AI agents are still a pretty big topic in crypto, a lot of projects doing what you described. did you try https://github.com/ccxt/ccxt

Semantic Web is now revived into its new marketing incarnation, called Knowledge Graphs. There's actually a lot of work on building KGs with LLMs, specially in the RAG space e.g., Microsoft's GraphRag and llama_index's KnowledgeGraphIndex

For MIDI:

https://github.com/Music-and-Culture-Technology-Lab/omnizart and https://basicpitch.spotify.com/

They work better if you apply some source separation before (e.g, https://github.com/sigsep/open-unmix-pytorch, https://github.com/facebookresearch/demucs, or https://mvsep.com)

Still, I think the best results are from proprietary models (specifically https://www.ableton.com/en/manual/converting-audio-to-midi/ and https://www.celemony.com/en/melodyne/what-is-melodyne)

To train small gpt-like models, there's also aitextgen: https://github.com/minimaxir/aitextgen

As the creator of aitextgen, I'm mixed on continuing support since there doesn't seem to be as much demand as expected for small GPT models given the success and cost-effectiveness of GPT-3/ChatGPT, unfortunately.

I still have a few ideas there (including another secret approach at better text generation) but it's hard to determine ROI.

I think what you have created still has great demand. It give devs who do not have the budget or need for the gigantic models, something to train and use for their own specific language tasks.

Not everyone is trying to replicate CHATGPT results for certain tasks.

You can call any of the huggingface models [1] using their API [2]. A few examples:

- https://huggingface.co/EleutherAI/gpt-j-6B

- https://huggingface.co/t5-base

- https://huggingface.co/facebook/opt-66b

- https://huggingface.co/bigscience/bloomz-3b

There are also other companies offering large models as a service:

- https://www.forefront.ai

- https://nlpcloud.com

- https://www.goose.ai

- https://cohere.ai/generate

[1] https://huggingface.co/models

[2] https://huggingface.co/inference-api

I wouldn't have suggested those models. Just use a semantically fine-tuned BERT.

> GPT-3 Embeddings by @OpenAI was announced this week. I was excited and tested them on 20 datasets. Sadly they are worse than open models that are 1000 x smaller

https://twitter.com/Nils_Reimers/status/1487014195568775173

Get models here: https://sbert.net/docs/pretrained_models.html

Indeed. He managed to disrupt another industry, this time the fabric of space and time! What a time to be alive.

Ableton Live has a Audio to Midi [1] feature that works pretty well to extract notes from audio. Melodyne [2] is a very powerful tool to mess around with notes in general.

There's also some open source projects out there, such as aubio [3] and Omnizart [4].

[1] https://www.ableton.com/en/manual/converting-audio-to-midi/

[2] https://www.celemony.com/en/melodyne/what-is-melodyne

[3] https://aubio.readthedocs.io/en/latest/cli.html#aubionotes

[4] https://music-and-culture-technology-lab.github.io/omnizart-...

Check out https://github.com/fedden/RenderMan and https://github.com/DBraun/DawDreamer

thanks, I'll check these out!

I have been using a Cubot Mini (4") for sometime now (1+ year), and am pretty happy with the experience overall. Feels so good to have a phone that fits in any pocket.

https://cubot.net/Smartphones/king-kong-mini/8

I was interested until I saw Android 9. That seems to be the common problem with a lot of these older phones: dated software. We have a car with android auto integration that actually finally works well with h android 11 and better with 12. 9 and 10 were the reason we paid for the built in navigation in the infotainment system.

Just noticed they have a new version with android 10: https://cubot.net/Smartphones/KingKong-mini-2

I see visible screws on the backside.

Is it possible to screw a belt clip there? And I don't mean in a case, but physically attached to the back of the phone with screws?

Some Chinese work phones have that feature, and it's stupidly addictive to be able to clip your phone anywhere!

The screws have a very easy direct access, I can see them being repurposed for other means with a bit of effort.

Looks tempting... two questions: 1. How's the keyboard? I'm specifically wondering about this because at the taller aspect ration the screen is narrower than that of an iPhone 5. 2. How's the battery life? It's obviously not a phone that's designed for hanging out on social media all day, but I'd like to be able to make a few phone/WhatsApp calls without having to run for the nearest charger.

The keyboard is a little small, and does take a bit to get used. Not recommended for big fingers.

Battery life is pretty good; I charge it at night, and usually is half-battery. My usage is pretty decent, but mostly on wifi.

Looks as exactly the right phone for me. But the amazon comments say it would break after some month/weeks. I can imagine that happens when you treat is as a phone for a construction area, as it looks. But maybee it just dies because it's bad quality. Does anyone have experience with the phone?

Can you install AOSP/Lineage on that phone?

Seems it uses a Mediatek chip, so probably not.

:(

The field of Open Information Extraction has been trying to do that in a generic way for a long time, but the results are still far from good. A few references: OpenIE [1] Graphene [2] MinIE [3].

If you already have a Knowledge Graph (KG) and want to populate its instances from documents, that's called KG Population, and Knowledge-net [4] is a good reference.

Relation Extraction is another interesting approach if you know which kind of relations you're interested in, OpenNRE [5] a good example.

[1] https://github.com/dair-iitd/OpenIE-standalone

[2] https://github.com/Lambda-3/Graphene

[3] https://github.com/uma-pi1/minie

[4] https://github.com/diffbot/knowledge-net

[5] https://github.com/thunlp/OpenNRE