Move over, Large Language Models: Microsoft thinks a Small Language Model is often a better fit. They’ve just released something called Phi-3-mini, which can easily run on a smartphone or laptop computer. This makes custom models cheaper to build and supports applications where a network connection is unavailable.
Martech keeps taking larger bites out of marketing budgets: 17.3% last year, 19.9% this year, 23.5% next year, and 30.9% five years from now, according to the latest CMO Survey. This despite barely more than half (56.4%) of current tools being used and nearly half (48.8%) of the survey respondents reporting worse-than-expected results. Oddly enough, marketers rate selecting marketing technologies as the thing they do best.
Security is another reason to build your own custom models. Cybersecurity firm Riversafe surveyed 250 U.K. CISOs and found that around 20% have had sensitive corporate data leaked through employee use of generative AI. Without proper controls, data entered into a model like ChatGPT may be visible to other users of the tool.
If you still think that MCP stands for Male Chauvinist Pig, well, the 1970’s want their disco ball back. Today’s hep cats know that MCP stands for Model Context Protocol, a new-but-widely supported standard that lets large language models access structured data as context for their prompts. Early adopters include Salesforce and Adobe, and search platform Algolia is now using it to feed their data to customer-facing real-time AI agents.