以ChatGPT为代表的大型语言模型研究进展

大型语言模型是当今人工智能领域最前沿的研究方向之一,该方向旨在训练含有大规模参数的通用语言模型,使其能够遵循人类指令完成不同类型的自然语言处理任务。作为大型语言模型的代表,由OpenAI研发的ChatGPT在各个领域均展现出强大的自然语言生成能力,受到了全球各行各业的关注。本文从语言模型的发展历程出发,介绍了近年研究者在扩大语言模型规模上的探索,然后分析了大型语言模型带来的范式改变,并以ChatGPT为典型实例概述了其发展、技术和应用,接着介绍了后ChatGPT时代大型语言模型的前沿进展,最后从评价和治理两方面总结了目前大型语言模型的局限性及未来需要解决的挑战。

Artificial Intelligence Methods Applied to Catalytic Cracking Processes

Fluidic Catalytic Cracking(FCC)is a complex petrochemical process affected by many highly non-linear and interrelated factors.Product yield analysis,flue gas desulfurization prediction,and abnormal condition warning are several key research directions in FCC.This paper will sort out the relevant research results of the existing Artificial Intelligence(AI)algorithms applied to the analysis and optimization of catalytic cracking processes,with a view to providing help for the follow-up research.Compared with the traditional mathematical mechanism method,the AI method can effectively solve the difficulties in FCC process modeling,such as high-dimensional,nonlinear,strong correlation,and large delay.AI methods applied in product yield analysis build models based on massive data.By fitting the functional relationship between operating variables and products,the excessive simplification of mechanism model can be avoided,resulting in high model accuracy.AI methods applied in flue gas desulfurization can be usually divided into two stages:modeling and optimization.In the modeling stage,data-driven methods are often used to build the system model or rule base;In the optimization stage,heuristic search or reinforcement learning methods can be applied to find the optimal operating parameters based on the constructed model or rule base.AI methods,including data-driven and knowledge-driven algorithms,are widely used in the abnormal condition warning.Knowledge-driven methods have advantages in interpretability and generalization,but disadvantages in construction difficulty and prediction recall.While the data-driven methods are just the opposite.Thus,some studies combine these two methods to obtain better results.

Ultra-Short Wave Communication Squelch Algorithm Based on Deep Neural Network

The squelch problem of ultra-short wave communication under non-stationary noise and low Signal-to-Noise Ratio(SNR)in a complex electromagnetic environment is still challenging.To alleviate the problem,we proposed a squelch algorithm for ultra-short wave communication based on a deep neural network and the traditional energy decision method.The proposed algorithm first predicts the speech existence probability using a three-layer Gated Recurrent Unit(GRU)with the speech banding spectrum as the feature.Then it gets the final squelch result by combining the strength of the signal energy and the speech existence probability.Multiple simulations and experiments are done to verify the robustness and effectiveness of the proposed algorithm.We simulate the algorithm in three situations:the typical Amplitude Modulation(AM)and Frequency Modulation(FM)in the ultra-short wave communication under different SNR environments,the non-stationary burst-like noise environments,and the real received signal of the ultra-short wave radio.The experimental results show that the proposed algorithm performs better than the traditional squelch methods in all the simulations and experiments.In particular,the false alarm rate of the proposed squelch algorithm for non-stationary burst-like noise is significantly lower than that of traditional squelch methods.