Artificial Intelligence (AI) is a very big umbrella term for a wide range of programs, and the popular definition has only gotten fuzzier as tech companies jump on the hype and make bold claims about the abilities of their products. Encyclopedia Britannica's definition is "the ability of a digital computer or computer-controlled robot to perform tasks commonly associated with intelligent beings ... [especially] systems endowed with the intellectual processes characteristic of humans, such as the ability to reason, discover meaning, generalize, or learn from past experience." While we are a far cry away from computers that can think like people, there are already very specialized programs that can out-preform human experts in certain tasks, and billions of dollars are being spent on expanding and refining the technology. This Libguide is focused on the basics of how these machines work, the strengths and weaknesses of the most popular programs, and how they can affect our campus.
ChatGPT and Large Language Models
Large Language Models (LLMs) are some of the most significant AIs to our campus as they include popular text-generating AI like ChatGPT and Google's Bard. They are generative AIs, meaning that they seemingly create new works like stories, letters, realistic conversations, and more when prompted. While this can give the illusion of a computer with a human's creativity and expertise in any subject, these systems actually work like the world's most advanced set of magnetic poetry. LLMs work by being fed and trained on billions of pages of text, including huge swaths of the freely accessible internet, public domain books, etc, and then break them down into strings of symbols called tokens. Once its broken down, a LLM studies what patterns the tokens follow: It might learn that the letters that spell out "Hello, how are you today?" are commonly put together in that order, so when you ask it to write a greeting, it follows that common pattern. By knowing these patterns, programs like ChatGPT can create responses that read like clear English without actually having to understand any of the words; it just has to move letters around to look like what it's read before.
Just How Powerful is AI?
There are lots of programs being labelled artificial intelligence, but there is a wide gap between sci-fi dreams and the tech being developed today. AI's can be put into two categories: artificial narrow intelligence and artificial general intelligence, or "weak AI" and "strong AI"
Strong AI or artificial generalized intelligence would be a computer program that can take in information, process and learn from that information, and act accordingly without human involvement. It would be a "generalized" intelligence because it could take in lots of different ideas and observations and develop new skills and creative ideas based on them. This is the AI you see in science fiction and sometimes in bold claims by current engineers. However, no AI today has been proven to be anywhere close to this level of autonomy or complexity.
Weak AI is often called "narrow" intelligence because it is about creating a program that can respond to a focused set of inputs and be specialized at producing related outputs. Whenever you hear about AI beating a human grandmaster at chess, that is an example of a program that is extremely skilled at a narrow scope of actions. However this program is built entirely around the task it was designed to excel in: a chessbot only knows the world of chess, is entirely designed for chess, and cannot be given or even tweaked enough to take on new task like driving a car.
ChatGPT and other LLMs are extremely complex and able to create a wide scope of different text outputs, from poetry to chatbots to short stories, but are still very limited in the types of information they can process and then act on, making them a very powerful form of weak AI.
What're the Issues?
Despite their popularity, Large Language Models are still a new technology and suffer from several pitfalls. Here are a few important limitations and warnings to keep in mind before using any form of AI in your college career.
How Can It be Helpful?
The best uses of AI and LLMs as a student are as productivity tools, not as writing replacements. Here are some ways a program help inspire and streamline your own work.
2300 HIGHLAND ROAD • BATESVILLE, ARKANSAS 72501 • PHONE: (870-307-7205) • FAX: (870) 307-7279