CALL MY AGENT
I
magine this: You are out in the field during a training exer- cise, wrestling with a stubborn equipment issue that has resisted every solution you’ve tried. Frustrated, you call the help desk, hoping they can guide you to the answer.
Tey answer and, with friendly camaraderie, begin probing your situation before providing a step-by-step walkthrough for your exact issue. Ten, much to your surprise, you are asked to rate your interaction with an artificial intelligence (AI) chatbot. Until now, you were under the impression that you were speaking to a human. After all, they sounded and responded exactly like a real person while providing the optimal solution.
AI advancements are progressing rapidly, making this scenario increasingly realistic. What appears to be a single AI entity is actually a team of AI entities working in tandem. Tese entities, known as AI agents, each specialize in one task and cooperate to provide the best response. One agent gathers information from your speech, another quickly queries thousands of documents to find and rank the best solutions and a third agent uses a power- ful language model to generate understandable, accurate and context-rich responses. Like people, AI agents work well in teams, producing greater value than the sum of their parts.
LARGE LANGUAGE MODELS When you call a company’s support hotline, you are often greeted by an automated voice system. While cumbersome, they do a decent job of gathering information from callers and directing them to the appropriate department. Tese systems may soon see significant upgrades with AI agents powered by large language models (LLMs). Instead of the caller needing to “listen carefully because the menu options have changed” or preemptively press- ing “5” for options in Spanish, callers may converse with an AI agent that can respond automatically in a preferred language.
LLMs are also able to sift through huge volumes of text data from sources such as articles, forums and social media platforms to understand and generate human-like language. Tese models can be fine-tuned for specific tasks. For example, a model trained on software code can learn best coding practices from billions of lines of code. Similarly, a model trained on help desk support can mimic field support representatives or data system engineers and generate relevant remedies to user issues by analyzing a specific collection of documents, such as user manuals and technical documentation. Tis is especially useful in situations where the availability of human experts is limited.
FIRST THINGS FIRST
Before documents can be read by large language models, they must have their text content extracted and then converted into a series of numerical decimal values, called embeddings. Larger decimal values create more specific embeddings, which can increase the accuracy retrieval. These embeddings are stored in a database specifically optimized for processing these values. (Graphic by Turhan Kimbrough, PEO C3T and USAASC)
38
Army AL&T Magazine
Fall 2024
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60 |
Page 61 |
Page 62 |
Page 63 |
Page 64 |
Page 65 |
Page 66 |
Page 67 |
Page 68 |
Page 69 |
Page 70 |
Page 71 |
Page 72 |
Page 73 |
Page 74 |
Page 75 |
Page 76 |
Page 77 |
Page 78 |
Page 79 |
Page 80 |
Page 81 |
Page 82 |
Page 83 |
Page 84 |
Page 85 |
Page 86 |
Page 87 |
Page 88 |
Page 89 |
Page 90 |
Page 91 |
Page 92 |
Page 93 |
Page 94 |
Page 95 |
Page 96 |
Page 97 |
Page 98 |
Page 99 |
Page 100 |
Page 101 |
Page 102 |
Page 103 |
Page 104 |
Page 105 |
Page 106 |
Page 107 |
Page 108 |
Page 109 |
Page 110 |
Page 111 |
Page 112 |
Page 113 |
Page 114 |
Page 115 |
Page 116 |
Page 117 |
Page 118 |
Page 119 |
Page 120
