If your organisation is a modern, data-driven business, chances are you’re considering how AI can help you if you’re not already using it. Well, you’re in luck - AI is compatible with GCP thanks to its various managed AI services. We’ll cover a few of them in this blog. Continue reading to find out more about the functionality of AI integrated with GCP and how you can benefit from it!
Why use GCPs AI tools?
AI tools on GCP are great for two main target audiences
Companies that have some data but don’t know what to do with it. This is called ‘data ideation’ GCP has some powerful tools to help you bring an AI process to life quickly and easily.
Companies that are already running an AI-based process but want to ‘streamline’ them and use best-in-class cloud tech to do so.
Before you continue 🤚
AI is usually the last step or the second last step in an organisation’s data maturity development. If you don’t fit these prerequisites, you should consider increasing your data maturity first. Here's what we think the 'ultimate GCP data stack' looks like, and as you can see, AI comes right at the end.
What to do when you don’t have a data strategy?
If you’re not capturing any data or don’t have any sort of warehousing strategy, you’re going to have a very tough time with AI. You could be working with well-defined, clean datasets when training AI models; such as your warehouse inventory or transactional data, but if that data isn’t well structured, changes all the time, or is inaccurate, then you’re going to be in a world of pain.
We recommend using BigQuery for your analytical storage and using any AI service on top of that.
Lack data capacity
You can only take things so far as a lone-AI user. Despite Google offering fantastic serverless managed AI services, which means no patching, updating, infrastructure, or all that good stuff, it doesn’t mean that you can sit back and put your feet up. Business needs change, people change, and data changes. These things happen all the time. If you don’t have the capacity to deal with change, you’re in for a nightmare trying to get production AI workloads to change with business requirements.
You may ask, how can I increase data capacity?
Hire.
Get a data partner (not to pat our own back, but we’re pretty good!)
Reduce your current workload. You’d be surprised at how many things you can bin to free up time.
Automate or transfer ownership internally. There are technologies and strategies you can put in place to decrease the load on your data team.
Don’t have defined benchmarks of success
Ask yourself - what are you trying to achieve? Hopefully, this article helps inspire you, but ultimately you know your data and your business goals better than anyone. You need to know what you want before you commit to using technology. Ask critical questions like; What do we want? When do we want it? Why do we want it?
What are you trying to achieve?
We love that question at cobry, so ask yourself that first. Once you have an answer, cross reference it with the headings below.
We want to deploy or use a large language model
Vertex AI is your friend.
We want to apply machine learning to an analytical database
If you're a data scientist, Vertex AI is your friend, if you're a developer, you have more freedom. But we need to dive a bit deeper.
Are you trying to do prediction? Anomaly detection?
We want to apply AI to media
Ok, cool. What kind of media?
Vision 👁️
If you’re collecting image data in one way or another, you have incredible information at your fingertips. However, processing image data conventionally is very difficult, and typically involves convolutional neural networks. This often puts image processing out of reach for companies that don’t have expert data scientists on their teams. GCP’s Vision service enables you to process image data without having to do the hard work of training a model. With Vision, you can classify and label images. It’s perfect for detecting people or specific objects, recognising text, and detecting graphic content and sentiment.
I was able to use the Vision demo to detect a person, a hat, and a piece of broccoli in this image all without writing a single line of code.
If you want a bit more control over the model’s behavior, you can also use Vertex AI AutoML Image, which allows you to train and optimise Vision’s output by providing it with your own training data. GCP’s toolset is consistently designed for the best of both worlds. Whether you want total control over the code, the model, and the training, or you couldn’t care less, image processing with Vision is painless.
For a really cool story about Vision being used in real life, read how Texas A&M University uses AutoML Vision to track environmental change
Natural Language 😛
Natural language processing is a particularly challenging area of machine learning, and implementing an effective solution on your own can often prove impossible unless you have a lot of experience using machine learning. Just as with Image Recognition, NLP is another area of ML best suited for CNNs. Cloud Natural Language gives you powerful text analysis to pull insights from emails, chats, reviews, and social media: literally any source of text. For instance, you could pull the sentiment from reviews effortlessly to inform your next UX changes for your app.
I used the Natural Language Demo to analyse the sentiment from a passage from my favorite book - ‘Welcome to Night Vale’.
Just like with Cloud Vision, you can use GCP’s AutoML feature to take care of all the heavy lifting and setup. Again, you can hone in and configure things yourself if you have the know-how.
Cloud Natural Language also comes with a discrete package for Healthcare use cases. You can extract information from the unstructured medical text to determine insights in a medical context, all without code.
Here’s a story in which Hearst Newspapers used the NLP API to sort, label, and categorise an average of 3,000 new articles every day. Neat!
Everything else…
The AI product stack on GCP is vast, and it covers tens, if not hundreds, of use cases. Whether you need translation, classification, text-to-speech, speech to text - it does all of that.
Vertex 🧠
Vertex AI is GCP’s brand-new unified AI solution. You can define your entire machine learning workflow from the very start to end, all from the Vertex UI. It boasts pre-trained models for image processing, natural language processing, as well as your more conventional machine learning techniques such as classification. It also offers a seamless connection with BigQuery, and can apply labels straight to your data without it leaving BigQuery. Again, AutoML takes care of the tricky stuff for you, and there are pre-trained models available to you to use immediately.
If you want to do some of the work yourself, Vertex AI comes with custom toolkits that augment the common machine learning libraries, such as sci-kit-learn, TensorFlow, and PyTorch, which leads to approximately 80% fewer lines of code needed to build a model. Managing the lifecycle of your model is also fully contained within Vertex, allowing you to effortlessly monitor, tune and serve your model from the cloud console.
GCP is putting AI at your fingertips. We could go into a lot more detail about it here, but to see exactly how your business can start using AI, you should book a discovery call
Book a discovery call if you want to get ahead of the curve and benefit from GCP & AI. There are many tools out there to help with data management and analytics. At Cobry we believe in providing services well aligned with your organisation's needs to ensure that you get the most from your data.
Start your AI journey with Cobry! Book a discovery call & find out how you can get the most from your company's data.
