Exploring the Diverse Applications of Large Language Models
The second application area is search.
Talking to a Machine You May Not Notice
Many people now meet large language models without knowing it. They type a question into a chat window, or ask a spoken assistant on a phone, and receive a written or spoken reply. Behind some of these experiences is an LLM, which stands for Large Language Model. This article focuses on one specific application area: using LLMs to power chatbots and virtual assistants, and to improve traditional search engines.
The central idea is simple. An LLM can power a chatbot, and a chatbot powered this way can answer user queries. A query is whatever the user asks or types. The LLM is the component that produces the answer. The source does not describe the internal design of products such as ChatGPT or Gemini, so the most useful model at this level is the flow of information between the user, the LLM, and the reply.
Following One Query Through a Chatbot
Read the diagram below from left to right. The user's query enters the system, the LLM handles it, and an answer comes back out. Follow the same three stages each time: the query comes in, the LLM answers, and the answer goes back to the user. The diagram shows only the parts the source supports. It does not show extra stages such as storage, retrieval, or filtering, because the source does not describe them.
What do you think happens?
In the chatbot flow above, which step actually processes the user's query?
Reveal answer
The LLM
The source says that in an LLM-powered chatbot, a user query goes to the LLM, which can answer it. The LLM is the component that produces the answer; the other steps move the query in and the reply out.
Where the LLM Sits Among Chatbots and Assistants
A virtual assistant is a computer program that assists users, and LLMs can power it. A chatbot is an application that answers user queries, and LLMs can power it too. The source treats these two kinds of application together because both are named as applications that LLMs can power and both answer user queries.
A virtual assistant is the broader category: a program that helps the user. An LLM is one technology that can supply the assistant's ability to understand and answer requests. ChatGPT and Gemini are the source's examples of this kind of product. The diagram below shows the relationships the source provides and nothing beyond them.
Augmenting Traditional Search Engines
The second application area is search. The source says LLMs augment traditional search engines like Google Search and Microsoft Bing. The important word is augment. The source does not say that LLMs replace search engines. It says LLMs can improve traditional search engines. The relationship is one of enhancement: the search engine remains a traditional search engine, and the LLM contributes to making it better.
Tracing a Student's Question Step by Step
A student asks an LLM-powered chatbot a question
A student opens a chatbot powered by an LLM and types a question. Identify what happens to the question and where the LLM fits.
- 1The query comes in
The student types a question. The question is the query, and it enters the chatbot.
- 2The LLM answers
The chatbot passes the query to the LLM, which produces a reply. This is the only step that produces the answer.
- 3The answer goes back
The reply is returned to the student as a written or spoken response.
- 4The student reads the reply
The flow ends with the student reading the answer. The same three stages repeat for each new question.
The LLM sits in the middle stage, between the incoming query and the outgoing reply. The chatbot carries the query in and the answer back out, and the LLM is the component that answers.
Common Misreadings About LLM Applications
Common mistake Treating an LLM-powered chatbot as a search engine
What it looks likeAssuming a chatbot that answers a question is the same product as a search results page.
Why it failsThe source describes the chatbot case as one where a query goes to the LLM, which answers it. It describes search as a separate application area where the LLM improves an existing search engine.
Do this insteadKeep the two applications distinct: a chatbot answers queries directly, while search is improved by the LLM.
Common mistake Believing the LLM replaces search engines
What it looks likeReading 'LLMs augment traditional search engines' as 'LLMs take over search'.
Why it failsThe source uses the word augment and says LLMs improve rather than replace traditional search engines.
Do this insteadDescribe the relationship as enhancement of an existing tool.
Common mistake Adding internal components the source does not describe
What it looks likeStating that ChatGPT or Gemini contain a specific retrieval database or filter.
Why it failsThe source explicitly says it does not describe the internal design of these products.
Do this insteadDescribe only the flow the source supports: query in, LLM answers, reply out.
Common mistake Assuming the source claims guaranteed correctness
What it looks likeConcluding that because an LLM can answer a query, its answer is always right.
Why it failsThe source states that LLMs can power chatbots and improve search. It does not claim that every reply is guaranteed to be accurate.
Do this insteadDescribe what the source says LLM applications can do, and do not attribute guarantees to them that the source does not make.
Trace a Different Query
A user asks a virtual assistant on their phone to answer a question by voice, and hears a spoken reply. Trace the query through the flow: name the stage where the query enters, the stage where the LLM answers, and the stage where the answer returns. Then state whether this scenario is a chatbot case or a search-augmentation case, and justify your choice using the source.
Hints
- Look at the three stages: the query comes in, the LLM answers, the answer goes back.
- The source names virtual assistants among the applications an LLM can power.
- Search augmentation is specifically about improving a traditional search engine such as Google Search or Microsoft Bing.
Key Takeaways From the Applications
- 1An LLM can power a chatbot, and a chatbot powered this way answers user queries: the query goes to the LLM, which produces the reply.
- 2A virtual assistant is a computer program that assists users, and LLMs can power it. ChatGPT and Gemini are the source's examples of this kind of product.
- 3LLMs augment traditional search engines such as Google Search and Microsoft Bing, improving them rather than replacing them.
- 4The source does not describe the internal design of these products or claim guaranteed accuracy, so describe only the flow and the relationships the source supports.
At a glance
- An LLM can power a chatbot, and a chatbot powered this way answers user queries by sending the query to the LLM and returning its reply.
- Virtual assistants are programs that assist users, and LLMs can power them; ChatGPT and Gemini are the source's examples.
- LLMs augment traditional search engines such as Google Search and Microsoft Bing, improving rather than replacing them.
- Avoid claims the source does not make, including details of internal design and guaranteed accuracy.