YEG Hack-GPT aims to unlock the power of large language models

· The Pulse

Developers and entrepreneurs will see what problems they can solve with the kind of technology that powers ChatGPT at a hackathon thought to be the first of its kind in Edmonton.

YEG Hack-GPT will see teams gather at Work Nicer Beaver House from July 22 to 23 to build something using large language models such as GPT-3 and 4, LLaMA, or Falcon.

"You're ultimately building a little startup within two days," said Roberto Moreno, vice-president of artificial intelligence for PLACE, the real-estate technology and services company backing the hackathon.

More than 30 participants have registered so far for the event, which is geared toward AI enthusiasts, data scientists, recent graduates, and similarly tech-minded people. It's sponsored by AI companies such as Weaviate, OpenAI, LangChain, and Hugging Face, making it possible to award $3,000 in prizes for the best app and the most unorthodox use of large language models. While participants can work on solutions to any problem they choose, Moreno said those who work on something related to real estate may receive additional recognition.

"The beauty of hackathons is that the technologies allow you to build a lot," he told Taproot. "You may think it's not a lot of time, but it really makes people focus on what the problem is and maybe just solving that problem instead of all these other things."

Moreno said it's Edmonton's first hackathon focused on large language models. He organized it to help nurture talent by giving people a chance to network, he said.

"There's a lot of people looking for jobs in this space, especially in Edmonton," he said. "It's going to allow them to build something that they can show on their resume and network with other people and kind of build a community around that."

An AI-generated image of six people sitting at a table with laptops open. There are three white men, two Black men, and a woman whose race is difficult to discern.

YEG Hack-GPT organizer Roberto Moreno generated this image using Midjourney with the prompt "Group of hackers of multiple ethnicity sitting around a table with computers at a hackathon." Large language models tend to reflect biases in the data they're trained on, which may account for the gender imbalance in the picture. (Supplied)

Large language models use deep learning algorithms trained on huge amounts of data to perform natural language processing. They can answer questions, summarize text, and analyze sentiment. Chatbots like ChatGPT use a large language model to understand a prompt and generate text in response; programs like Midjourney use large language and diffusion models to generate images.

Large language models can help real estate professionals with tasks such as customer service, writing marketing posts, and scheduling, Moreno said.

"A real estate team or agents have a lot of repetitive things they do, and some of these AI technologies can help alleviate that repetitiveness and help these agents do what they're good at," he said. "Like showing properties, meeting customers, writing offers, that kind of thing."

While these are early days for generative AI and there are many ethical issues to grapple with, Moreno said it is ultimately something people can leverage for their own benefit.

"I think people need to embrace it. It is a tool, it's not going to solve everything, but it's going to allow you to work on things that maybe you want to work on," he said. "Yeah it's scary, but it's going to remove a lot of those repetitive things, and kind of mundane things."