传统的机器学习（ML）还相关吗？目前在机器学习方法方面是否有活跃的研究？
今天我看到的一切都与大型语言模型（LLMs）有关。我个人在日常活动中使用LLMs，它们表现得非常出色。但我不禁想问，传统的机器学习方法怎么突然就消失了！那些垃圾邮件分类器、情感分析模型、word2vec、递归神经网络（RNN）、卷积神经网络（CNN）、长短期记忆网络（LSTM）、前馈神经网络（FFNN）现在都在哪里？2010年代末的典型数据科学家或机器学习工程师现在在做什么？他们训练的决策树、设计的神经网络、进行的准确性评估、超参数调优，现在都去哪儿了？感觉突然间一切都与LLMs有关。那些为特定问题提供轻量且易于部署的传统机器学习模型现在都过时了吗？
查看原文
Everything I see today is only about LLMs. I personally use LLMs in my daily activities and they are doing great job. But I wonder what happened to traditional machine learning methods all of a sudden! Those hamspam classifiers, sentiment analysis models, word2vecs, RNNs, CNNs, LSTMs, FFNNs, where are they now? What is a typical data scientist or a ML engineer of late 2010s doing now? The decision trees they trained, the neural nets they architected, the accuracy evaluations, hyperparameter tunings where are they all now? It feels like it is all about LLMs suddenly. The traditional ML models which served specific problem with light weight and easily deployable loads are all obsolate now?