“While Amazon doesn’t have a ChatGPT-type product to advertise its AI capabilities, the breadth of its cloud services will give it an advantage in selling generative AI to others. The breadth of AWS is going to be interesting,” argues Steven Dickens, CEO and principal analyst of the firm specialized in the global technological panorama, HyperFRAME Research. He also explains that the company’s line of chips will help make the software of AI offering: “Silicon has to be a key part of any hyperscaler’s strategy in the future. Amazon has been developing its custom chip for longer than its competitors.”
Garman says an increasing number of AWS customers are moving beyond demonstrations to building commercially viable products and services that incorporate generative AI: “One of the things we’re most excited about is customers designing their own AI experiments.” and proofs of concept. He asserts that many customers are much less interested in pushing the boundaries of artificial intelligence than in finding ways to make the technology cheaper and more reliable.
New AWS tools
AWS presented its new service Model Distillationwhich produces a model that is smaller, faster and less expensive to run, while still having similar capabilities to a larger one. “Let’s say you’re an insurance company, you can take a whole set of questions, feed them into a really advanced model, and then use that to train the smaller model to run whatever you ask it to,” Garman says.
Another cloud tool is Bedrock Agents– Can be used to create and manage AI agents that automate useful tasks with customer support or order processing and analysis. It includes a master agent that will manage a team of AI subordinates, providing reports on their performance and coordinating changes: “You can basically create an agent that says you are the boss of all the other agents.”
Garmin hopes businesses will be excited about Amazon’s new tool, which will “ensure that the results of a chatbot be accurate.” LLMs are prone to hallucinations, and existing methods to keep them sane are imperfect. “Insurers can’t afford to make mistakes with their AI model, they’re crying out for this kind of safeguard. When you ask, ‘Is this covered by my insurance?’, you don’t want the model to say no when it is or yes when it isn’t,” the CEO explains to WIRED.
Amazon’s new verification tool, called Automated Reasoningis similar yet different to a product OpenAI launched earlier this year. It is based on logical reasoning to analyze the results of a model; To make it work, a company has to convert its data and policies into a format that allows logical analysis: “We take natural language, translate it into logic, prove or disprove the claim, and then we can provide an argument for why the claim is made.” statement is true or not,” said Bryon Cook, AWS scientist and vice president of the company’s Autonomous Reasoning Group.
Cook confirms that the same type of formal reasoning has been used for decades in areas such as chip design and cryptography. He adds that the method could be used to create chatbots who manage airline ticket refunds or provide human resources information without mistakes. Companies can combine multiple systems that incorporate automated reasoning to create more sophisticated applications and services, including those that incorporate autonomous agents. “Now you have conversational agents that perform formal reasoning and communicate their reasons. Reasoning will become a very important issue,” Cook concludes.
Article originally published in WIRED. Adapted by Alondra Flores.
#Amazon #works #supercomputer #Anthropic
