With Google in mind, OpenAI introduces SearchGPT.

With Google in mind, OpenAI introduces SearchGPT.

According to the company's website for the search AI version, OpenAI has finally entered the AI search space with the new SearchGPT prototype model, which will assist you in finding answers from reliable sources.

https://openai.com/index/searchgpt-prototype/

The goal of SearchGPT is to bring together the ChatGPT conversational experience with answers from the internet and other platforms. OpenAI describes the function of its AI search model as follows: "SearchGPT will quickly and directly respond to your questions with up-to-date information from the web while giving you clear links to relevant sources."

https://chatgpt.com/search

OpenAI appears to be stepping into the search arena. This move aims to challenge various players, including newcomers like Perplexity, as well as established names like Google & Bing.

On Thursday, the company introduced SearchGPT, a new search feature meant to deliver “timely answers” derived from web sources.

When it comes to its user interface, SearchGPT resembles OpenAI's chatbot platform ChatGPT. You simply enter a question, and SearchGPT provides information, images from the web, and links to pertinent sources. Shortly after, you can ask follow-up questions or delve into additional related searches found in a sidebar.

Some searches also consider your location. A support document reveals that SearchGPT “collects and shares” general location data with third-party search providers to enhance the accuracy of results. For instance, it can display lists of nearby restaurants or provide weather forecasts. Through a toggle in the settings menu, users have the option to share more specific location details.

SearchGPT operates using OpenAI’s models, namely GPT-3.5, GPT-4, and GPT-4o. Launching today for “a small group” of users and publishers (there’s a waitlist here), OpenAI labels SearchGPT as a prototype. Future updates will likely see certain features integrated into ChatGPT as well.


“Getting answers online can often take quite a bit of effort,” notes OpenAI in a blog post. “It usually requires several tries to find relevant information. We think that by improving our models’ conversational abilities with actual web data, finding what you need can become faster & easier.”

Speculation about OpenAI creating something akin to a Google competitor has circulated for some time. In February, The Information reported that such a product—or at least its pilot—was being developed. However, launching SearchGPT now seems poorly timed because AI-driven search tools face valid criticism regarding plagiarism & inaccuracies.

Google’s AI-powered search feature, called AI Overviews, notoriously suggested applying glue on pizza. Arc Search from The Browser Company bizarrely claimed that cut-off toes would regenerate. Moreover, Genspark's AI once recommended lethal weapons rather freely. Additionally, Perplexity copied news articles from outlets like CNBC and Bloomberg without proper credit.

These AI-generated summaries pose a risk of diverting traffic from the original content creators—this has already been documented since one study indicated that about 25% of publisher traffic could suffer due to reduced emphasis on article links.

In light of these issues—copying and lack of attribution—OpenAI aims to position SearchGPT as a more responsible option compared to others in this space.

According to OpenAI, SearchGPT “prominently cites & links” to publishers while ensuring “clear, in-line, named attribution.” Additionally, they are collaborating with publishers to design the user experience and give website owners control over how their material appears in search outcomes.

“Importantly,” OpenAI states in its blog post, “SearchGPT focuses on search; it is distinct from training our generative AI foundation models.” This means sites can appear in search results even if they opt out of generative training. They express commitment to maintaining a healthy ecosystem for creators & publishers.

However—and it's important to note—it can be difficult to trust a company that previously scraped millions of YouTube transcripts without permission for model training purposes. Still, we shall observe how the story of SearchGPT develops moving forward.

Enjoy Friends.


要查看或添加评论,请登录

社区洞察

其他会员也浏览了