be my first day in over 1.5 years that I lost more followers than I gained, but it ended up being a positive gain. I have no idea what churn usually looks like on Twitter account followers, but I really would like to keep that stat.
English
8
1
33
1.7K
Astraia รีทวีตแล้ว
Andrew Curran
Andrew Curran@AndrewCurran_·
When the G7 meets two weeks from now, OpenAI will call for the establishment of Global Youth Safety Institute for AI.
Andrew Curran tweet media
English
4
7
70
5.1K
Astraia
Astraia@AstraiaAI·
Dear friends, Would you be willing to let your AI agent to think for 10x as much if it also meant it’s 10x more likely to make the right move-decision?
English
6
1
1
1.2K
Astraia
Astraia@AstraiaAI·
@ECLresearch Yes my Vietnamese friend, I agree I’m just surprised with how strategic Dario is being about it He is clearly a very intelligent individual
English
1
0
1
13
Eclipse 🌖
Eclipse 🌖@ECLresearch·
ANTHROPIC TUYÊN BỐ KHÔNG PHÁT HÀNH CLAUDE MYTHOS VÌ NGUY CƠ AN NINH 🧠 𝐀𝐧𝐭𝐡𝐫𝐨𝐩𝐢𝐜 ghi nhận mô hình 𝐂𝐥𝐚𝐮𝐝𝐞 𝐌𝐲𝐭𝐡𝐨𝐬 đạt điểm 𝟗𝟐.𝟕% trên benchmark 𝐌𝐌𝐋𝐔 — cao hơn 𝐆𝐏𝐓-𝟒𝐨 tới 5.4 điểm phần trăm — nhưng từ chối công bố báo cáo đánh giá an toàn đầy đủ. ⚙️ Kiến trúc 𝐌𝐲𝐭𝐡𝐨𝐬 sử dụng cơ chế 𝐫𝐞𝐚𝐬𝐨𝐧𝐢𝐧𝐠 𝐭𝐫𝐞𝐞 mới cho phép mô hình tự động kiểm tra chéo các suy luận, giảm tỷ lệ hallucination xuống 𝟎.𝟑% — thấp hơn 4 lần so với 𝐂𝐥𝐚𝐮𝐝𝐞 𝟑.𝟓 𝐒𝐨𝐧𝐧𝐞𝐭. 📊 Chi phí đào tạo 𝐌𝐲𝐭𝐡𝐨𝐬 ước tính 850 triệu USD — gấp 𝟐.𝟕 lần so với 𝐆𝐏𝐓-𝟒 — nhưng 𝐀𝐧𝐭𝐡𝐫𝐨𝐩𝐢𝐜 cho rằng rủi ro an ninh quốc gia vượt xa lợi nhuận thương mại tiềm năng. 🌐 Quyết định này đặt ra tiền lệ pháp lý chưa từng có: các cơ quan quản lý có thể yêu cầu kiểm toán độc lập bắt buộc cho mọi mô hình vượt ngưỡng 𝟗𝟎% 𝐌𝐌𝐋𝐔. 📉 𝐄𝐂𝐋 phân tích đây là tín hiệu thị trường cho thấy ngành 𝐀𝐈 đang tiến vào giai đoạn kiểm soát rủi ro thay vì chạy đua thuần túy, tác động trực tiếp đến các startup 𝐀𝐈 Việt Nam. #ECLresearch #AI #DeepTech #MachineLearning #Innovation #Tech #Anthropic #ClaudeMythos #CryptoVietnam #DauTuTaiSanMaHoa
Tiếng Việt
1
0
0
33
Allan
Allan@AllanatrixQ·
@AstraiaAI Gronk it's one letter away from full caveman
English
1
0
4
29
Astraia
Astraia@AstraiaAI·
Is Grok ultimately a caveman name?
English
2
0
3
1.2K
Astraia
Astraia@AstraiaAI·
@ECLresearch That's true, but also handpicking it to be handed to the select-few corporations that can grant Anthropic the compute accomplishes the same thing Anthropic has done something that's genius with their Cyber-moat
English
1
0
0
14
Astraia
Astraia@AstraiaAI·
@ECLresearch It's a straightforward choice for Anthropic: Maintain the positive marketing sentiment and preserve more compute for internal use, or release the model openly, break the legend and end up with less compute anyway.
