Unlocking Innovation: How OpenAI’s New ‘Deep Research’ Tool is Transforming AI Landscape

0
22
ETtech Explainer: How OpenAI is moving the needle with new 'deep research' tool

OpenAI Enhances ChatGPT with Deep Research Capabilities and New Model Release

On Sunday, OpenAI announced significant upgrades to ChatGPT, introducing ‘deep research’ agentic capabilities aimed at revolutionizing the user experience. Additionally, the company launched o3-mini, a new modeling architecture touted as the most cost-efficient reasoning agent yet. These updates follow the buzz generated by the new Chinese competitor, DeepSeek.

Understanding ‘Deep Research’ Capability in ChatGPT

With the introduction of ‘deep research’, ChatGPT can now execute intricate, multi-step research tasks by sifting through extensive online resources. According to OpenAI, “Give it a prompt and ChatGPT will find, analyze, and synthesize hundreds of online sources to create a comprehensive report in mere minutes, an endeavor that would take a human many hours.”

Early testers have utilized the feature for diverse applications, including literary analysis and extensive research reports that sometimes exceed 10,000 words. One user described it as a glimpse into “the future of human-AI collaboration in knowledge work,” while another likened it to having an almost-PhD-level researcher at one’s disposal.

Despite its impressive capabilities, some users reported drawbacks, such as a lack of consistent source citations and the challenge of restarting the model when it continues to respond after a query has been completed.

The ‘deep research’ feature is currently accessible to Pro users, with Plus and Team users set to gain access soon.

What is the o3-mini Model and How Does It Differ?

OpenAI’s recently introduced o3-mini model offers a specialized alternative to its earlier reasoning model, o1, which focused on general knowledge. The new model is designed for technical domains, providing faster and more efficient performance.

Available as an API for developers and integrated into ChatGPT for Plus, Team, and Pro users, the o3-mini is also accessible for free users to experiment with. Developers can choose from varying reasoning effort tiers, including low, medium, and high, allowing for tailored application based on project requirements.

Cost and Accuracy Comparison

OpenAI claims that the o3-mini model marks a significant reduction in intelligence costs—now 95% lower than GPT-4—without sacrificing quality. In a cost comparison, the o3-mini model is priced at $1.10 per million input tokens, significantly less than its predecessor o1, which cost $15 per million. Bulk API requests can reduce the cost further to $0.55 per million tokens, which aligns with the pricing of DeepSeek’s R1 model.

When evaluated for accuracy using the ‘Humanity’s Last Exam’ benchmark, OpenAI’s deep research model achieved a notable score of 26%, whereas DeepSeekR1 received a score of 9.4%. While DeepSeekR1 outperformed o1, scoring 9.1%, it still fell short of the o3-mini’s scores of 10.5% and 13% in its medium and high versions, respectively.

It’s worth noting that unlike DeepSeek, which offers open-source options for developers, OpenAI’s models are not open-source, limiting users in terms of customization.

Impact of DeepSeek on the Market

The recent updates from OpenAI were largely motivated by the launch of the open-source Chinese LLM, DeepSeekR1, which stirred the competitive landscape by operating on considerably less computational power.

OpenAI CEO Sam Altman acknowledged the significance of having a new competitor, stating, “We will obviously deliver much better models, and it’s legitimately invigorating to have a new player in the field! We will ramp up our releases.”

This sentiment indicates a burgeoning competition focused on producing more affordable, efficient AI models, contributing to the global AI race. With rising interest in alternative solutions, Microsoft, a key partner of OpenAI, has even integrated DeepSeekR1 into its cloud platform Azure and code-hosting platform GitHub, alongside other tech giants like Nvidia and AWS.

As companies like Indian startup Krutrim host the new Chinese model on their AI platforms, the dynamics of AI competition continue to evolve rapidly.

Conclusion

The recent advancements made by OpenAI with the deep research capability and the launch of the o3-mini model underscore its commitment to remain at the forefront of AI technology. As they navigate an increasingly competitive landscape prompted by innovations from new entrants like DeepSeek, OpenAI’s offerings are poised to attract attention for both their affordability and efficiency.

Q&A

  1. What is the primary feature introduced with ChatGPT?
    Deep research capabilities allow ChatGPT to conduct complex online research, synthesizing multiple sources quickly.
  2. How does the o3-mini model differ from the earlier o1 model?
    The o3-mini is specialized for technical fields, offering faster performance and more efficient processing compared to the broader o1 model.
  3. What is the cost per million tokens for the o3-mini model?
    The o3-mini model costs $1.10 per million input tokens, which can drop to $0.55 with batch API requests.
  4. How did OpenAI’s deep research model perform on the ‘Humanity’s Last Exam’ benchmark?
    It achieved a score of 26%, which is significantly higher than the scores of both DeepSeekR1 and the o1 model.
  5. Is DeepSeek’s model open-source?
    Yes, unlike OpenAI’s models, DeepSeek offers an open-source solution for developers.

source