AI Infrastructure Landscape
ai-infra.fun
ai-infra.fun
Huggingface provides model hub and also HF space for inference & deployment.
The landscape generator tool supports tag-based filter. I will open an issue to keep track of it.
disclaimer: maintainer of the landscape https://github.com/tensorchord/ai-infra-landscape
categories and lists and blogs about anything just for the sake of it
AI, OTOH, can provide meaningful value in most industries. Lump them together at your own peril.
However, while definitely overhyped and its capabilities far exaggerated, its applications already in several fields have been significant. I think the “gen AI” thing is pointless, but I got an eerie feeling watching 4 fairly senior engineers the other day huddled around a chatGPT terminal they were asking questions.
To me it felt like some profound moment, like maybe how the guys that went around manually lighting street lamps felt when they first saw electric lights.
Anyway, the answer usually falls in the middle somewhere between hype and doomsaying. It’s improved my workflow a bit. Not too worried that it’ll replace me, but I do believe there will be less work because of it, and to management that usually means someone’s getting fired.
instant frat-guy funding.. they have networks of insta-money. There is a cult in SF connected to Seattle..
meanwhile, thousands upon thousands of competent coders are looking for a work gig.. the price for skilled coders is dropping in almost all categories..
He describes how scientific revolutions aren’t really incremental progress on previous theories. Think Galileo’s theory earth revolved around the sun - that destroyed the conventional understanding of astronomy (and maybe even physics?). After these revolutions there is a flood of “normal” science that fills in the gaps based on these new big axioms.
Now, blockchain nor AI are scientific revolutions by Kuhn’s meaning of the term. They are certainly “normal” operating within the established domains of statistics, computer science, cryptography etc. But I think they are an analogs . Major breakthroughs were made that gave us AI and blockchain and then afterwards there is a flood of (probably decreasingly small) extensions or applications of them.
I guess my point is that I don’t feel like anything is broken or there are bad incentives. Thats just how innovation works. You have massive progress followed by decreasingly useful work. And certainly in the case of AI there is more large scale progress to be made.
Open-source contributions can make it better. :-)
https://github.com/underlines/awesome-ml
But it lost a bit of traction lately.
It needs re-work for the categories, or better, a tagging system, because these products and libraries can sit in more than one space.
Plus it either needs massive collaboration, or some form of automation (with an LLM and indexer), as I can't keep up with it.
One of the things that is still TBD about this wave of ai companies: what are the pain points in building products that people actually want to use?
I guess my warning would be that, much like integrating an analytics platform a few years back, it's easy to spend a lot of time and money on someone else's idea of how to improve your business.
I feel like I've had to learn this lesson at every level of the stack. There's no ORM that will save you from thinking about database schemas. There's no analytics platform that will save you from thinking about what's the most important thing your product is. I suspect, there's no training or evaluation platform that will save you from trial and error; prompt and check techniques.
** Prompt: Is this classification correct: LLM Providers are { Hugging Face, Open AI, Anthropic }. Think step by step. Be very precise and return with YES or NO only for each Provider in the list.
Response: For each provider in the list:
Hugging Face: No, Hugging Face is not a large language model provider. They are a technology company that provides a platform for NLP and AI, offering a wide range of pre-trained models, including large language models, through their transformers library. OpenAI: Yes, OpenAI is a large language model provider, as they have developed several large language models, like GPT-2 and GPT-3. Anthropic: Yes, Anthropic is a large language model provider, as they have developed large language models, like Claude.
In summary, the classification "LLM Providers are { Hugging Face, Open AI, Anthropic }" is not entirely accurate. A more precise classification would be "NLP organizations/companies { Hugging Face }" and "LLM Providers { OpenAI, Anthropic }". **
About a year ago I toyed with writing a web app that was essentially a front-end to a diffusion image generator (very original I know). The site used socket-io -> flask -> redis queue -> distributed pytorch processes.
Am I correct that several of these services are selling some equivalent of the '-> redis queue -> model' component? Is part of the value proposition here that you don't need to staff people with pytorch/equivalent familiarity?
From the maintainer of the project.
https://landscape.lfai.foundation
They have graduated levels like CNCF :)
https://landscape.lfai.foundation
Would love to collaborate with the author of this as I help run CNCF/LF landscape infrastructure.
I don't see that on https://landscape.lfai.foundation/. Also, I don't see a space for Code or IDE tooling the Linux Foundation page. ai-infra.fun contains TabNine and Tabby.
Would things like TabNine and Tabby make sense on the Linux Foundation one? Would love to collaborate on this!
These startups are offering tools to maximize workflow efficiencies for high scalability deployments, but most of the world is still trying to understand WTF is a transformer.
A computer like device that can "generate text", "generate audio", "generate video" and also "train".
One for each category: audio, text, video, image.
One analyzer to coordinate everything.
This would be my API that I could access with mobile devices.
Here's a scenario:
I could talk to my phone about ideas, in the background it would create apps prototypes, create posters, make music based on something i whistle, teach me ask i ask question about a topic.
We could delegate the mundane stuff to it.
A processor has different cores, Computers may have several hard-drives, 4 sticks of ram.
Each component can run in parallel.for example, if a long video processing task is underway and your text generation component is idle, it could assist.
Should the audio component fail , only that specific part would be affected.
I'm surprised that Apple hasn't jumped on this yet. All the building blocks are in place to make a revolutionary device possible. It just needs to be very polished, and have great UX, which is what Apple typically excels at.
Now our smartphones are 1000 times more powerful than the ENIAC and use less power.
Do you think Apple likes to jump on things? Apple usually tries not to be first, but definitely likes to polish .
Still waiting on 'consumer' AI training facilities I can practically use: it's a bit arcane, not up to speed on that. I can generate text up to Llama 70b if I'm patient, and generate images and sort-of-video quite easily.
the main reason for me would be no need to pay and also no one censoring the content
Apologize for the allergic reaction to running something on somebody else’s computer. As much as I appreciate our connected world, I prefer my data and compute local and private.
There’s few things in this world that infuriate me on a daily basis more than a slow or lost internet connection.
Edit: grammar
I was commenting more on the idea to run AI in the cloud instead of on device.
from the guide:
Category 1 - Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat. Duis aute irure dolor in reprehenderit in voluptate velit esse cillum dolore eu fugiat nulla pariatur. Excepteur sint occaecat cupidatat non proident, sunt in culpa qui officia deserunt mollit anim id est laborum.
LAMP stack was easy to choose and onboard.
To whoever made this: Will you keep it updated as new stuff comes out?
EDIT: And it seems they just used this project as the base: https://landscape.cncf.io/