Image of the interior of an Amazon Go store, especially the entrance.  A shopper takes fruit from one of the shelves.

Why this VP of AWS believes Generative AI has the potential to transform our lives

Think back to when a new set of technologies or a tech-enabled gizmo completely captured your attention and imagination. The first personal computers. The advent of the internet and the web. E-mail. Smartphones. These things have changed our lives in ways that were hard to predict, and perhaps appreciate, until we had some time with these technologies under our collective belt.

We’re back in that time again with artificial intelligence (AI) and machine learning (ML). I believe AI and ML are the most revolutionary technologies of our time. That’s why, for over 20 years, Amazon has invested heavily in AI and ML developmentinstilling these extraordinary capabilities in nearly every business unit.

Why Generative AI is in the spotlight

AI was the domain of a small group of researchers and data scientists. Today you can’t open a newsfeed without some reference to AI and especially generative AI. It might come as a surprise, but AI concepts have been around since the 1950s.

So why is this technology, which has filtered through for decades, garnering so much interest now? Simply put, AI has reached a tipping point thanks to the convergence of technological progress and a greater understanding of what it can do. Couple that with the massive proliferation of data, the availability of highly scalable computing capacity, and the advancement of ML technologies over time, and the focus on Generative AI is finally taking shape.

Image of the interior of an Amazon Go store, especially the entrance.  A shopper takes fruit from one of the shelves.

Also, it is very likely that you already have experience using AI and ML. If you’ve listened to a Wondery podcast, asked Alexa for the day’s forecast, searched Prime Video for a new series, or visited a store featuring Just Walk Out technology, you’ve tapped into Amazon AI. More specifically, have you interacted with ML systems or models. It is these ML models that lie at the heart of the excitement and potential of generative AI.

So what exactly is Generative AI? And how does it differ from other AIs?

While based on the same concepts, there is a clear distinction between traditional AI machine learning techniques that we have been using for years, especially deep learning and generative AI. As the name suggests, Generative AI is a type of artificial intelligence that can create new content and ideas. It can be text, images, video, voice and even code. Like all AI, Generative AI is powered by machine learning models, very large ML models that are pre-trained on large amounts of data and commonly referred to as foundation models (FMs).

Before getting FMs to work, traditional forms of machine learning allowed us to take simple inputs, like numeric values, and map them to simple outputs, like predicted values. With more advanced ML techniques, especially deep learning, we could take somewhat more complicated inputs, like video or images, and map them to relatively simple outputs. You might be looking for an image in a video stream that conflicts with guidelines or analyzing a document for sentiment. With this approach, you get insights into the data you feed the model, but you don’t generate anything new. With Generative AI, you can harness massive amounts of data by mapping complicated inputs to complicated outputs and creating new content of all kinds in the process.

Even traditional ML models tend to be task specific. If I wanted to do translations with a deep learning model, for example, I would have access to a lot of specific data related to translation services to learn how to translate from Spanish to German. The model would only do the translation work, but it couldn’t, for example, continue to generate recipes for paella in German. It could translate an existing Spanish to German paella recipe, but not create a new one.

Spanish to German translation of the instructions on how to prepare a paella.  A plate of chicken and rice serves as the background image behind the text.

A unicorn on the beach at sunset

Now, with Generative AI, anyone can use AI without manual data preparation. The large models that power generative AI applications are built using a neural network architecture called Transformers. It arrived in AI circles around 2017 and greatly shortens the development process.

Using the Transformer architecture, AI models can be pre-trained on huge amounts of unlabeled data of all types: text, images, audio, etc. There is no manual data preparation involved, and due to the huge amount of pre-training (basically learning) the models can be used out of the box for a wide variety of generalized tasks. It’s kind of like AI’s Swiss Army Knife.

A majestic unicorn soars over the beach at sunset with snow capped mountains and a rainbow in the background.

Image generated using SDXL on Amazon Bedrock. The initial suggestion was: “A majestic unicorn stood on its hind legs on the beach at sunset.”

A model can learn in the pre-training phase, for example, what a sunset is, what a beach looks like and what the special characteristics of a unicorn are. With a template designed to take text and generate an image, not only can I ask for images of sunsets, beaches, and unicorns, but I can have the template generate an image of a unicorn on the beach at sunset. And with relatively small amounts of labeled data (we call that fine-tuning), you can tailor the same basic model for particular domains or industries.

Generative AI applications: Generative AI will transform the way every business and organization operates

The ability to customize a pre-trained FM for any task with just a small amount of labeled data – that’s what’s so revolutionary about generative AI. It’s also why I believe the biggest opportunity ahead of Generative AI is not with consumers, but in transforming every aspect of how businesses and organizations operate and how they deliver to their customers.

In healthcare, the legal world, the mortgage underwriting business, content creation, customer service and more, we expect expertly tuned generative AI models to have a role to play. Imagine if automated document processing made filing taxes quick and easy and making your mortgage application a simple process that took days, not weeks. What if conversations with a healthcare professional weren’t just transcribed and annotated in clear text, but offered the doctor potential treatments and the latest research? Or if you could explore the design of a new product, optimizing sustainability, cost and price with simple suggestions. All of this is not only possible, but probable with generative AI.

A doctor shows a patient an image describing the heart's functions on his laptop in a clinic room.

We are already seeing a pattern emerge in how Generative AI will present itself in enterprises through four main modalities.

  1. Improve the customer experience through features such as chatbots, virtual assistants, intelligent contact centers, personalization and content moderation.
  2. Strengthening employee productivity with conversational search, text summarization, and code generation, among others.
  3. Production of all kinds of creative content from art and music to text, images, animations and videos.
  4. Improve business operations with intelligent document processing, maintenance assistants, quality control and visual inspection, and synthetic generation of training data.

The key is to make sure you’re actually choosing the right AI-enabled tools and pairing them with the right level of human judgment and expertise. These models will not replace humans; they will make us all much more productive. More importantly, you need to optimize these models with your data in a secure way, so at the end of the day these models are tailored to your organization’s needs. Your data is the differentiator and key ingredient in creating great products, customer experiences, or improved business operations.

Like the Internet in 1995

These are still early days for generative AI. There is much more to invent and repeat. It reminds me of the internet circa 1995 when the web was just starting to come into being and we heard about this thing called a web browser.

Image of a retro 80's computer with two hands typing on the keyboard.  The computer screen shows the code and an old phone and a slinky are on either side.

When you step back and look at where we are today and what is yet to come, generative AI has the potential to revolutionize our lives, whether at home, school or work. With these tools that Amazon and our customers are building, we will all be able to spend more time on what we do best and less time on routine work. It’s quite powerful, and that’s what will make this moment so incredible.

For more information on Generative AI and the latest AWS tools, Check out my keynote at the AWS New York Summit.

#AWS #believes #Generative #potential #transform #lives
Image Source :

Similar Posts

Leave a Reply

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