Surviving Science Fiction: Chat GPT and Large Language Models
How are large language models changing our world, and why are so many people worried about them?
In November 2022, it felt like we started living the script of a science fiction movie when the tool known as ChatGPT 4 was released to the world. For technologists, this was the evolution of a continuing phenomenon; previous versions of ChatGPT had come and gone. But none of them were capable of capturing the eye of the public like this one. Something had fundamentally shifted.
At first, things were heady and fun. People asked for biographies, poetry. But concerns surfaced quickly. And by the time products like Bing Chat surfaced that used ChatGPT to power their experiences, things got REALLY weird. Suddenly, billionaires were warning the world to slow down. What? Aren’t they the people with the power to slow things down? Weren’t they the people who invested in these things in the first place?
This is the second post in my “Surviving Science Fiction” series where I attempt to provide a more human context for the AI concepts driving so much of today’s news. For this second entry, let’s dive in and demystify the Large Language Model concept at scale:
- What are Large Language Models (LLMs)? (review from Part 1)
- How are LLMs built, and why is the creation of these models so controversial?
- How is using an LLM different from just talking to a human or traditional chatbot?
- Where are LLMs being used today?
- What are the legal risks around using LLM content?
- When do LLMs become dangerous?
What are Large Language Models (LLMs)?
The market research firm Gartner defines a large language model as:
“a specialized type of artificial intelligence (AI) that has been trained on vast amounts of text to understand existing content and generate original content.”
A large language model is (to be very reductive) the code equivalent of creating the Genie of the Lamp from the story of Aladdin: Huge language power, itty bitty living space, and typically limited by certain parameters about the language “magic” created.