Revolutionizing Creativity: Amazon's Bedrock Generative AI Service Goes Public

0:00Today, Amazon has officially launched Bedrock, which is its centralized service for curated selection of generative AI models. And this is really, I think, opening the door for AWS customers to incorporate these AI models into their applications and operations. So first introduced in April, right? This is not new. This is something that Amazon kind of, I think, when the whole AI wave started, they made the announcements and kind of started talking about way back in April. This is a trend I think we're seeing a lot in the industry. Even today, you know, I'm seeing announcements out of Google and Microsoft where they're like, yeah, we're going to do this really cool thing, but it's going to launch in Q1 of next year.

0:35So, you know, I think people are kind of, there's a lot of hype around AI. They're making their announcements now, marking their territory, and then they kind of work towards those. And that's kind of what Amazon did. But Bedrock now is live and it offers a fusion of Amazon's own generative AI models and those from third party partners, which is all accessible through an API. So the service aims to provide businesses with the tools they need to develop AI agents that can perform tasks ranging from booking travel to managing inventories and even things like processing insurance claims. So Amazon's bedrock appears at a, I think, really critical time in the adoption of AI.

1:13It's definitely providing a big array of, you know, options to brands and developers. And according to the company, the platform is soon going to feature Llama 2, which of course is an open source large language model developed by Meta. This is going to be alongside models from other industry players. You know, I think they have AI21 Labs, Anthropic, Cohere, Stability AI, a lot of these big players, right? And Amazon purports that Bedrock will be the first quote unquote fully managed generative AI service to essentially offer both the 13 billion and 70 billion parameter version of Llama 2. So for the uninitiated on this, parameters essentially constitute the part of an AI model that are learned from historical data.

1:55So determining the model's efficacy and problem solving tasks like text generation, for example. So however, all that being said, I think it's a noteworthy. The Lama 2 isn't a newcomer to cloud based generative AI services. It's also available on Google's Vertex AI. So Google kind of beat Amazon and AWS to that. but when we're kind of looking I think to the future looking at the difference between Bedrock versus Vertex AI. Vertex AI is definitely is kind of like Google's platform for this but it's something I've talked about a lot in the past. I wanted to talk a little bit about what I think is perhaps the AWS edge here and so I think while comparisons between Bedrock and Google Vertex AI are inevitable Amazon makes a really compelling case for its platform's unique advantages.

2:40So Swami Sylvester Bramian is the VP of data and AI at AWS. And they kind of pointed out that Bedrock seamlessly integrates with existing AWS services such as AWS PrivateLink. So this enables a secure connection between Bedrock and a company's virtual private cloud. and so you know swami was recently saying quote over the last year the proliferation of data access to scalable compute and advancements in machine learning have led to a surge in interest in generative ai sparking new ideas that could transform entire industries and reimagine how work gets done so i think it is worth noting that this supposed edge might be more um you know perception than reality.

3:27The purported advantages kind of depend largely on a customer's current cloud infrastructure, which is an aspect that I think Swami does not kind of directly acknowledge. So when we're kind of looking at this, I think at the end of the day, this really does provide more tools in the arsenal. And so what I mean by that is that right now in tandem with Bedrock's announcement, Amazon has also rolled out its Titan embedding models. So this first party model focuses on converting text into numerical representations known as embeddings. And this is done to enhance search and personal personalization applications.

4:06So the Titan embeddings model boasts support for around 25 languages, and it can process text chunks, or even entire documents of up to over 8 ,000 tokens, which roughly is around 6 ,000 words. So this places it right on par with the latest embedding models from OpenAI. But you know, sometimes we're seeing bigger ones in other places. In any case, there's definitely a lot of hurdles to overcome here. And some of these initial ones, I think they're really focusing on. So Amazon's bedrock services definitely has not been without its challenges. Bloomberg reported in May that the platform struggled in its early days, lacking widespread access among cloud customers.

4:48And this was six weeks after its debut with an ambitious kind of press release and I think just one testimonial. But in any case, Amazon appears to be doubling down on its investment in AI, especially following its recent multi-billion dollar stake in AI startup Anthropik. I think originally, right, they gave Anthropik over a billion dollars and said they'll invest up to$4 billion. dollars and this is kind of through AWS making sure that Anthropic is on you know AWS's platform for their compute. So I think today's announcement kind of marks Amazon's aggressive play into this growing market for generative AI a sector which is seeing increasing interest and investment across industries.

5:29I think with Bedrock Amazon is really aiming to kind of democratize access to generative AI models. It's offering a really robust platform for businesses ranging from startups to establish enterprises. I think it's going to make a big difference in the industry moving forward because AWS, of course, is the number one cloud provider for a majority of businesses. And I think, you know, having these tools built right in is going to really accelerate adoption and, you know, make it a lot easier for businesses to implement a lot of these tools. So very interesting to see how this progresses in the future, but definitely big steps from Amazon.