NEWSLETTER
“Perform a comprehensive review of the literature on the state of art in automatic learning and energy consumption. (…) “
With this notice, I tried the new deep research function, which has been integrated into the Operai O3 reasoning model since the end of February, and I did a avant -garde literature review in 6 minutes.
This function goes beyond a normal web search (for example, with chatgpt 4): the research consultation is broken down and structure, information is sought on the Internet, which is then evaluated, and finally, a structured and complete report is created.
Let's take a close look at this.
table of Contents
1. What is Operai's deep research and what can you do with him?
2. How does deep research work?
3. How can you use deep research? – Practical example
4. Challenges and risks of the deep research function
Final thoughts
Where can you continue learning?
1. What is Operai's deep research and what can you do with him?
If you have an Operai Plus account (the $ 20 per month plan), you have access to deep investigation. This gives access to 10 consultations per month. With the Pro Subscription ($ 200 per month), it has extended access to deep research and access to the previous view of GPT-4.5 research with 120 consultations per month.
Opadai He promises that we can carry out several steps investigations using public web data.
Duration: 5 to 30 minutes, depending on complexity.
Previously, this investigation generally took hours.
It is intended for complex tasks that require deep search and thoroughness.
How are concrete use cases?
- Make a literature review: carry out a review of literature on automatic learning and state -of -the -art energy consumption.
- MARKET ANALYSIS: Create a comparative report on the best marketing automation platforms for companies in 2025 based on current market trends and evaluations.
- technology and Software Development: Investigate Languages and Programming Mark for the Development of ai applications with performance analysis and use cases
- Investment and financial analysis: conduct research on the impact of ai negotiation on the financial market based on recent reports and academic studies.
- Legal research: Make an overview of data protection laws in Europe compared to the US, including relevant decisions and recent changes.
2. How does deep research work?
Deep Research uses several deep learning methods to carry out a systematic and detailed analysis of the information. The whole process can be divided into four main phases:
1. Decomposition and structuring of the research question
In the first step, the tool processes the research question using natural language processing methods (NLP). Identify the most important key concepts, concepts and subcues.
This step ensures that ai understands the question not only literally, but also in terms of content.
2. Obtain relevant information
Once the tool has structured the research question, it specifically seeks information. Deep Research uses a mixture of internal databases, scientific publications, API and web scraping. These can be open access databases such as Arxiv, Pubmed or Semantic Scholar, for example, but also public websites or news sites such as The Guardian, New York Times or BBC. In the end, any content that can be accessed online and is publicly available.
3. Analysis and interpretation of the data
The next step is that the ai model summarizes large amounts of text in compact and understandable responses. The transformers and care mechanisms ensure that the most important information is prioritized. This means that not only creates a summary of all the content found. In addition, the quality and credibility of the sources is evaluated. And cross validation methods are normally used to identify incorrect or contradictory information. Here, the ai tool compares several sources from each other. However, it is not known exactly how this is done in deep investigations or what criteria there is for this.
4. GENERATION OF THE FINAL REPORT
Finally, the final report is generated and shown. This is done using natural language generation (NLG) so that we can see easily legible texts.
The ai system generates diagrams or tables if requested in the application and adapts the user's style response. The main sources used are also listed at the end of the report.
3. How can use deep research: a practical example
In the first step, it is better to use one of the standard models to ask how to optimize the message to conduct deep research. I have done this with the following message with chatgpt 4:
“Optimize this notice to carry out a deep investigation:
Performing a search for literature: perform a search for literature on the state of art on automatic learning and energy consumption. “
The 4O model suggested the following notice for the deep research function:
Then, the tool asked me if I could clarify the scope and approach of literature review. Therefore, I have provided some additional specifications:
Chatgpt then returned the clarification and began the investigation.
Meanwhile, I could see progress and how more sources were added gradually.
After 6 minutes, the review of the avant -garde literature was complete and the report, including all sources, was available to me.
Deep Research Example.mp4
4. Challenges and risks of the deep research function
Let's take a look at two research definitions:
“A detailed study of a subject, especially to discover new information or reach a new understanding.”
“Research is a creative and systematic work carried out to increase the stock of knowledge. It implies the collection, organization and analysis of evidence to increase the understanding of a topic, characterized by particular attention to bias and error control sources. “
The two definitions show that research is a detailed and systematic investigation of a topic, with the aim of discovering new information or achieving a deeper understanding.
Basically, the deep research function meets these definitions to some extent: it collects existing information, analyzes it and presents it in a structured way.
However, I think we must also be aware of some challenges and risks:
- Danger of superficiality: Deep research is mainly designed to search, summarize and provide existing information in a structured way (at least in the current stage). Absolutely excellent for general research. But what about digging deeper? Real scientific research goes beyond mere reproduction and analyzes criticism of sources. Science also thrives in generating new knowledge.
- Reinforcement of existing biases in research and publication: Articles are more likely to be published if they have significant results. The “non -significant” or contradictory results, on the other hand, are less likely to be published. This is known to us as publication bias. If the ai tool now mainly evaluates the documents cited frequently, reinforces this trend. Rare or less extended but possibly important findings are lost. A possible solution here would be to implement a mechanism for the evaluation of weighted sources that also take into account the less cited but relevant documents. If the methods of ai mainly cite sources that are cited frequently, less extended but important findings can be lost. Presumably, this effect also applies to us humans.
- Quality of research work: While it is obvious that a bachelor's, mastery or doctoral thesis cannot be based solely on the research generated by ai, the question I have is how universities or scientific institutions deal with this development. Students can obtain a solid research report with a single message. Presumably, the solution here must be adapted the evaluation criteria to give greater weight to reflection and in -depth methodology.
Final thoughts
In addition to OpenAI, other companies and platforms have also integrated similar functions (even before OpenAI): for example, <a target="_blank" href="https://www.perplexity.ai/de/hub/blog/introducing-perplexity-deep-research” rel=”noreferrer noopener” target=”_blank”>Perplexity ai It has introduced a deep research function that performs and analyzes searches. Also Gemini by Google He has integrated such a deep research function.
The function gives you an incredibly rapid general vision of an initial research question. It remains to be seen how reliable the results are. Currently (as of March 2025), Operai himself writes as limitations The fact that the characteristic is still at an early stage, can sometimes hallucinate facts in answers or draw false conclusions, and has problems distinguishing authorized information from rumors. In addition, uncertainties cannot be proven precisely.
But it can be assumed that this function will be extended even more and will become a powerful tool for research. If you have simpler questions, it is better to use the standard GPT-4o model (with or without search), where you get an immediate answer.
Where can you continue learning?
Do you want more tips and tricks on technology, Python, Data Science, Data Engineering, Automatic Learning and ai? Then, regularly receive a summary of my most read items in my replacement, curator and for free.
(Tagstotranslate) Deep Lear
