CITIC Capital's research note points out that Pangu's pre-trained big model capabilities have now been validated in areas including energy, retail, finance and industry. Meanwhile, Pangu NLP Big Model achieves the goal of less sample learning through migration learning, incorporates the latest technologies such as P-tuning into Pangu's fine-tuning framework, and is trained through a huge industry knowledge base, making it more suitable for applications in complex commercial scenarios with stronger capabilities than GPT-3.
But those who want to wait for toC can look forward to nothing more.
It is the level of GPT-3 only, the target customers are set to central enterprises and Yan4Fu, and toB is the only way out.
Interesting stuff! It would be interesting to see whether large language models could also help explain certain strategies or interface with traditional financial techniques. For instance, you might be able to have the language model interface with traditional portfolio optimization techniques using something like toolformer or differentiable portfolio optimizers (https://github.com/cvxgrp/cvxpylayers).
FinGPT is on the horizon, but not much emphasis is being placed on what information is going to be the most valuable. Exploration into the working domains of AI-powered financial models needs to be increased and shared for better understanding of how these black boxes work. Garbage in gets garbage out, so it’s important to not be blinded by the results of small scale tests and continued commitment and focus should be spent in domain research/exploration.
I was intrigued by a use case of analyzing financial documents like SEC filings (10-k,10-Q). Like in this video https://youtu.be/UzUJzdn7c2w, Prof. Aswath Damodaran teaches how to read these documents and value of company. It was mostly information retrieval. With the LLMs, we can automate this process and get a valuation of the company.
SuperSymmetry Technologies has collaborated with Fudan University to release BBT-2-Finance large language model. It trains on a large amount of Chinese financial corpus data to help users extract key information from lengthy texts and write summaries.
https://arxiv.org/pdf/2302.09432.pdf
I guess it will. Financial big data has various sources, twitter, yahoo, meta, papers, etc. These data helps for building a large language model, and may help investors for making decisions. It also will encounter challenges, such as processing picture/speech/video data
I hope it helps. Duxiaoman develops an ai for financial platform, which applies large language model LLM in the interpretation of credit report.
https://www.duxiaoman.com/
1) BloombergGPT:
a) Introducing BloombergGPT, Bloomberg’s 50-billion parameter large language model, purpose-built from scratch for finance:
https://www.bloomberg.com/company/press/bloomberggpt-50-bill...
b) BloombergGPT: A Large Language Model for Finance:
https://arxiv.org/abs/2303.17564
2) Stochastic AI:
a) xFinance: How to outperform BloombergGPT without breaking the bank:
https://www.stochastic.ai/blog/xfinance-vs-bloomberg-gpt/
b) xTuring website:
https://xturing.stochastic.ai/
c) xTuring: Easily build, customize and control your own LLMs:
https://github.com/stochasticai/xturing