On February 2, during a live announcement, OpenAI introduced a new AI agent named "Deep Research," which aims to revolutionize how users engage with ChatGPT for in-depth investigationsThe purpose of this innovative feature is to facilitate complex research tasks across various professional domains, including finance, science, policy, and engineeringThis powerful tool promises to provide comprehensive, accurate, and reliable research assistance, all while extending its usefulness to individuals making significant consumer decisions, such as purchasing automobiles, appliances, or furniture.

Deep Research is specifically designed for users who seek more than just quick answers or summariesIt caters to those who require an extensive analysis that considers multiple websites and diverse information sources, helping users dig deeper until they uncover the most suitable and pertinent answers

Advertisements

Sam Altman, the CEO of OpenAI, expressed enthusiasm for the new capabilities, stating, "It can scour the internet, perform complex research and reasoning, and then compile a report for youThis is an exceptional tool meant to execute tasks that typically take hours or even days and can cost hundreds of dollars." He further remarked on its impressive ability, despite its computational demands and processing speed, making it the first AI system capable of handling such a wide array of complex and valuable tasks.

The announcement ignited excitement among users eager to test this new featureSome were already prepared to implement it in their workflows, while others sought early access by sharing their accountsHowever, curiosity also sparked questions about the effectiveness of ChatGPT's Deep Research compared to similar capabilities offered by GoogleConcerns about privacy and specific functionalities, such as the ability to capitalize the first letter of each sentence in the generated reports, were also raised.

Currently, the Deep Research feature is only accessible to subscribers of ChatGPT Pro, with a cap of 100 queries per month

Advertisements

Altman indicated that this function would eventually roll out to Plus and Team subscribers, with further expansion toward enterprise-level users anticipated in the futureThe Plus subscription is expected to launch in about a month and will see significantly increased query limits for paid usersHowever, many users commented that the $200 monthly fee for the Pro subscription seemed exorbitant, asking rhetorically, "Who cares if it's not free?" Others speculated on the potential future pricing for AGI, joking about a possible $2000 monthly fee with only ten allowed queries.

Interestingly, Altman mentioned that free users would have limited access to the feature, though specifics on the extent of that accessibility remain undisclosedIt's worth noting that the Deep Research functionality has only been released in selected regions, and as of now, OpenAI has not announced when it will be available for users in the UK, Switzerland, or the European Economic Area

Advertisements

Furthermore, any output from Deep Research currently exists exclusively in text formatOpenAI plans to incorporate embedded images, data visualizations, and other analytical formats in the futureThey also intend to support connections to additional specialized data sources, including subscription-based databases and internal resources.

A pertinent question looms over the introduction of this feature: Will analysts face widespread unemployment as a result? The key to this concern hinges on whether the accuracy of ChatGPT's Deep Research can meet the demands of professional investigationsAfter all, artificial intelligence is not infallible; it has been known to produce erroneous or misleading information, a phenomenon commonly referred to as "hallucination." In light of this, OpenAI has committed to providing complete documentation of each Deep Research query, including clear citations and reasoning summaries, to facilitate user reference and information verification.

However, skepticism persists regarding whether such measures will adequately prevent AI from misleading users

• Gold Prices Surge Dramatically!
• A-Share Surge Exceeds 2500%!
• AI Milestone Achieved
• Japan's Chip Production Hits Snags
• Positive Outlook for U.S. Manufacturing

Previous reports have indicated that AI search functions within ChatGPT have returned inaccurate answers, with some evaluations revealing that ChatGPT Search did not perform as reliably as Google in certain inquiriesIn recognition of this challenge, OpenAI has introduced a specialized version of its o3 "reasoning" AI model to enhance depth research precisionThis model has undergone reinforcement learning training and is specifically designed for real-world tasks that necessitate the use of a browser and Python tools.

OpenAI claims that the o3 model is optimized for web browsing and data analysis, adept at searching, interpreting, and analyzing vast quantities of text, images, and PDFs from the internet while flexibly adjusting its strategies based on new information it discoversFurthermore, this advanced model can examine user-uploaded files, create and iteratively refine charts, and incorporate generated visuals and images from web sources into responses, citing specific sentences or paragraphs from the sources consulted.

To gauge the effectiveness of their model, OpenAI employed a rigorous assessment method termed the "Ultimate Human Exam," comprising over 3,000 challenging academic questions—outpacing other AI evaluation benchmarks to ensure its difficulty remains ahead of the model's progression