Leveraging LangGraph and AutoGen for Agentic AI Frameworks

IEEE Access

Artificial intelligence is changing the world, especially the interaction between machines and humans. Learning and interpreting natural languages and responding have paved the way for many technologies and applications. The amalgam of machine learning, deep learning, and natural language processing helped Conversational Artificial Intelligence (AI) to change the face of Human-Computer Interaction (HCI). A conversational agent is an excellent example of conversational AI, which imitates the natural language. This article presents a sweeping overview of conversational agents that includes different techniques such as pattern-based, machine learning, and deep learning used to implement conversational agents. It also discusses the panorama of different tasks in conversational agents. This study also focuses on how conversational agents can simulate human behavior by adding emotions, sentiments, and affect to the context. With the advancements in recent trends and the rise in deep learning models, the authors review the deep learning techniques and various publicly available datasets used in conversational agents. This article unearths the research gaps in conversational agents and gives insights into future directions. INDEX TERMS Artificial intelligence, machine learning, natural language processing, affective computing, mood or core affect, sentiment analysis, emotion theory, emotion in human-computer interaction, emotional corpora, intelligent agents, semantics, syntax, feature extraction, text processing. I. INTRODUCTION 16 Today everything we have in our society is the result of 17 intelligence; therefore, supplementing our human intellect 18 with artificial intelligence has the potential to help soci-19 ety thrive like never before-as long as we can make 20 the technology helpful. Healthcare, manufacturing, customer 21 services, e-commerce, education, media, from every facet, 22 it has transformed human life. One of the important branches 23 of artificial intelligence is conversational AI which makes 24 machines capable of understanding, processing, and respond-25 ing to humans in natural language. Conversational agents 26 The associate editor coordinating the review of this manuscript and approving it for publication was Utku Kose. have remained the center of the AI revolution in the past few 27 years, powered by Natural Language Processing (NLP) and 28 Machine Learning (ML) technologies. 29 A conversational agent [1] is an Artificial Intelligence (AI) 30 program that originated to imitate human conversations using 31 spoken or written natural language over the Internet. Many 32 alternative terms are used for conversational agents. Ear-33 lier, dialogue system, this term was popular. But nowadays, 34 chatbots, smart bots, intelligent agents, intelligent virtual 35 assistants/agents, interactive agents, digital assistants, and 36 relational agents are used alternatively in research articles 37 [1], [2]. Conversational agents are the practical implementa-38 tion of AI technology in industries or businesses. Conversa-39 tional agents can be seen being used in various applications 40

arXiv (Cornell University), 2023

The massive successes of large language models (LLMs) encourage the emerging exploration of LLM-augmented Autonomous Agents (LAAs). An LAA is able to generate actions with its core LLM and interact with environments, which facilitates the ability to resolve complex tasks by conditioning on past interactions such as observations and actions. Since the investigation of LAA is still very recent, limited explorations are available. Therefore, we provide a comprehensive comparison of LAA in terms of both agent architectures and LLM backbones. Additionally, we propose a new strategy to orchestrate multiple LAAs such that each labor LAA focuses on one type of action, i.e. BOLAA, where a controller manages the communication among multiple agents. We conduct simulations on both decision-making and multi-step reasoning environments, which comprehensively justify the capacity of LAAs. Our performance results provide quantitative suggestions for designing LAA architectures and the optimal choice of LLMs, as well as the compatibility of both. We release our implementation code of LAAs to the public at https://github.com/salesforce/BOLAA.

