Amazon Introducing Rufus AI + Arc's AI Browser + OLMo 7B by Microsoft

In the last couple of days, there have been some major breakthroughs in the world of AI. From Amazon shopping assistant Rufus to AI2's open source AI tool Olmo 7B in Ark's innovative web searching browser, we're seeing groundbreaking changes in how AI is used and shared. First, let's talk about Amazon's latest innovation in the realm of online shopping, Rufus, an AI powered shopping assistant designed to revolutionize the shopping experience. Rufus, named in homage to the dog of Amazon's early editor in chief, symbolizes the company's commitment to innovation and customer centricity. Rufus is engineered to assist shoppers in making informed decisions, leveraging a vast array of data, including Amazon's product catalog, customer reviews, community Q&As, and web information. The tool operates on a large language model, allowing it to process and respond to customer queries in a conversational manner. This feature enables Rufus to provide in depth product comparisons, answer specific queries and offer customized recommendations. To access Rufus, customers can simply type or speak into the search bar of Amazon's mobile app. A chat window then appears, enabling users to interact with Rufus in a natural conversational style. This interaction could range from asking about the differences between product types like trail and road running shoes, to requesting comparisons between products such as Drip and pour over coffee makers. Rufus employs its AI capabilities to draw insights from various data sources, providing relevant and comprehensive responses. Amazon plans a phased rollout for Rufus, initially targeting a small subset of U.S. customers before expanding nationwide in the weeks to follow. This strategy allows Amazon to fine tune Rufus based on user feedback and usage patterns. While currently available through the Amazon mobile app, there are plans to extend Rufus to other platforms. The launch of Rufus is part of a broader strategy by Amazon to accelerate its AI initiatives. CEO Andy Jassy has indicated the company's intent to integrate generative AI across all its business areas. This move aligns with the tech industry's trend towards incorporating advanced conversational AI into online platforms and services. Amazon's exploration of AI extends beyond Rufus. The company has released various AI tools and services recently, spurred by the momentum generated by OpenAI's ChatGPT. These initiatives include AI tools for responding to shopper questions, summarizing reviews, assisting third party sellers in writing product listings, and beyond its retail domain, introducing Q a chatbot for businesses and Bedrock, a generative AI service for cloud customers. Amazon faces competition in the AI powered shopping assistant domain. Walmart, in partnership with Microsoft, has introduced its own generative AI shopping assistant. Startups like Seattle based Spiffy are also entering the market with similar tools and technologies for a broader array of retailers. However, Amazon's Rufus stands out due to its integration with the company's extensive product catalog and its vast customer base. Alright now, AI2, a non profit started by Microsoft's Paul Allen, is changing how AI research is done by opening up their work to everyone. They've just released a new AI tool called Olmo 7B and they're sharing everything about it, from the training data to the source code on websites like GitHub and Hugging Face. In the past, AI researchers used to share their findings freely, but when big AI models like OpenAI's GPT4 came along, things changed. Investors wanted to keep the details secret because they were valuable, but AI2 is going against this trend. AI2's senior director Hannah Hajisirzi explains that they're making everything about Olmo 7B public, including all the steps and details of its development. This openness means that AI researchers can see exactly how Olmo 7B works, which can help them make better AI tools and solve problems with current ones. Sophie Labrecht, the COO of AI2, says that to understand and improve these AI models, researchers need full access to all the data. This way they can figure out why and how the AI makes its decisions. Researchers have been trying to link specific AI responses to the data it was trained on and this openness might help. One problem right now is that many AI researchers can only use big AI models from companies like OpenAI or Google, which are expensive to train and use. These researchers can't see the inner workings of these models, so they don't fully understand the results they get. Hajisherzi compares this situation to an astronomer trying to study the solar system with only newspaper pictures. Even open source models from companies like Meta aren't fully open. They share the model but not the training data or all the details. Olmo is a medium sized model with 7 billion parameters and was trained on 2 trillion tokens. Librecht points out that keeping AI research secret means other researchers often try to redo the same work we which wastes energy and has environmental impacts. By sharing their work, AI2 hopes to reduce the carbon footprint of AI research and avoid duplicated efforts. Meta's chief AI scientist Yann LeCun supports open sourcing new AI models. He believes that an open source community is the best way to advance AI. Haji Shirzi agrees, but notes that even meta's open source models don't share everything like the training data and all the code details. In short, AI2's move to make Olmo 7B completely open is a big step towards more collaboration and progress in AI research. And it could also help the environment by reducing duplicated work. Okay, lastly, let's talk about Ark, a tech company that is creating something really cool. An AI that can search the web for you. Normally we use Google or other search engines to find stuff online, but Ark wants to change that with their new tool. They're making a browser called the Ark browser that does the searching for you. Soon they'll release a tool where you just tell the browser what you're looking for and it finds all the info for you by searching the web itself. Imagine telling it you need a table reservation at a restaurant and it finds all the available times for you. You can then book a table with just one click. Ark has already started making some of these cool features. They've got this new iPhone app called Ark Search. It has a feature called Browse for Me where the app reads several web pages about what you're looking for and then makes a new page with all the important info, including photos and videos. They've also got this new thing called Instant Links. Say you're looking for the gladiator. Two Trailer Curtains fall on Maximus Epic. Instead of showing a bunch of search results, it takes you straight to the trailer on YouTube. And if you want reviews for a product like the Apple Vision Pro, it makes a folder with all the reviews from different websites. Later, they'll add a live folder feature. It's like a regular folder, but it updates automatically when new stuff gets posted online. For example, if there's a new blog post on a topic you're interested in, the folder updates with it. It's a bit like keeping up with news feeds and website changes. But we're not sure if it can track every little change on a webpage. The idea behind Arc Browser is to change how we use the web. They want to make it faster to find what you're looking for without going through a bunch of search results. They're using AI in a different way, not just for writing stuff, but for things like renaming tabs and showing previews links. One big question is how this AI powered browser will affect the websites it gets info from. And how will it decide what's the best result for each person since everyone likes different things. With companies like OpenAI and Perplexity and even big search engines like Google, Microsoft and DuckDuckGo using AI for searching, Ark thinks it's the perfect time to offer a new way to search on the web. They promise not to sell user data and they're thinking about making special features for teams, but they haven't shared much about that yet. Alright, that wraps up our video. If you liked it, please consider subscribing and sharing so we can keep bringing more content like this. Thanks for watching and see you in the next one.