Director of RedRealities, Trekkie, History fan
Brisbane, Queensland
Joined March 2009
- Tweets2.8K
- Following363
- Followers157
- Likes3.4K
❌ Alex Chernov ❌ retweeted
Crowdstrike has broken the cardinal rule of development, never push updates on a Friday
❌ Alex Chernov ❌ retweeted
One of the reasons we need 'experts' is the gullibility & ignorance of non-experts. This latest poll, showing 75% of Russians support the war, is worthless. Under conditions of political terror & zero social trust, only loyalists & the brave will reply truthfully to a pollster.1/
❌ Alex Chernov ❌ retweeted
My heavily modified Linux 6.1 running on a 6x6mm chip which only has 8MB RAM.
It would be a lot easier if people use RAM wisely.
❌ Alex Chernov ❌ retweeted
Лёша, спасибо тебе за 26 лет абсолютного счастья. Да, даже за три последних года счастья. За любовь, за то, что всегда поддерживал, за то, что смешил даже из тюрьмы, за то что ты обо мне всегда думал.
Я не знаю, как жить без тебя, но я постараюсь так, чтобы ты там, наверху, радовался за меня и гордился мной. Я не знаю, справлюсь или нет, но я буду стараться.
Мы обязательно встретимся однажды. У меня столько для тебя нерассказанных историй, и у меня столько для тебя сохраненных в телефоне песен, глупых и смешных, вообще, честно говоря, ужасных песен, но они про нас, и я так хотела дать тебе их послушать. И так хотела смотреть, как ты их слушаешь, смеешься, а потом обнимаешь меня.
Люблю тебя вечно. Покойся с миром.
Wow, this is just mind blowing
Neural Network Diffusion
Diffusion models have achieved remarkable success in image and video generation. In this work, we demonstrate that diffusion models can also generate high-performing neural network parameters. Our approach is simple, utilizing an autoencoder and a standard latent diffusion model. The autoencoder extracts latent representations of a subset of the trained network parameters. A diffusion model is then trained to synthesize these latent parameter representations from random noise. It then generates new representations that are passed through the autoencoder's decoder, whose outputs are ready to use as new subsets of network parameters. Across various architectures and datasets, our diffusion process consistently generates models of comparable or improved performance over trained networks, with minimal additional cost. Notably, we empirically find that the generated models perform differently with the trained networks. Our results encourage more exploration on the versatile use of diffusion models.
❌ Alex Chernov ❌ retweeted
A client of mine building a world class data/information operation said -
"AI won't replace people. People who use AI will replace people who don't use AI."
This is a great illustration of that concept.
I think at this point in the game there is only one country that can defeat #USA realistically. It’s USA itself. Through civil conflict and potential infighting their global standing and network of alliances could be destroyed or degraded.
Wouldn’t be a first time in history.
❌ Alex Chernov ❌ retweeted
When I was a little boy, the Japanese military attacked Pearl Harbor. It was a surprise attack, and thousands of U.S. servicemembers perished. As a nation, we were stunned. And we vowed to strike back. Revenge was understandably on everyone’s mind, including many Americans of Japanese descent who opposed the emperor and were peaceful and law-abiding U.S. citizens and residents.
In its zeal to exact that revenge, however, the U.S. government overreacted, out of fear and bigotry. They targeted everyone who happened to look like the people who had carried out the attack. Those of us who had done nothing wrong were forced to pay the consequences for the decisions of others far away and disconnected from us. We were interned for years, in open-air prisons, while America went off to fight Japan, Germany and Italy.
It’s so important that we carry the lessons of the past through to today. Merely because one group commits atrocities and acts with depravity does not mean vast hundreds of thousands or even millions of others should be lumped together with them and made to suffer. We must never paint with the brush of justice and retaliation too broadly, or the toll of human suffering will rise immeasurably.
❌ Alex Chernov ❌ retweeted
With many 🧩 dropping recently, a more complete picture is emerging of LLMs not as a chatbot, but the kernel process of a new Operating System. E.g. today it orchestrates:
- Input & Output across modalities (text, audio, vision)
- Code interpreter, ability to write & run programs
- Browser / internet access
- Embeddings database for files and internal memory storage & retrieval
A lot of computing concepts carry over. Currently we have single-threaded execution running at ~10Hz (tok/s) and enjoy looking at the assembly-level execution traces stream by. Concepts from computer security carry over, with attacks, defenses and emerging vulnerabilities.
I also like the nearest neighbor analogy of "Operating System" because the industry is starting to shape up similar:
Windows, OS X, and Linux <-> GPT, PaLM, Claude, and Llama/Mistral(?:)).
An OS comes with default apps but has an app store.
Most apps can be adapted to multiple platforms.
TLDR looking at LLMs as chatbots is the same as looking at early computers as calculators. We're seeing an emergence of a whole new computing paradigm, and it is very early.
❌ Alex Chernov ❌ retweeted
This is an artificial womb that successfully grew lamb fetuses and could help prematurely born babies
[read more: buff.ly/2FRy8Zs]
❌ Alex Chernov ❌ retweeted
That was a point well made by @Nuclear4Climate in a talk he gave last night.
Australia has to remove the ban on nuclear power or we will end up as the quarry of the world with our environment destroyed.
If we want manufacturing, and jobs, we need clean, cheap power.
Пригожину с его 50к+ закаленными в боях, надо брать Москву. Все сейчас в Украине. 5к оставить устраивать показуху в точке базирования, 5к блокировать ЖД и дороги, 40к кремль, белый дом и Останкино. Путина казнить или под суд. Сам диктатор.
Я этого не хочу, но самый выигрышный шаг
Если просто продолжить бухтеть громко то точно выпилят. А сейчас на росте популярности у определенных слоев, централизации власти и минимуме оборону самое время брать риски. В худшем случае он начнет гражданскую войну и хаос в котором ему будет сильно проще выжить.
