Update on the Security Incident
The attack has been mitigated, and we have temporarily shut down the application to isolate the security incident.
1. None of the connected wallets and CEXes were impacted.
2. Thanks to the immediate incident reponse from the CoinStats team, only 1.3% of all CoinStats Wallets were affected, totaling 1,590 wallets. The list might change as the investigation is ongoing but we don’t expect significant changes.
3. If your wallet address is in this affected list, please move your funds immediately using your exported private key (if you have exported previously): https://t.co/NhctjkpECt
4. We are actively investigating the extent of funds moved and will provide updates as soon as they become available.
We're actively working to bring the app back online as quickly as possible. Thank you for your patience.
Who’s hyped for the 1st $MENTO token airdrop? 🌐
The Mento Governance Token is live!
With the introduction of the MENTO token, it's time to bring local currency stablecoins to the masses...
Like, RT, & reply “MENTO” to get a secret code in your DM 👀
How to qualify 👇
My thoughts on Mistral's stellar rise:
- Successful setup: founded at good timing in the OSS vs closed AI debate, closed a $400M Series A at $2B valuation, and driven by a lean team.
- There're dozens of models coming out every month, but only a handful actually have staying power, and even fewer evoke a sense of yearning from the public. LlaMA and Vicuna are 2 other examples.
- I think what Mistral did right is to optimize the hell out of 7B-tier models, instead of much larger capacity. 7B and the 7B-MoE (equivalent to 12B dense) are much easier for grassroot AI engineers to build on.
- MoE is the right path forward. It flexibly trades off between knowledge/memorization and efficiency of a small model. OpenAI has been doing this for > 1 yr since GPT-4 training. I'm honestly surprised the AI community didn't route more attention to MoE.
- An LLM is a snapshot of a civilization. There will be a LOT more localized LLMs that represent different cultures, political spectrum, religious beliefs, and region-specific regulations. Mistral makes multi-lingual front & center. Not surprising as they're a French startup!
- The release itself was done with swag. It's literally the reverse order of what folks expect:
(1) Drop a magnet link with no explanation. Magnet is the new clickbait!
(2) Submit a pull request to the open-source vLLM project to help the community integrate the Megablocks CUDA kernels. Rad move!
(3) Only *then* you release the blog post.
- Releasing hosted API endpoint is the best way to quickly gather customer feedback, iterate on in-the-wild use cases, and most importantly, monetize OSS models. Mistral does it immediately.
- "Mixtral" is such a clever name 👏
Launch blog: https://t.co/bTYFj97Ydy
API platform: https://t.co/JNbSvEWetR
NVIDIA basically compressed 30 years of its corporate memory into 13B parameters. Our greatest creations add up to 24B tokens, including chip designs, internal codebases, and engineering logs like bug reports. Let that sink in.
The model "ChipNeMo" is deployed internally, like a shared genie:
- EDA scripts generation. EDA stands for "Electronic Design Automation", a core software suite for designing the next-gen GPUs. These scripts are the keys to a $1T market cap 🦾;
- Engineering assistant chatbot for GPU ASIC and Architecture engineers that understands internal hardware design specs and is capable of explaining complex design topics;
- Bug summarization and analysis as part of an internal bug and issue tracking system;
- Domain-finetuned retriever that achieves much better accuracy over internal knowledge.
And we publish a whitepaper to share ChipNeMo's creation process: https://t.co/dMG5AFMXTH
Official blog: https://t.co/MMjLbKxE34
Congrats to Haoxing "Mark" Ren's team for the outstanding work!
A team just made OpenAI Whisper 6x faster, 49% smaller, while keeping 99% of the accuracy.
The model is already available on the HuggingFace Transformers library:
model_id = "distil-whisper/distil-large-v2"
You can also use their web UI to transcribe from URLs, files, or audio recordings.
Model: https://t.co/I65gFW9cxw
Demo: https://t.co/BPEmo0i1wx
Paper: https://t.co/xpjbAZuQdr
@srush_nlp
💥 Happy #Bitcoin Whitepaper Day 💥
On this day 15 years ago, Satoshi Nakamoto released the 9-page "Bitcoin: A Peer-to-Peer Electronic Cash System" whitepaper on a cryptography mailing list - forever changing the world of finance 🙌
Yesterday 1Password released an Incident Response Report believing that when Okta was breached (again) the Threat Actor(s) tried to pivot to them.
They noted they used MalwareBytes™ FREE AV
BREAKING: BlackRock tells SEC it will seed its spot #Bitcoin ETF in October in preparation for approval.
576 ETF applications.
575 ETFs approved.
They know what they're doing... 👀