This is interesting as a first large diffusion-based LLM.
Most of the LLMs you've been seeing are ~clones as far as the core modeling approach goes. They're all trained "autoregressively", i.e. predicting tokens from left to right. Diffusion is different - it doesn't go left to right, but all at once. You start with noise and gradually denoise into a token stream.
Most of the image / video generation AI tools actually work this way and use Diffusion, not Autoregression. It's only text (and sometimes audio!) that have resisted. So it's been a bit of a mystery to me and many others why, for some reason, text prefers Autoregression, but images/videos prefer Diffusion. This turns out to be a fairly deep rabbit hole that has to do with the distribution of information and noise and our own perception of them, in these domains. If you look close enough, a lot of interesting connections emerge between the two as well.
All that to say that this model has the potential to be different, and possibly showcase new, unique psychology, or new strengths and weaknesses. I encourage people to try it out!
The critical bug has been fixed. The individual security researcher who discovered and reported it is now under significant pressure. His years-old Twitter, Telegram, and LinkedIn accounts were all deactivated, and he remains worried about being sued. How can this be?
sending my thoughts and prayers to the investment partners that have to explain why their portco hacked an american exchange, stole 3m dollars, and laundered it through an ofac-sanctioned protocol
What if CertiK employees must adhere to the company's designated workflow? Now that the critical bug has been fixed, the individual security researcher who helped you discover and report it is under great pressure due to your accusations made without knowing the truth.
This individual discovered the bug in our funding system, and leveraged it to credit their account with $4 in crypto. This would have been sufficient to prove the flaw, file a bug bounty report with our team, and collect a very sizable reward under the terms of our program.
At the recent IC3 (@initc3org) event, Suning (@msfew_eth) from @SecurFi introduced ODESEC.
Through this on-chain database and its corresponding mechanism, ODESEC enables projects to filter out spam, promptly receive valid emergency vulnerability reports, and respond accordingly.
Thrilled to announce that ODESEC (Onchain Database of Emergency Security Event Contact) has won the ‘@RiscZero - Best zkVM Application’ at the Scaling Ethereum 2024 Hackathon hosted by @ETHGlobal.
🧵⬇️
Satya wins.
Reflexes of a startup CEO.
Resources of a trillion dollar company.
Pulls this together in 48 hours from a cold start.
Gets it signed and over the line before markets open.
Twitter is now the king of social media.
But 99% don't know about its "Advanced Search" feature.
10 Twitter Advanced Search tips that will make you a pro! (with examples):
These diagrams are the best explainers of the Ethereum block and state structure I have ever seen.
The first part needs to be updated for PoS, but it is still informative.
Credits to StackExchange user Lee: https://t.co/J4pnxRM1gn
Today marks the 100th anniversary Lee Kuan Yew's birth, the founding father of modern Singapore. He transformed Singapore "From Third World to the First." One the greatest statesman of the 20th century, it's mind-blowing when you compare LKY to politicians like @GaryGensler, who have no vision for the future, build nothing, attack innovative industries, and are turning America "From First World to the Third."
But it's ok, 2024 is going to be the year of reckoning for the anti-crypto, anti-business, anti-progress Democratic party and they will have their pal Gary to thank for it. Onward! 🚀