Lecture Notes in Computer Science, 1999

The leading edge of computer science research is notoriously fickle. New trends come and go with alarming and unfailing regularity. In such a rapidly changing field, the fact that research interest in a subject lasts more than a year is worthy of note. The fact that, after five years, interest not only remains, but actually continues to grow is highly unusual. As 1998 marked the fifth birthday of the International Workshop on Agent Theories, Architectures, and Languages (ATAL), it seemed appropriate for the organizers of the original workshop to comment on this remarkable growth, and reflect on how the field has developed and matured. The first ATAL workshop was co-located with the Eleventh European Conference on Artificial Intelligence (ECAI-94), which was held in Amsterdam. The fact that we chose an AI conference to co-locate with is telling: at that time, we expected most researchers with an interest in agents to come from the AI community. The workshop, which was planned over the summer of 1993, attracted 32 submissions, and was attended by 55 people. ATAL was the largest workshop at ECAI-94, and the clear enthusiasm on behalf of the community made the decision to hold another ATAL workshop simple. The ATAL-94 proceedings were formally published in January 1995 under the title Intelligent Agents, and included an extensive review article, a glossary, a list of key agent systems, and-unusually for the proceedings of an academic workshop-a full subject index. The high scientific and production values embodied by the ATAL-94 proceedings appear to have been recognized by the community, and resulted in ATAL proceedings being the most successful sequence of books published in Springer-Verlag's Lecture Notes in Artificial Intelligence series. ATAL-95 was held at the International Joint Conference on AI, which in 1995 was held in Montreal, Canada. The number of submissions leapt to 55, and the workshop was attended by 70 people. Since many international conferences fail to attract this many submissions and delegates, it was decided at this point to make ATAL an annual event. It was also decided to hold ATAL-96 in Europe, following the successful model of ATAL-94 by co-locating with ECAI-96, which was held in Budapest, Hungary. We received 56 submissions, and the workshop was attended by about 60 delegates. For ATAL-97, it was felt that the workshop was sufficiently mature that it merited its own event, and so the conference was located immediately before the AAAI-97 conference in Providence, Rhode Island. It was attended by about 75 delegates. ATAL-98 was co-located with the "Agents World" series of events, held in Paris in July 1998. 90 submissions were received, and 139 delegates registered for ATAL. In the five years since ATAL-94, the landscape of the computing world has changed almost beyond recognition. Even seasoned veterans of the historically fast-moving IT environment have been surprised by the current rate of change. Perhaps the simplest way we can sum up these changes is by noting that the first ATAL was also the last not to have a World Wide Web (WWW) page. In 1999, on the eve of the new millennium, it would be unthinkable for a serious academic conference or workshop not to have a dedicated WWW site. The changes brought about by the explosion of the Internet into worldwide public and corporate awareness are well documented, and it is not appropriate for us to add to the mountain of comment (and hyperbole). However, it is important to note that the rise of the Internet had a significant impact on the development of the agent field VI Preface itself. By the summer of 1994 it was becoming clear that the Internet would be a major proving ground for agent technology (perhaps even the "killer application"), although the full extent of this interest was not yet apparent. The emergence of agents on and for the Internet gave rise to a new, associated software technology, somewhat distinct from the "mainstream" of agent research and development. In the summer of 1994, a California-based company called General Magic was creating intense interest in the idea of mobile agents-programs that could transmit themselves across an electronic network and recommence execution at a remote site. At the time, General Magic were distributing a widely-read white paper that described "Telescript"-a programming language intended to realize the vision of mobile agents. In the event, it was not Telescript, but another programming language that caught the imagination of the Internet community: Java. When Netscape incorporated a Java virtual machine into their Navigator browser, and hence brought the idea of applets into the mainstream, they gave Java an enormous impetus, both as a way of animating the Internet, but also as a powerful, well-designed object-oriented programming language in its own right. A number of mobile agent frameworks were rapidly developed and released as Java packages, and interest in Telescript rapidly waned. As we write this preface in late 1998, Java is the programming language of choice not just for agent systems, but also, it seems, for most other applications in computing. Mobile agent technology was not the only other agent technology beginning to make its presence felt at the time of the first ATAL. The summer of 1994 saw the publication of a special issue of Communications of the ACM with the title "intelligent agents". Many of the articles in this special issue described a new type of agent system, that acted as a kind of "expert assistant" to a user working with a particular class of application. The vision of agents as intelligent assistants was perhaps articulated most clearly by Pattie Maes from MIT Media Lab, who described a number of prototype systems to realize the vision. Such user interface agents rapidly caught the imagination of a wider community, and in particular, the commercial possibilities of such technologies was self-evident. A number of agent startup companies were founded to commercialize this technology (many of which have by now either been sold or gone bust). Current interest in such agents comes, to a great extent, from the possibility of using them in electronic commerce scenarios, where they negotiate on behalf of their "owner". The commercial interest in agents in the latter half of the 1990s has not been limited to venture capitalists and "small scale" agent systems. Perhaps one of the most encouraging long-term trends for agent technology is the idea of agents as a software engineering paradigm. The level of interest in this concept has been evidenced in several ways. For example, the number of large-scale industrial-strength agent systems being developed and deployed is an obvious indicator. However, the degree of interest is perhaps best illustrated by the attempts currently underway to develop international standards for agent communication. Although some tentative steps towards standard agent communication languages were taken by the KQML/KIF community in the USA in the early 1990s, it is the FIPA initiative, started in 1995, which currently appears to be the best candidate for a standard agent communication framework. Turning more specifically to the ATAL workshops, a number of smaller scale trends have emerged, echoing to some extent the more visible changes in the computing world Preface VII itself. One obvious indicator that agent technology is beginning to mature is that far fewer new agent architectures are being developed. It seems that authors are taking architectures off the shelf, rather than developing their own. In this vein, the belief-desire-intention (BDI) class of architectures has become particularly prominent. This work represents a paradigm example of theATAL ethos-there is a well-defined theory, which relates more or less directly to specific architectures or programming languages. On the theoretical side, there has been an increasing trend towards more integrated models; that is, theories which cover a wider proportion of an agent's decision making and acting capabilities. We noted above that five years sometimes seems like a long time for an academic workshop. Incredibly, when ATAL began, there were no conferences dedicated to agent technology. In contrast, the agent research community is now served by at least two major international scientific conferences (the International Conference on Multi-Agent Systems and the International Conference on Autonomous Agents), as well as a dedicated journal (Autonomous Agents and Multi-Agent Systems). That agent technology is able to comfortably support this degree of interest tells us that agents have a good chance of succeeding as a technology. We hope that ATAL will continue to play its part in this development, maintaining its current high level of scientific and production values, and serving a vibrant, rich research and development community. To close, we would like to take this opportunity to thank those who have made ATAL the success we sincerely believe it is today. In particular, our thanks go to Jörg Müller, Munindar Singh, Anand Rao, and Milind Tambe, who have all acted as organizers for ATAL, and helped to shape it through their dedication and vision. In addition, we would like to thank those who have played various other special roles throughout the first five years, including

