Jarretera/Shutterstock
There have been few standout platforms in the race for ultimate AI supremacy. OpenAI’s GPT-3 (the artificial brain behind ChatGPT and Bing Chat) has been the basis for a great deal of the web apps and services powered by large language models, owed mostly due to its substantial backing, a massive dataset, and open source accessibility.
The tech is maturing faster than any of us could have imagined, meaning we’ll see a startling explosion of new AI-based tools in 2023 and beyond, as well as sweeping improvements to existing platforms. Search giants stand to gain the most out of the deal. As smart as these search engines already are, AI will make them much more effective. The aforementioned Bing Chat is already tightening its screws in preview status, and Google is hard at work behind the scenes on Bard.
But Brave says it wants in. The company’s browser and search engine aim to offer everything the incumbent stalwarts don’t — faster speed, more security, and a truly private browsing experience. And with the launch of Braze Summarizer, a feature that understands natural language questions and produces easily digestible answers, it doesn’t plan to be far behind in the AI department.
Brave Summarizer extracts the information you care about
In a post filled with cheeky shots fired at the likes of ChatGPT and Bing Chat, Brave says its AI-powered Summarizer isn’t prone to outdated information, spreading misinformation, and flat-out lying. That’s because it gets its information purely from web results, and you’ll always be able to trace its claims back to the source thanks to ample citation. In practice, it’s not unlike Bing Chat’s core functionality today.
If you aren’t satisfied with the answer Summarizer gives or you’re skeptical of the reliability of the information and sources it landed on, the search engine also uses its AI capabilities to highlight the relevant snippets of information in the standard search results’ descriptions.
“Unlike many others that have released similar features recently, we do not rely on third parties, nor do we limit access due to scalability concerns,” writes Brave in a blog post announcing the feature. “The Brave Summarizer relies on our owned and operated models that are highly tuned to be as efficient as possible at inference time.” Still, Brave admits that its large language model is still in its infancy, and it can be prone to some of the same pitfalls ailing other generative systems, such as “hallucinations” (AI making things up) and using offensive text in the resulting query responses.
Read More
Tami Mischke