After the rise of the Large Language Model, developers can do what most NLP engineers are doing now, such as text classification, entity extraction, and reasoning, as long as they use the large language model. It is foreseeable that with the continuous improvement of LLM capabilities, it may do better than NLP engineering.
And the SAM (Segment Anything Model) released by Meta also brings CV to an end. SAM makes Engineering out of the box to the extreme. Before the advent of SAM, for midstream and downstream practitioners, to complete a segmentation task with almost no academic significance required at least: 1. A lot of image annotation work 2. A machine with reasonable computing power. After the advent of SAM, the most basic segmentation tasks can be achieved with zero annotation or even zero code. And there will be stronger algorithms and models than SAM in the future.
Therefore, NLP engineers and CV algorithm engineers may not exist in the future.
Just like there are no Office engineers now.
LLM and other open source algorithms have narrowed the gap between large and small factories, and made it possible for small and medium-sized enterprises whose data and computing power cannot compare with large factories to have the possibility of mechanism innovation. Developers without any algorithm background can connect to the algorithm work as seamlessly and smoothly as possible.
Creativity + engineering ability (executive ability) can produce great products.
Therefore, the next thing BitEagle has to do is to use LLM to build its own disruptive products.
What we have to do is to keep up with the latest developments in the industry: open source project developments, large company product developments, awesome start-up company ideas, and conduct our own product research and development.
At present, in the field of deep learning, scale training has peaked, but vertical training is booming.
Standing on the shoulders of giants, you can get twice the result with half the effort.
resources to focus on
ChatGPT Application Development Guide
Wu Enda's Prompt course
Microsoft's Deep Speed Chat
open source project LangChain
langchain-ChatGLM : ChatGLM and other large language model applications based on local knowledge base
ChatGLM-6B
open source project ChatGLM-Tuning
Wenda : an LLM calling platform. Find and design automatic execution actions for the plug-in knowledge base of small models, and realize the generation ability no less than that of large models
HuggingGPT
uses ChatGPT as a controller to connect various AI models in the HuggingFace community to complete multimodal complex tasks
https://mp.weixin.qq.com/s/Uo2jgou3Wv0PRDETCpodfw
https://huggingface.co/spaces/microsoft/HuggingGPT
https://github.com/microsoft/JARVIS
Forever Voices
has created AI female Internet celebrities , AI Jobs and other products
vector database
https://www.pinecone.io/
https://milvus.io/
https://zilliz.com/