IntechOpen eBooks, 2023

Business communication today is driven by the organizational financial performance and customer best quality service provision. The rise of 24/7 ubiquitous access to Internet and Applied Artificial Intelligence (AAI) creates a platform for future fully-automated cyberspace. Role of Computing has become an essential part of research, innovation and development of future cutting-edge technologies that will transform the way we live to a next level, where Machine-to-Machine Communication will drive Human-to-Human Communication. Today we see the technology that has brought the Computational Machines much closer to us people, and LEXA is becoming an integrated part home companionship and the entertainment for many young families who are tech-savvy and love to play with the high-tech. Naturally, there are many dimensions to technological innovation and application of AI that are fascinating and inspire scholars and industry developers to push the edge of advancement of next generation technologies that will promote a Cyber Automation and Ultra Smart Living for everyone. The Machine learning (ML) as a subdomain of artificial intelligence enables smart computational device to abstract patterns from data without explicit programming. In addition to AI and ML, we see the rise of Humanoid Robotics that has potential to act like, look like, talk like and reason like human, and eventually becoming an essential part of an organizational business infrastructure worldwide [1]. Some experts compare level of intelligence of new Humanoid Robot called Sophia to that of a one month old new born baby that is at the very beginning of discovering the world around and is beginning to observe and to learn to communicate with his or her parents. Yet, very young and tiny, babies have a Natural Intelligence and Human Reasoning that may be compared to a computational capacity of most performant Supercomputer in the world today. Many experts suggest that there are fundamental differences between the AI and Natural Intelligence (NI) and that future evolution of AI driven Software Agents and Humanoid Robotics may have strong impact on the future Industry 5.0 and 6.0, Business to Business (B2B) and Mobile Commerce (mC). Given recent COVID relates crisis, applications of AI in business, industry, government, academia and other sectors have become ubiquitous showing more and more examples of Smart Software Agents application. The next section, presents examples of the AI driven applications including chatbots and ChatGPT, and discus what, how and where they are.

2024

Can emergent language models faithfully model the intelligence of decision-making agents? Though modern language models exhibit already some reasoning ability, and theoretically can potentially express any probable distribution over tokens, it remains underexplored how the world knowledge these pretrained models have memorized can be utilized to comprehend an agent's behaviour in the physical world. This study empirically examines, for the first time, how well large language models (LLMs) can build a mental model of agents, termed agent mental modelling, by reasoning about an agent's behaviour and its effect on states from agent interaction history. This research may unveil the potential of leveraging LLMs for elucidating RL agent behaviour, addressing a key challenge in eXplainable reinforcement learning (XRL). To this end, we propose specific evaluation metrics and test them on selected RL task datasets of varying complexity, reporting findings on agent mental model establishment. Our results disclose that LLMs are not yet capable of fully mental modelling agents through inference alone without further innovations. This work thus provides new insights into the capabilities and limitations of modern LLMs.