AI Cloud Computing: Faster and Easier…Right?

The Mental Burden of Google Cloud and Azure:

With the extreme cognitive burdens that have been placed on data scientists / researchers and data engineers to keep up with AI, these AI cloud companies aren’t making it easier… but they should be. The community needs the barriers to entry to be reasonable. 

Big Cloud Companies started with large, complicated ecosystems. Now, they’ve tacked robust AI services onto them creating a vast ecosystem. It’s overly complicated. To compound the situation, the documentation is all over the place. data engineers need to reference many screens of information to unblock problems. 

To punctuate the amount of overwhelm, Google has a recent Blog Post

"The overwhelmed person' guide to Google Cloud"

It almost seems like Google has accepted this state of being with this title. 

We need to get to flow state quicker!

New users need to quickly understand the overarching mental model of how the many elements of the ecosystem work together so we can quickly get moving on our projects.

Give us simple, clear way-finding based on common projects so that we can achieve momentum more quickly. Don’t bury us in documentation in multiple locations when we’re trying to work at the speed of AI.

Aspire to have UX that's as delightful as favorite consumer apps

While Data Engineers / Data Scientists are great at digging through problems and workflow blockers they already have too much on their plate. Let’s get back to the early days of the internet when we wanted to delight people with user experiences that anticipated their needs ahead of time. 

Data Scientists - Engineers are the salesforce

Due to the complexity of AI technology data scientists and data engineers create the product, test the product, put the product into production, and estimate the cost of the product. When you compare this to any other product category it’s an unusual situation. Therefore, friction and blockers in the user experience prevents us from closing the deal faster with the business people.

Pricing is too complicated and too stressful!

Data Engineers - Scientists have historically dealt with a lack of transparency and confusion around AI service pricing. You can go to any Discord and see developers asking the community lead to clear up a billing situation. There’s also historically been some elusiveness on billing. 

Again this is a weird situation that we’re in. Because the technology is so complicated it's difficult for someone that doesn’t understand the tech to trouble shoot billing issues. 

We carry this emotional baggage with us when we need to figure out the pricing of the Big Cloud Services. And I can tell you my emotions were triggered when using the free trials of Google Cloud with Vertex AI and Azure with AI Studio. Billing is its own ecosystem, and it's deep and complicated. 

I stopped using Vertex AI because they charged by the hour and I had accumulated $200 in billing of the $300 free trial just to figure out how to use the ecosystem. I hadn’t even yet deployed an LLM or trained any models. 

At the time of this writing I hadn’t gone as far into Azure’s billing but I was told by someone there that they didn’t think that Azure Studio had hourly billing. This is just one small example of many variables of a complex billing ecosystem. The bottom line is that we just don’t want to mess up the estimates and billing.  

Don't leave us stranded!

Getting onto these platforms feels daunting. You just want to figure out where you are and where you need to go. 

We strive to get a test project done to understand how the ecosystem thinks and then we can go further with other projects. 

We need a better map.