Using AI Chatbots: A Gentle Introduction
Using LLMs for knowledge access, content generation, and analytical reasoning
I still use a corded ball mouse from the 90s at my home desktop computer. And that's the only computer I ever use; never owned a laptop. As luddite my lifestyle is, I work in tech—as CEO of an AI startup no less—and I do use AI in my personal life. This is because I like books and am curious about things. And the best AI chatbots are like having at hand a full-blown professor with 24/7 office hours.
For folks like me, who are otherwise late to AI, I organize here a simple three-part division of the kinds of useful interactions one can have with an AI chatbot. In increasing sophistication, we have: knowledge access, content generation, and analytical reasoning. I then discuss briefly where the next kind of AI chatbot usecase appears imminent: executive assistance.
Knowledge Access
Everyone uses Google. Even if you don't search on google.com, queries you run on e.g. some news site are often powered by Google. When you use a search engine here, you are asking it to find relevant content. That is its purpose. But often you are not looking for content. You are looking for an answer. Consider the following question:
Q: In the United States, do I have a right to clean air?
Googling that question gives a page of results all relating to the Clean Air Act. One would have to go read some of the webpages cited to check if the Clean Air Act gives one the right to clean air. Asking ChatGPT the same question gives a two-sentence summary of the Clean Air Act, and very interestingly also another three factors that impinge on the right; the answer reads like a short brief:
A: In the United States, there is no explicit constitutional right to clean air, but there are laws and legal precedents that protect air quality:
Clean Air Act (1970) – This is the primary federal law regulating air pollution. It authorizes the Environmental Protection Agency (EPA) to set air quality standards and enforce regulations to limit pollutants.
Public Trust Doctrine – Some legal scholars argue that the government has a duty to protect natural resources, including air, for public use. However, this is not always recognized in courts as an enforceable individual right.
State Constitutions – Some states, like Pennsylvania and Montana, explicitly recognize a right to a clean environment in their constitutions.
Environmental Lawsuits – Courts have occasionally recognized citizens' rights to sue for environmental harms, though these cases depend on specific legal arguments rather than a broad "right to clean air."
With Google, you get search results. With ChatGPT, you are getting an answer; an answer that is both more succinct and more complete. The AI typically explains how it arrived at its answer. In fact, I commonly ask follow-up questions on concepts I am new to.
Content Generation
As yet, I do not know what title to give this post. I am planning on having AI read this and suggest some titles. I'm not likely to use one of the suggestions, but I might borrow a word or turn of phrase from one of them. This is the "generative" mode of AI use, and includes asking AI to generate an image depicting some scene of interest. An AI chatbot can even draft an entire blog post, or email, or computer program; that a user can then edit to their desired result. Though my style is to write my own draft first.
AI chatbots often offer useful suggestions even when not asked explicitly to do so. I asked ChatGPT, following up on my earlier question:
Q: My issue is that the government is building an air polluting bus depot close to my residence. Does the 9th Amendment help in suing them to stop? Please explain.
After explaining that the 9th amendment does not help, ChatGPT noted other avenues worth exploring; including a non-legal one. Again, this is "generative"; perhaps in the space of ideas:
A: …
Instead of relying on the 9th Amendment, you may have better success using:
Environmental Laws
The Clean Air Act (CAA) requires federal and state agencies to ensure that projects do not violate air quality standards. If the depot will significantly increase pollution, you could challenge it based on noncompliance.
Environmental Impact Assessments: If the depot is federally funded, it may need an environmental review under the National Environmental Policy Act (NEPA). You could challenge the review if it underestimates the pollution risk.
State & Local Laws
Many states have their own environmental protection laws that may provide grounds for a lawsuit.
Zoning and land use laws could be relevant if the depot violates local regulations.
Public Nuisance Laws
If the depot will substantially interfere with your ability to use and enjoy your property due to pollution, you may be able to file a public nuisance lawsuit.
…
Engage the Community: Public opposition and media attention can pressure decision-makers.
Analytical Reasoning
The earliest computers were typically used for doing arithmetic. When I attended university, I found computer labs there that were running much more complex computations, e.g. to solve equations modeling airplane flight. But we never thought of these machines as performing reasoning. They were simply executing instructions that us humans had coded and provided them. This has changed with the advent of modern AI.
Modern LLMs are able to solve mathematical problems they have not been explicitly coded to do so. This breakthrough occurred in recent months with the advent of the reasoning versions of LLMs12.
If you are a lawyer, you might think that this does not apply in your work. But this same ability to reason through a mathematical problem, allows an LLM to reason through a legal argument. When writing an appellate brief, an LLM in reasoning mode can comb through the prior trial proceedings to look for defects. At Colit AI, we are hard at work bringing this functionality to our users.
And Beyond...
Solving math problems, and reasoning in general, are at first sight off-label uses of AI chatbots. Unless...we focus on the AI part of AI chatbots. The Artifical Intelligence behind chatbots has facility with language, but also other modes of reasoning. What we have here is intelligence on tap, a phrase Dario Amodei, CEO of Anthropic used in his recent essay on the topic3.
Unlike humans', to be ‘tapped’, this artificial intelligence needs detailed directions towards specific goals. But this is changing rapidly. We have just witnessed the release of ‘Deep Research’ versions of LLM-based AI4. These carry out tasks that require multiple steps of searching the Internet and assembling disparate results.
Soon there may be AI executive assistants that will do even more: they will take actions. They will carry out tasks such as "I want to repeat last year's Thanksgiving trip to my parents' this year; please make the same arrangements where possible." Said AI EA will look up your email confirmations from last year and book the same flight using your credit card, choosing your favorite airplane seat. Such AI EAs could even show up at your office, to handle simpler projects for you. They might communicate with you on Slack, or email, to report progress and ask for further guidance.
To the extent such AI assistants do a good job, AGI5 beckons ever closer.
OpenAI: https://openai.com/index/learning-to-reason-with-llms/
Google: https://techcrunch.com/2024/12/19/google-releases-its-own-reasoning-ai-model/
DeepSeek: https://github.com/deepseek-ai/DeepSeek-R1
xAI: https://x.ai/blog/grok-3
OpenAI Unveils New A.I. That Can ‘Reason’ Through Math and Science Problems.
https://www.nytimes.com/2024/12/20/technology/openai-new-ai-math-science.html
Machines of Loving Grace: How AI Could Transform the World for the Better. Dario Amodei.
https://darioamodei.com/machines-of-loving-grace
OpenAI: https://openai.com/index/introducing-deep-research/
Google: https://blog.google/products/gemini/google-gemini-deep-research/
Perplexity: https://www.perplexity.ai/hub/blog/introducing-perplexity-deep-research
Artifical General Intelligence. When we think machine intelligence has approached humans’, generally speaking, without material qualifications.