Introducing Answer with AI, Brave’s New AI Assistant
Digest
This episode of the Brave Technologies podcast features Josep, the lead of Brave Search, and Remi, the principal engineer at Brave, discussing the new Answer with AI feature launched by Brave Search. They delve into how Answer with AI differs from other chatbots like ChatGPT, its implications for traditional search, Brave's privacy policy regarding the engine, how they handle hallucinations, and the challenges encountered during its development. Josep emphasizes that Answer with AI is not meant to replace traditional search but rather to augment it, serving as a complementary tool for specific types of queries. He highlights Brave's commitment to providing a comprehensive search experience, encompassing both traditional search results and AI-powered answers, all while prioritizing user privacy. Remi discusses the technical aspects of Answer with AI, including its ability to handle a significant volume of requests per second and the automatic triggering mechanism that determines when AI-generated answers are most appropriate. He also addresses concerns about hallucinations, explaining how Brave's approach, grounded in the search index, helps mitigate this issue. The episode concludes with a discussion about the future of AI in search, with Josep and Remi hinting at potential integrations with Brave's in-browser AI, Leo, to provide personalized and contextually relevant answers.
Outlines
Introducing Answer with AI
This Chapter introduces the new Answer with AI feature launched by Brave Search and welcomes Josep, the lead of Brave Search, and Remi, the principal engineer at Brave, as guests on the podcast. They discuss the purpose and functionality of Answer with AI, highlighting its unique approach compared to other chatbots like ChatGPT.
Answer with AI: Evolution, Not Replacement
This Chapter explores the relationship between Answer with AI and traditional search engines. Josep argues that Answer with AI is not intended to replace traditional search but rather to evolve it, serving as a complementary tool for specific types of queries. He explains that not all queries are suitable for AI-generated answers and that traditional search remains essential for tasks like navigating websites or finding specific information.
Brave's Strategy: A One-Stop Shop
This Chapter delves into Brave's strategic approach to search, emphasizing their commitment to providing a comprehensive search experience that encompasses both traditional search results and AI-powered answers. Josep explains that Brave aims to be a one-stop shop for users, offering a unified system that caters to diverse search needs and avoids the need for users to switch between different platforms.
Answer with AI: Unique Benefits and User Experience
This Chapter highlights the unique benefits of Brave's Answer with AI feature, emphasizing its ability to provide users with the best possible answers and a seamless search experience. Josep discusses how Brave's approach to AI integration differs from competitors, particularly in terms of user privacy and the integration of AI-generated answers within the search results page.
Privacy and Answer with AI
This Chapter addresses concerns about user privacy in relation to Answer with AI. Josep explains that Brave treats AI queries the same way as traditional search queries, meaning they do not collect or store user history or personal identifiable information. He emphasizes Brave's commitment to privacy as a core principle and highlights the ethical implications of data collection and potential government access to user data.
Brave's Independence: A Key Differentiator
This Chapter explores the significance of Brave's independence in the search landscape. Josep and Remi discuss the challenges of building a truly independent search engine and the benefits it offers in terms of privacy, control, and flexibility. They highlight the limited number of independent search engines with their own indexes and the potential for censorship or limitations imposed by relying on third-party APIs.
Traffic and User Feedback
This Chapter provides insights into the traffic and user feedback received for Answer with AI. Remi shares data on the number of requests per second and concurrent generation sessions, highlighting the growing adoption of the feature. He also discusses the automatic triggering mechanism that determines when AI-generated answers are most appropriate and the user feedback received, including both positive and negative responses.
Addressing Hallucinations
This Chapter addresses concerns about hallucinations, a common issue with LLMs. Remi explains how Brave's approach, grounded in the search index, helps mitigate this issue by providing relevant information to the model and limiting its scope. He emphasizes the importance of critical thinking and fact-checking, even with AI-generated answers, and acknowledges the ongoing challenges of ensuring accuracy and reliability.
Keywords
Brave Search
Brave Search is a privacy-focused search engine developed by Brave Software. It aims to provide users with a more private and secure search experience compared to traditional search engines like Google and Bing. Brave Search uses its own index of websites and employs various techniques to protect user privacy, such as not tracking user search history or collecting personal identifiable information.
Answer with AI
Answer with AI is a new feature launched by Brave Search that leverages artificial intelligence to provide users with concise and informative answers to their queries. It differs from traditional chatbots like ChatGPT by being grounded in the search index, meaning it uses relevant information from the web to generate answers. Answer with AI aims to provide users with the best possible answers, while also offering traditional search results for more comprehensive exploration.
ChatGPT
ChatGPT is a large language model chatbot developed by OpenAI. It is known for its ability to engage in natural-sounding conversations and generate human-like text. ChatGPT is trained on a massive dataset of text and code, enabling it to respond to a wide range of prompts and questions. However, it is important to note that ChatGPT can sometimes generate inaccurate or misleading information, known as hallucinations.
Hallucinations
Hallucinations in the context of AI refer to instances where a language model generates incorrect or fabricated information. This can occur due to the model's training data or its inherent limitations in understanding and interpreting complex information. Hallucinations can be a significant issue, particularly when AI is used for tasks that require accuracy and reliability.
Leo
Leo is Brave's in-browser AI, designed to provide users with a more personalized and contextually relevant AI experience. It is intended to complement Brave Search by offering a more interactive and in-depth AI interaction. Leo is still under development, but it has the potential to revolutionize how users interact with information online.
