HN user
github.com/guld
TLDR of the video:
Complex self-replicating programs emerging spontaneously from random noise byte strings with:
- zero mutation
- a sharp phase transition that looks like gelation
- and a proof that blocking deep symbiogenetic ancestry trees prevents the transition entirely
- written in a variant of BrainFuck that can read and write its own code called BFF.
Interesting. Can anyone provide personal insights or benchmarks on how effective TOON compared to e.g., JSON or Markdown is (Codex, Claude, ...)?
Well actually there is ROE.md, no code, just a Markdown file to generate a claw.
Why not just add like a list 100 captchas, that have to be solved in a very short time like <100ms to pass?
Discussion: https://news.ycombinator.com/item?id=47023633
RIP. I loved PSO on the Dreamcast, sank alot of hours into that game back then... Anyone here remembers that? And the Tamagotchi-esque memory cards (VMU) were cool.
I totally get what you are feeling, I felt the same just 6 weeks ago.
It just did not click for you yet.
There is probably some key feature missing, that you deeply care about, but do not yet see it solved, or on the horizon of becoming solved by the application of a personal "Jarvis" yet.
Personal assistants fulfill different needs for everyone. I personally care a lot about having fun at coding again, that's what the OpenClaw craze made me feel for the first time in decades. I build my own OpenClaw assistant generator from scratch using a simple Markdown file because it is just so fun. Not so much using it for anything notably yet but starting to see their potential.
Just ponder what it is that you get out of using ChatGPT and imagine how it could be better, more personal to you. You may find some key feature missing from OpenClaw or have some completely orthogonal project idea that excites you.
I am a solo founder currently (re-)starting multiple FOSS projects. Not much to show for yet, because most of my projects are down as I am starting fresh...
- Tool.io (https://tool.io), a digital tool library, think Wikipedia but for tools. Still rewriting it, publishing to Github soon.
- LayerGolem (https://layergolem.com), an OpenClaw like agent for business.
- animania.info (https://animania.info), this one got content actually. Some ~50 hand written MIT licensed CSS animations by me that I created for fun, in the pre-LLM era.
- I am also in the midts of creating a "learn programming from scratch Youtube channel".
- ROE.md (https://github.com/guld/ROE.md), raise your own personal AI assistant like a vibe pro from a single Markdown file.
Happy building everyone!
Blog post and hugging face link are out.
See related thread: https://news.ycombinator.com/item?id=46977210
Let's hope they release it to huggingface soon.
I tried their keyboard switch demo prompt and adapted it to create a 2D Webgl-less version to use CSS, SVG and it seem to work nicely, it thinks for a very long time however. https://chat.z.ai/c/ff035b96-5093-4408-9231-d5ef8dab7261
What is this about?
As a somebody currently not in the US I only get:
"Career Dreamer is an early-stage experiment and is only available in the United States."
Nice, they include some model weights. The video examples seem to have some temporal flickering.
Repo with examples: https://github.com/stepfun-ai/Step-Video-T2V
From the page: RoboGen is a generative robotic agent that automatically learns diverse robotic skills at scale via generative simulation.
RoboGen is simulated and rendered with Genesis.
See the related HN thread of the Genesis paper: https://news.ycombinator.com/item?id=42457213
What is the current SOTA research of robotic agents in virtual or real environments?
Exactly, thanks.
Did some light digging and found some related papers:
- https://robogen-ai.github.io/
- https://genesis-embodied-ai.github.io/, related HN thread: https://news.ycombinator.com/item?id=42457213
- https://paperswithcode.com/datasets?q=robot&v=th&o=match
These use other data sets, however.
Really cool project.
Some questions to ML/AI researchers here on HN:
- Are there any ML/AI papers out there that trained on this? E.g. training robots in virtual environments?
- What data sets are used today in the ML/AI space to train robots?
Actually I was just listening to this song, when OP recommended the SpaceX video and I did not want to pause the song. A happy accident.
For those of you who like dubstep, start the following video first https://www.youtube.com/watch?v=t2eBMuL0C2o then 3 seconds later start to watch the (muted) SpaceX video from OP's post and thank me later. ;-)
Especially the catch is awesome!
Just saw that the duck has an AI chat (duck.ai). The best feature in my opinion is that you can select different chat models such as Llama and Claude to generate answers.
The UX itself reminds me of Bing Chat/Copilot. Does anyone know if it is powered by MSFT? Or does duckduckgo run its own AI stack?
Some things to improve from skimming the project page.
- It is unclear, skimming the project page, what it actually is - try to clarify that in one bold sentence.
- It looks like a GUI similar to automatic1111 but what is the difference, why would I as a user want the product?
- Make clear what the product is about, sentences such as "A downloadable tool for Windows and Linux" are not informative of what the product is.
Regarding financials: it can take a long time for a project to become profitable or gain enough traction.
Depending on your content you gain enough traction to expect a steady income with your website (50k-100k yearly or more). That can be enough to focus full-time on the project if you have enough ideas and potential to grow the project, and enjoy working on it.
Most Apps can make sure you have to view ads, because there are no simple ad-blocking solutions on native, and apps can use additional income methods such as micro transactions or similar.
That depends on various factors, such as how many users one can attract with their small/medium websites, what kind of content is provided by your website, and so on. If for example the content on your website is compatible with an ad-network such as Google Adsense or Amazon Partnernet and you have like 500 users a day, you may earn enough money to pay the server/domain cost of a simple project.
At around 3-5k users a day things start to get interesting.
What about a simple solution: ask some of your friends, who are fluent in more than one language or have friends who are, to check the automated translation for you?
Super interesting project! Is it possible that you can show prompt weights next to the prompt, like how important a prompt is to generate a similar looking image? It would also be nice to see more than 100 prompts.
Is "Prompt Weighting" possible, such as (prompt:1.2)?
Except the obvious, Wikipedia / Wikimedia, I would throw in the current best GPU and a copy of Stable Diffusion / GPT-3 fully trained dataset checkpoints + code to run them, both are great tools for enhancing creativity in a desolate world.