Atlassian does not use Chinese AI models in production, its CEO says
Mike Cannon-Brookes, co-founder and CEO of Atlassian, has stated that the company does not utilize any Chinese AI models in its products. In an interview with 20VC podcast, hosted by investor Harry Stebbings, Cannon-Brookes addressed the growing concern over Chinese open models in the AI landscape.
Customer Preferences and Open Models
According to Cannon-Brookes, while there is customer pushback against "Chinese models," Atlassian prioritizes open models. The company relies on models from Nvidia and Mistral, among others, choosing the most suitable model for each task at the lowest cost for its users.
"There is certainly some customer pushback in some places to, quote unquote, Chinese models. There is not customer pushback to open models."
Collaboration and Model Sources
Atlassian maintains a deep partnership with Nvidia, utilizing their Nemotron models. Cannon-Brookes also noted that many US models are trained on Chinese data without explicitly stating so, suggesting a complex web of model dependencies.
The Future of Specialized Models
The CEO expressed skepticism about the long-term benefits of specialized models, stating that they would need to consistently outperform general models in terms of price and performance to gain widespread adoption.
Cost and AI Gateways
Cost is a significant consideration for large customers, and Atlassian has established an AI gateway team to efficiently allocate tasks to the most suitable models. Cannon-Brookes advises product teams not to worry excessively about costs at the early stages of feature development.
Conclusion
In a broader context, the interview highlights the ongoing debate around AI model development, deployment, and transparency. As AI continues to evolve, the balance between cost, performance, and ethical considerations will remain a crucial aspect of shaping the future of this technology.