Privacy
Privacy is a fundamental concern in the digital age, particularly with the increasing use of AI and data collection. Brave prioritizes user privacy and aims to provide a more secure and private online experience. Brave Search does not track user search history or collect personal identifiable information, ensuring that user data remains protected.
Independent Search Engine
An independent search engine is a search engine that operates independently of major tech companies like Google and Microsoft. It uses its own index of websites and does not rely on third-party APIs. Independent search engines offer users more control over their data and provide an alternative to the dominant search engines.
LLMs (Large Language Models)
LLMs are a type of artificial intelligence that are trained on massive datasets of text and code. They are capable of generating human-like text, translating languages, writing different kinds of creative content, and answering your questions in an informative way. LLMs are a rapidly evolving field with significant potential for various applications.
Q&A
How does Brave Search's Answer with AI differ from other chatbots like ChatGPT?
Answer with AI is grounded in the search index, meaning it uses relevant information from the web to generate answers, unlike ChatGPT which relies on its pre-training data. This makes Answer with AI more contextually relevant and less prone to hallucinations.
Is Answer with AI intended to replace traditional search engines?
No, Answer with AI is meant to augment traditional search, serving as a complementary tool for specific types of queries. Traditional search remains essential for tasks like navigating websites or finding specific information.
What is Brave's approach to user privacy with Answer with AI?
Brave treats AI queries the same way as traditional search queries, meaning they do not collect or store user history or personal identifiable information. They prioritize user privacy and believe that it should not be sacrificed for the sake of AI functionality.
What are some of the challenges faced in building Answer with AI?
Challenges include ensuring performance at scale, given the large language models involved, and maintaining high quality answers, which requires a combination of automated testing, human assessment, and continuous improvement. The need to extract relevant information from a vast index and present it in a concise and informative way also poses a significant challenge.
What are some potential future developments for Answer with AI?
Brave is exploring integrations with Leo, their in-browser AI, to provide personalized and contextually relevant answers. This could involve leveraging local information stored in the browser to enhance search results and provide a more tailored experience.
How can users provide feedback on the quality of Answer with AI?
Users can provide feedback through a feedback button on every search result page. They can also report offensive or dangerous content through community forums or social media platforms. Brave also has a dedicated email address for users to submit requests related to the right to be forgotten.
What is the significance of Brave's independence in the search landscape?
Brave's independence allows them to prioritize user privacy and control over their data. It also gives them the flexibility to innovate and develop new features without being constrained by the terms of service or limitations imposed by third-party APIs.
How does Brave handle hallucinations in Answer with AI?
Brave's approach, grounded in the search index, helps mitigate hallucinations by providing relevant information to the model and limiting its scope. They also emphasize the importance of critical thinking and fact-checking, even with AI-generated answers.
What is the current traffic and user feedback for Answer with AI?
Answer with AI is handling around 100 requests per second, representing about a third of Brave Search queries. User feedback has been positive, with many users finding the feature valuable and helpful. However, some users have expressed concerns about the AI's presence and have requested the ability to disable it.
Show Notes
Josep M. Pujol (Chief of Search at Brave) and Rémi Berson (Principal Engineer at Brave) discuss Brave Search’s latest feature, Answer With AI, which is the first privacy-preserving answer engine available at scale. They explore what this new assistant’s capabilities mean for traditional search, and its potential to reshape our search habits.
Key Takeaways:
-
How Answer with AI is different from chatbots
-
The challenges of AI hallucinations and ensuring accurate responses
-
Balancing personalized search experiences with user privacy and anonymity
-
The plan for integrating Answer with AI with Brave’s browser-based Leo AI experience
Guest Bio: Rémi has been working with AI since 2014, beginning with a Paris-based start-up that offered custom NLP solutions to various media groups, and pioneered large-scale automatic text generation (before transformers made an appearance). Currently, Remi is a senior software engineer at Brave Search, where he has been involved in scaling the ML infrastructure to serve continuously growing traffic and increasingly advanced needs, as well as in the development of Brave Search’s latest feature, Answer with AI.
Josep is the Chief of Search at Brave, which he joined in 2021. Josep first started working on building an alternative search engine in 2014; prior to Brave, he served as chief scientist at Cliqz, and led the Tailcat project (which Brave later acquired). Before being involved with search, he was a research scientist at multiple institutions, working on AI, distributed systems, and complex networks. He has authored over 30 papers and holds 4 patents. He also holds a PhD in AI from the Technical University of Catalonia.
----------------------------------------------------------------------------------------
About this Show:
The Brave Technologist is here to shed light on the opportunities and challenges of emerging tech. To make it digestible, less scary, and more approachable for all!
Join us as we embark on a mission to demystify artificial intelligence, challenge the status quo, and empower everyday people to embrace the digital revolution. Whether you're a tech enthusiast, a curious mind, or an industry professional, this podcast invites you to join the conversation and explore the future of AI together.
The Brave Technologist Podcast is hosted by Luke Mulks, VP Business Operations at Brave Software—makers of the privacy-respecting Brave browser and Search engine, and now powering AI everywhere with the Brave Search API.
Music by: Ari Dvorin
Produced by: Sam Laliberte
Table of contents
United States
