Postcards from tomorrow: Powering the agentic era through open source
Organizations initially adopted generative AI primarily through hosted APIs, The New Stack reported on November 5. The publication noted that open-weight models represent an alternative approach to engaging with generative AI.
When organizations first began to use generative AI, the predominant method involved accessing models through a hosted API. This means that a third-party service provided and managed the AI model, with organizations connecting to it over the internet.
This approach allows organizations to utilize generative AI capabilities without necessarily needing to host or maintain the complex AI infrastructure themselves. The report introduces open-weight models as a contrasting method for engaging with this technology.
Open-weight models differ in that their core numerical parameters are publicly accessible. This allows organizations to download and run these models on their own infrastructure, potentially offering more direct control over the deployment and operation of the AI.
Share this article
Related reading
6 stories
OpenAI’s Dots are coming up against the price of privacy, and the price of compute
We previously explored the compute and privacy costs associated with OpenAI's hosted models, a contrast to open-weight.
Remote Launches Virtual Expense Cards for Global Employees
AI bubble bursts, fifth plenum resolution, Singles’ Day
Can AI Help Small Businesses Go Global Without a Big Finance Team?
TikTok Rolls Out a Standalone TikTok Shop App in Singapore and 3 Other Southeast Asian Countries
