background
Chagpt3|4 is very popular recently, causing many people in the industry to think that strong artificial intelligence will soon replace us engineers. According to our understanding, artificial intelligence built with large language models can indeed replace some jobs in the future, but does this mean that it can be replaced? slow down?
What is a large language model?
Large Language Models (LLM), full name: Large Language Models, LLMs, represent a major advancement in the field of AI and are expected to change the field through acquired knowledge. The size of LLMs has increased by a factor of 10 per year over the past few years, and their performance has continued to evolve as the complexity and size of these models have increased. However, the development and maintenance of LLM is not easy, which makes it impossible for most enterprises to use LLM. --- NVIDIA
Personal understanding: Similar to communication between different countries, communication is done through translation. The big language model is the translation in the middle of the type, which collects and abstracts the existing human language and characters into its own set of models. This model can be continuously improved, can understand the relationship between various languages and semantics, and is based on the current civilization. The underlying logic is automatically learned and iterated, and through data + manual assisted improvement to achieve strong artificial intelligence that can be understood and automatically learned.
What are the big language models?
type |
describe |
Remark |
GPT-3|4 (Generative Pretrained Transformer 3|4) |
GPT-3|4 (Generative Pretrained Transformer 3|4) developed by openApi. chatGPT is the product |
GPT3 parameters reached 175 billion, and GPT4 parameters reached 3.5 trillion~. Reference: https://openai.wiki/news-gpt4model-317.html |
BERT(Bidirectional Encoder Representation from Transformers) |
The Encoder of the two-way Transformer was developed by Google, because the decoder cannot obtain the information to be predicted. |
Reference: https://baijiahao.baidu.com/s?id=1651912822853865814&wfr=spider&for=pc |
XLNet (eXtreme Language understanding NETwork) |
The pre-trained language model jointly developed by CMU and Google has 150 million parameters and is capable of performing natural language processing tasks such as text classification, question answering, etc. |
Reference: https://zhuanlan.zhihu.com/p/147250033 |
ERNIE Bot(文心一言) |
A new generation of knowledge-enhanced large language model developed by Baidu |
reference https://baike.baidu.com/item/%E6%96%87%E5%BF%83%E4%B8%80%E8%A8%80/62642976 |
T5(Text-to-Text Transfer Transformer) |
Also a pre-trained language model developed by Google |
reference: https://zhuanlan.zhihu.com/p/89719631 |
... |
There are many more, you can learn by yourself |
at last
I personally think that the singularity of artificial intelligence has arrived or passed, and it can gradually replace some human work, such as writing. Recently, some thesis studios fired many teachers in this field, because they used to guide the writing of a thesis. It takes at least a few weeks or months, and some students of chatgpt have created an original article in less than half an hour in this way, and some script writing and other related positions have been affected as far as I know. Of course, be careful Students who know about it may also know about AI drawing, which has recently caused many artists to be replaced~ This is definitely a real person, I am not trying to create tension, but just want to say that with the popularization of AI in the future, a large part of our work will be Instead, we should learn how to learn more efficiently and how to use AI to improve efficiency at work. However, I still feel rather scared. One day AI will have consciousness, so if it is not controlled, it will not treat people like grass every minute~
Finally, I would like to ask everyone, will artificial intelligence replace us much in the future? What jobs can you do for a lifetime?
Related literature:
https://wenxin.baidu.com/
https://wenku.baidu.com/view/00f6ccdfe63a580216fc700abb68a98270feac1b.html?_wkts_=1679881096196&bdQuery=%E4%BB%80%E4%B9%88%E6%98%AF%E5%A4%A7%E8%AF%AD%E8%A8%80%E6%A8%A1%E5%9E%8B
https://www.163.com/news/article/I01RQEEF00019UD6.html
https://zhuanlan.zhihu.com/p/52061158
https://baijiahao.baidu.com/s?id=1651912822853865814&wfr=spider&for=pc
https://blog.csdn.net/fs1341825137/article/details/120247499
https://blog.csdn.net/anonymous_me/article/details/129812778
https://blog.csdn.net/WUTab/article/details/73657905
Related papers:
https://www.docin.com/p-2163732292.html
https://www.jianshu.com/p/8d6341ded7eb
https://arxiv.org/pdf/1409.0473.pdf
https://arxiv.org/pdf/1706.03762.pdf
https://arxiv.org/pdf/2203.15556.pdf
https://arxiv.org/pdf/1810.04805.pdf
https://arxiv.org/pdf/1706.03762.pdf
https://baijiahao.baidu.com/s?id=1760876469298017714&wfr=spider&for=pc
https://blog.csdn.net/u011414416/article/details/51057789