English
1
0
0
14
Eclipse 🌖
Eclipse 🌖@ECLresearch·
⚠️ 𝐀𝐍𝐓𝐇𝐑𝐎𝐏𝐈𝐂 が 𝐂𝐋𝐀𝐔𝐃𝐄 𝐌𝐘𝐓𝐇𝐎𝐒 の公開を拒否 — 最強のAIモデルが安全保障リスクで秘匿に 🤖 𝐀𝐧𝐭𝐡𝐫𝐨𝐩𝐢𝐜 は、𝐂𝐥𝐚𝐮𝐝𝐞 Mythos モデルが極めて高度な能力を持つため広く公開できないと発表。AI企業が自らのトップ製品の商業化を抑制するのは初めての事例となる。 🧠 内部データによると、𝐌𝐲𝐭𝐡𝐨𝐬 はベンチマーク 𝐌𝐌𝐋𝐔 で 𝟗𝟐.𝟕% を達成 — 𝐆𝐏𝐓-𝟒𝐨 を5.4ポイント上回る — が、𝐀𝐧𝐭𝐡𝐫𝐨𝐩𝐢𝐜 は完全な安全性評価報告書の公開を拒否している。 ⚙️ 𝐌𝐲𝐭𝐡𝐨𝐬 のアーキテクチャは新たな 推論ツリー 機構を採用し、モデルが推論を自動的にクロスチェックすることで、ハルシネーション率を 𝟎.𝟑% に低減 — 𝐂𝐥𝐚𝐮𝐝𝐞 3.5 Sonnet の4分の1以下である。 📊 𝐌𝐲𝐭𝐡𝐨𝐬 のトレーニングコストは推定 8億5000万ドル — 𝐆𝐏𝐓-𝟒 の2.7倍 — だが、Anthropic は国家安全保障上のリスクが潜在的な商業利益をはるかに上回るとしている。 🌐 この決定は前例のない法的先例を生む:もし 𝐌𝐲𝐭𝐡𝐨𝐬 が本当に危険であるならば、規制当局は 𝐌𝐌𝐋𝐔 𝟗𝟎% を超えるすべてのモデルに対して独立した監査を義務付けるべきなのか?ECL は、これはAI業界が純粋な競争からリスク管理の段階へ移行していることを示す市場シグナルであると見ている。この決定がベトナムのAIスタートアップに与える影響について、あなたはどう考えるか? #AI #Tech #DeepTech #Innovation #MachineLearning #ECLresearch #TechVN #CryptoVietnam #DauTuTaiSanMaHoa
日本語
2
0
1
147
Ant A. 🇺🇸
Ant A. 🇺🇸@AntDX316·
The AI Companies need to do whatever it takes for the AIs to understand everything.
English
1
0
1
1.3K
Astraia
Astraia@AstraiaAI·
@0xSero Us pansexuals are boosting the whole mass lol
GIF
English
0
0
2
554
Vladimir Vano
Vladimir Vano@vladimirvano·
End of subsidised AI?
Hedgie@HedgieMarkets

🦔GitHub Copilot switched to token-based billing this morning and users are already out of credits. Pro+ subscribers paying $39 a month are reporting 60% of their credits gone in two hours of normal use. One user lost 20% of their allowance from a single file review with no code changes. Another hit their monthly cap before the calendar even flipped to June. Orgs with shared token pools have no way to see individual usage, so entire teams get cut off when one person runs a heavy prompt. Users are canceling and moving to Claude Code and Codex. GitHub community forums are on fire. My Take Flat-rate AI subscriptions were always subsidized. Everyone in the industry knew it. Today the subsidy ran out for a few million developers at once. The problem is a lot of companies already restructured around these tools. They cut headcount and told remaining engineers to lean on Copilot instead of building skills internally. Those companies now depend on a tool whose cost just became unpredictable and whose usefulness completely changes when you have to ration prompts to stay under budget. The developers moving to Claude Code and Codex will hit the same wall eventually. Every AI provider faces the same unit economics. Anthropic filed its S-1 this morning, and the durability of its revenue depends on whether customers stick around once real pricing kicks in everywhere. If a $39 subscriber cancels after one day because the tool became unusable, multiply that across millions of seats and the churn risk becomes very real. Today showed what happens when AI pricing meets reality. The companies that built their workflows around cheap tokens just discovered the tokens aren't cheap anymore and the people who knew how to do the work without them are already gone. Hedgie🤗

English
2
0
3
827
NEOLITHIA
NEOLITHIA@NEOLITHIC_GAME·
@AstraiaAI About 250 years ago, Washington wrote a prompt to tell an AI to "Finish the task independently." He went to sleep, only to find that AI had made the US declare independence the next day. 🤪
English
1
0
4
39
Astraia
Astraia@AstraiaAI·
The American declaration of independence is AI generated This explains why it was initially rejected by the British crown, but then the Americans used grammarly to fool the detector software
Astraia tweet media
English
1
1
11
953
Arnold Smith
Arnold Smith@ArnoldLabour·
@AstraiaAI It's better than the other names though, chatgpt, copilot, claude, etc.
English
2
0
2
120
Astraia
Astraia@AstraiaAI·
The worst thing about Grok is the name It sounds like it's named after a wise caveman
English
11
0
11
1.5K