AI isn't surprising anymore. Why don't you have it?

Are you involved in the roadmap of a product? Speech, text recognition, and similar features should be there!

I'm a proud early adopter. That means you invest significant time and effort in understanding new technologies and making bets for the future. I nailed it when I invested in learning about Docker but got fewer rewards for some technologies that never went mainstream: Cloud Foundry and Blockchain (yes, there is a use case for Blockchain somewhere, but it's not a good one, so stick to your databases and data services unless you have a strong case). When Google open-sourced Kubernetes, I just thought "Cloud Foundry does that" and rode the wave very late. I got mixed results betting on AI, and in a world where everyone has a voice assistant at hand, I wanted to reflect back to some of the history and clarify why your products should adopt and implement is as any other user story.

Back in 2011, I was a proud IBMer and watched closely how Watson played and won Jeopardy. Back then, almost a decade ago, Natural Language Processing was, to put it mildly, frustrating. But somehow, these guys from IBM Research were able to break down NLP, add text processing, had to map huge amounts of training data, and built a system that answered natural language questions in seconds. This video is a very passionate review of how it was back then. To give some historical context, Siri was integrated into the iPhone in fall 2011 (and it was very beta).

I remember vaguely the white papers around the infrastructure set in place for it, but I do remember something very clear: Watson was an On-Premises solution! A full data center, 90+ last generation servers and all the storage needed to get the responses down to seconds and compete with Jeopardy legends.

Then Cloud happened, and we were in the middle of API economics, so Watson became an API exposed to the world. Ginni Rometty, IBM's CEO, announced to the world the start of the "Cognitive Era". A big marketing effort was put in place to dominate the market, and you could hear the word AI in every single technology conference in the world: cognitive health, cognitive banking, cognitive insurance, cognitive swimming pools.

Amazon, Google, and other smaller companies caught up to Watson and even expanded some features to give differentiate. You can find a conversation API in virtually any Cloud vendor nowadays, and you can build your conversations in user-friendly interfaces everywhere. You can have your chatbot in seconds, you can build your speech recognition system, you can get a very accurate description of your pictures from an API. Or you can buy an Amazon Echo: It works well for most needs if you ignore privacy concerns. Siri even has very lame dad jokes.

So, let's get a couple of things straight:

AI is not one thing. AI is a marketing name we are giving a bunch of technologies. Some of these we expect in most modern apps and are focused on user experience: speech recognition, text to speech service, conversation services, natural language processing. Do you have end-user interfaces? Adding a conversation-based interface is a couple of libraries away for your developers if you think it. You can also use categories extraction functionalities to get some metadata on the conversation or get a sense of the satisfaction your clients are getting. All those androids with human expressions? They have preprogrammed conversations, and perform the role of marketing tools.

There are also data-oriented services. Some are sophisticated on visualization skills, and data processing, but some are just glorified spreadsheets. I would say some of these are tools that resemble the promises of Data Analytics with different approaches. They will aid with decision making adding a layer of prediction to it or will adapt to your data sets for better responses. Data Scientists use these tools or build their own algorithms to have better use of data. Most businesses do need some level of quality in their models to account for, and although some services will do some of the jobs for you, I would recommend getting a good Data Engineer to work around it. Most of these technologies have been around for a long time, but with Cloud came the possibility to monetize models and algorithms through an API, and sometimes, that's good enough for your needs.

Something AI is not, is a self-learning brain. You still need to code and train these services or build your own custom machine learning algorithms. There's a future for coding, and software engineers are needed for a long foreseeable future.

I wrote this small walkthrough of AI today because I was asked to give an opinion of Salesforce Einstein and to be honest, Einstein is a smart name they use to showcase the implementation of all these technologies as features to their product. I can't say it's good or bad, but I will say it is something that we should be expecting from any product today.

To summarize:

• Cloud democratized AI, making it widely available. It stopped being rocket science a decade ago.
• Current offerings of AI are exposed as API, implemented through libraries and SDK's and any developer can code it.
• Be aware of the capabilities and new interactions this brings to your products. Anyone can build a web page, but think about the product lifecycle and include new interactions with these technologies.
• We need better comedians to write voice assistants jokes!

Interested? here are some links to offerings out there:

https://www.ibm.com/watson/products-services

https://aws.amazon.com/machine-learning/

https://cloud.google.com/products/ai

https://www.salesforce.com/products/einstein/overview/