Salesforce Einstein makes a breakthrough in machine understanding of language – and it only took 3 months…

I saw this article come up on my google feed 2 days ago and was intrigued to say the least.  I don’t yet completely understand how AI can work in tandem with CRM, but at the very least, the way that AI is beginning to understand what we do is interesting – and frightening…

This article talks about 3 big trends in AI programming – the first deals with document relationships – not just interpreting documents by their content, but by what questions were asked to access them in a search engine.  I think this is interesting in a Service model, using AI to determine WHY a specific knowledge article was accessed not just THAT it was accessed.  Customers asking for information on how to articles could be served up answers on functionality, because the system begins to understand that the question how to is really more about a function than self help…

The 2nd breakthrough involves contextual learning – which is kind of weird but in a way we use it every day.  If there’s a word in a sentence that we don’t understand, we’ve been trained since childhood to look at the words around that word to figure it out.  In this case, AI would do the same – and instead of forgetting that word immediately (like I do), remember it for future searched.  Pretty cool…

Finally,  Salesforce is claiming that they’ve made a breakthrough in combining natural language processing, with Visual recognition.  Imagine a system that figures out the context of an article, not just on keywords, but on the screen shot you add to the article – pretty slick stuff.

Lots to be proven here, and as this article says, the proof will be in the subscription based pudding, but all in all, a really interesting angle on an ever evolving, ever expanding business.  It’s good to be a Salesforce partner for sure…

 

read more here.

 

You can share this article by using your social accounts:

Related Posts

Leave a Reply

Your email address will not be published. Required fields are marked *

Archives