Open access to AI models and their training materials is necessary to make the technology widely available, NVIDIA engineer Chris Alexiuk said in a company blog published Nov. 5.
His argument extends beyond access to finished models to the data and methods used to build them. NVIDIA, which develops the Nemotron models he promotes, published his comments in “The Wizard of Open Weights: Chris Alexiuk Makes AI Easy With Open Models.”
“Fully open source AI is necessary for the democratization of intelligence,” said Alexiuk, a Toronto-based senior product research engineer in deep learning at NVIDIA.
He said developers should have access to model weights, the learned values that determine how an AI model responds, alongside training data and instructions for training. His case for openness draws on years spent teaching developers to work with open models.
Alexiuk teaches on YouTube, showing developers how to build, adapt, test and put applications into use with large language models, the technology behind many AI chatbots. His lessons also cover AI agents that perform tasks.
He said spreading that knowledge could bring more people into AI development, including people who have yet to encounter the technology. He described reaching as many people as possible as the motivation for his teaching.
Alexiuk said open models could serve developers with different needs, from individuals handling repetitive tasks to companies building autonomous agents for extended assignments. He cited Nemotron 3.5 Lightning for frequent execution tasks and Nemotron Ultra for research and enterprise work.
He said openness helps more people understand AI and expressed hope that wider access would encourage inventive uses. He also described AI tools as a way to amplify users’ skills and human expression.
His enthusiasm for NVIDIA’s models is explicit. Alexiuk calls himself “No. 1 Nemotron Fan” on LinkedIn and uses the name Joe Nemotron on social media. He said his support reflects his belief in openness and his view that Nemotron best advances it.
His interest in machine learning began in a computational physics class at Trent University in Ontario, Canada. He later learned to build image classifiers using TensorFlow and a graphics processor.
One early project alerted him when a friend’s pet bird appeared on a webcam during shared video game sessions. Alexiuk said he wanted to see the bird without watching the webcam continuously while playing.
He contrasted that project with his ability to build a website in minutes using a model running on his laptop. He said the change illustrated how much easier it had become to start learning and working with AI.
