AI for Everyone? Kind of

Recently, Salesforce started a campaign concerning their latest product offering, Einstein. This einsteinplane1campaign promises, “AI for Everyone”.

That is quite a claim and Salesforce is certainly not the first to make it. As a self-proclaimed AI Enthusiast/junkie, I can say that I have seen this type of claim before. However, this is the first time that I think it might actually be somewhat possible. At least as far as a specific area of AI known as deep learning is concerned. And, also if were talking about relating this just to Salesforce in particular.

So, does this mean that soon everyone will have personal robot butlers?

Absolutely not. We are not even beginning to talk about robotics here. Nor some other areas that fall under the rather large AI umbrella.

So what can Salesforce customers do with Einstein?

Well that will certainly change as the product evolves, but right now the most relevant thing you can do with it is to utilize the Predictive Vision Service (PVS). This can be used to classify images into categories using supervised learning and very specifically optimized machine learning algorithms. These algorithms were developed by a company called MetaMind, which was last year aquired by Salesforce and since then they have been working feverishly to offer their services on the Force.com platform.

If you are interested in learning more about how this works, check out the docs here or this recently released webinar, which does a very good job of laying out what is currently possible with the PVS.

Note that currently Salesforce is not offering a service that does Natural Language Processing (NLP).  But, I am sure that will be the next big thing Salesforce customers will be demanding. NLP is a huge field and one that has been around for many years, but with varying levels of success.  The most difficult challenge I suspect will come from the fact that the product will need to support several languages beyond English to be considered useful. It will also need to be able to handle untrained users with a high degree of accuracy, which is a very tall order to fill.

It appears to me that the majority of Einstein’s capabilities will be “Baked in” to many of Salesforce’s products and their use should be seamless to users. They will also be very specific to Salesforce CRM.

The most important thing to understand is that Einstein is NOT a general purpose AI engine. As enthusiastic as the Salesforce Marketing team obviously is, Salesforce has not reinvented the wheel and certainly not developed some new and unheard form of AI that will corner the market.

BUT, they have started to offer some very useful API’s that can be used to implement specific areas of AI that were once only accessible to the elite of AI researchers.

And the most promising news was just announced this month when a group from Salesforce Research created a neural network named the Dynamic Coattention Network and that model was the first to break the 80% mark when tested against the Stanford Question Answering Dataset. And for those of you that just said to yourselves, “and why should I care about that?”.

Well, ever since Stanford released their dataset, which now consists of questions posed by crowdworkers on a set of Wikipedia articles, lots of top AI researchers (including Microsoft, Google and IBM) have been racing to create models that will reach this golden threshold, but Salesforce was the first to reach it. It’s kind of a big deal.

I look forward to the next few years and seeing all the new services that will be added to the platform, bringing about the Enhanced Computing world I always envisioned.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s