All about Bedrock, Amazon’s answer to ChatGpt
This is where the projects of Amazon Web Services (Aws), the cloud division of the US group Amazon, are in the artificial intelligence sector
A new AI race has begun among major tech companies.
After OpenAI rose to prominence last fall with ChatGpt's (Microsoft-backed) generative AI, and Google rushed to launch Bard, and Meta (Facebook's parent company) developed a large-scale language model, last April Amazon also presented (almost quietly) the Bedrock service. But rather than build AI models entirely on its own, the e-commerce giant founded by Jeff Bezos is recruiting third parties to host models on AWS.
“We've been investing in AI for more than 20 years,” Matt Garman, manager, head of global sales and marketing for AWS, said in a recent interview with Faz .
From recommendations in online commerce are based on artificial intelligence, as well as the Alexa voice assistance system or the automatic route calculation of robots used in Amazon warehouses. “But we've been developing generative AI for a few years now,” says Garman. For a long time, the company kept the curtain down on all of its Ai work, until it quietly revealed several new things in a blog post in the spring.
All the details.
WHAT STATE IS AMAZON WEB SERVICES' BEDROCK PLATFORM?
As we mentioned, in April Amazon unveiled an AI platform called Bedrock, which gives AWS customers access to more large AI models. Two of them were developed by Amazon itself and are called Titan. One of them is a classic text-based template similar to Open AI's Gpt templates. The other is designed specifically to perform search functions and can provide more accurate results than a classic keyword search, according to the company led by Andy Jassy. Customers can then further customize the models to their liking and train them with their own data.
AVAILABLE TO SELECTED CUSTOMERS ONLY
“Bedrock” is still in the so-called beta version and is therefore only available to selected customers so far. Among them are also German companies. However, Amazon doesn't want to reveal which ones just yet. Among the official clients are the consultancy firms Accenture and Deloitte.
NEW MODELS COMING SOON
Amazon has been working on developing the software for “some time,” Garman tells the German newspaper. He doesn't want to reveal how long exactly. However, “there will certainly be other models added by us”. This is mainly related to the cost of running large AI models, says the AWS manager. The larger and less specific they are, the more expensive they are to use and vice versa.
BUT ALSO THIRD PARTY PROGRAMS
In addition to your own templates, however, partner programs are also available. These include former Open AI employee model Anthropic and AI firm Stable Diffusion, which is among the leaders in image generation. This makes Amazon's approach very different from that of Google or Microsoft, which have so far relied mainly on their own models. “We don't flatter ourselves that we're the only ones who come up with cool stuff,” says Garman. Customers will combine the various applications on the market, the AWS representative told Faz . "There won't be a model."
GOOGLE AND OPEN AI INTEGRATION? IT COULD BE DONE FOR AWS
Additionally, Garman said the company is even open to integrating Google or Open AI programs into the platform.
At some point, all models are likely to be available from different vendors, the Amazon manager argues, just “how Windows runs today in an AWS environment.”
Garman doesn't see much benefit for Microsoft or Google in the race for the technology of the future. "We only took three steps in a 10-mile run."
FOCUS ON PRIVACY
Finally, if nobody dominates in the race for artificial intelligence at the moment, according to Amazon, the company has identified data security as a competitive advantage. The main concern of companies regarding artificial intelligence is how to protect their intellectual property, points out the manager of Amazon Web Services. ChatGPT, for example, can only be used via interfaces. During this process, the data automatically travels across the Atlantic. Some companies, including Samsung and Apple, have even banned their employees from using the service, for fear of feeding sensitive data to the AI. In recent days, both Twitter, the micro-blogging platform owned by Elon Musk, and Reddit have deactivated free access to their APIs, which until now had allowed anyone who wished to download large quantities of posts precisely to avoid "scraping". data for training AI models.
And it was precisely against OpenAI, the company that developed the popular ChatGpt artificial intelligence software, that a class action lawsuit was recently launched by California which could change the development of this software in which the whole of Silicon Valley is involved. According to the accusers, ChatGpt would fish the data without asking the "permission" of the right holders, in short, the owners of the data that each artificial intelligence processes and reprocesses to work better and better. Hence the various accusations: from the actual theft to the violation of copyright passing through that of privacy.
This shouldn't be possible with Amazon. The company promises an interface within a private network that doesn't leave its AWS environment. “Corporate data doesn't go back to the model,” Garman promises, and takes a little lateral jab at the competition. “We didn't just think back on it, we thought about it from the beginning.”
This is a machine translation from Italian language of a post published on Start Magazine at the URL https://www.startmag.it/innovazione/tutto-su-bedrock-la-risposta-di-amazon-a-chatgpt/ on Tue, 11 Jul 2023 05:46:31 +0000.