Viewpoint: OpenAI’s new ‘deep analysis’ agent remains to be only a fallible software—not a human-level professional

February 12, 2025

The GIST Editors' notes

This text has been reviewed in line with Science X's editorial course of and insurance policies. Editors have highlighted the next attributes whereas guaranteeing the content material's credibility:

fact-checked

trusted supply

written by researcher(s)

proofread

Viewpoint: OpenAI's new 'deep analysis' agent remains to be only a fallible software—not a human-level professional

OpenAI's "deep analysis" is the most recent synthetic intelligence (AI) software making waves and promising to do in minutes what would take hours for a human professional to finish.

Bundled as a characteristic in ChatGPT Professional and marketed as a analysis assistant that may match a skilled analyst, it autonomously searches the net, compiles sources and delivers structured studies. It even scored 26.6% on Humanity's Final Examination (HLE), a troublesome AI benchmark, outperforming many fashions.

However deep analysis doesn't fairly reside as much as the hype. Whereas it produces polished studies, it additionally has severe flaws. In accordance with journalists who've tried it, deep analysis can miss key particulars, battle with latest info and typically invents information.

OpenAI flags this when itemizing the restrictions of its software. The corporate additionally says it "can typically hallucinate information in responses or make incorrect inferences, although at a notably decrease price than present ChatGPT fashions, in line with inside evaluations."

It's no shock that unreliable knowledge can slip in, since AI fashions don't "know" issues in the identical approach people do.

The concept of an AI "analysis analyst" additionally raises a slew of questions. Can a machine—regardless of how highly effective—actually exchange a skilled professional? What can be the implications for data work? And is AI actually serving to us suppose higher, or simply making it simpler to cease considering altogether?

What’s 'deep analysis' and who’s it for?

Marketed to professionals in finance, science, coverage, legislation and engineering, in addition to lecturers, journalists and enterprise strategists, deep analysis is the most recent "agentic expertise" OpenAI has rolled out in ChatGPT. It guarantees to do the heavy lifting of analysis in minutes.

Presently, deep analysis is simply out there to ChatGPT Professional customers in america, at a value of US$200 per 30 days. OpenAI says it’s going to roll out to Plus, Workforce and Enterprise customers within the coming months, with a less expensive model deliberate for the longer term.

In contrast to an ordinary chatbot that gives fast responses, deep analysis follows a multi-step course of to supply a structured report:

1. The consumer submits a request. This may very well be something from a market evaluation to a authorized case abstract.
2. The AI clarifies the duty. It could ask follow-up inquiries to refine the analysis scope.
3. The agent searches the net. It autonomously browses lots of of sources, together with information articles, analysis papers and on-line databases.
4. It synthesizes its findings. The AI extracts key factors, organizes them right into a structured report and cites its sources.
5. The ultimate report is delivered. Inside 5 to half-hour, the consumer receives a multi-page doc—doubtlessly even a Ph.D.-level thesis—summarizing the findings.

At first look, it seems like a dream software for data staff. A better look reveals important limitations.

Many early assessments have uncovered shortcomings:

• It lacks context. AI can summarize, but it surely doesn't totally perceive what's essential.
• It ignores new developments. It has missed main authorized rulings and scientific updates.
• It makes issues up. Like different AI fashions, it could possibly confidently generate false info.
• It could't inform truth from fiction. It doesn't distinguish authoritative sources from unreliable ones.

Whereas OpenAI claims its software rivals human analysts, AI inevitably lacks the judgment, scrutiny and experience that make good analysis helpful.

What AI can't exchange

ChatGPT isn't the one AI software that may scour the net and produce studies with only a few prompts. Notably, a mere 24 hours after OpenAI's launch, Hugging Face launched a free, open-source model that almost matches its efficiency.

The largest threat of deep analysis and different AI instruments marketed for "human-level" analysis is the phantasm that AI can exchange human considering. AI can summarize info, however it could possibly't query its personal assumptions, spotlight data gaps, suppose creatively or perceive totally different views.

And AI-generated summaries don't match the depth of a talented human researcher.

Any AI agent, regardless of how briskly, remains to be only a software, not a alternative for human intelligence. For data staff, it's extra essential than ever to put money into abilities that AI can't replicate: crucial considering, fact-checking, deep experience and creativity.

If you happen to do wish to use AI analysis instruments, there are methods to take action responsibly. Considerate use of AI can improve analysis with out sacrificing accuracy or depth. You would possibly use AI for effectivity, like summarizing paperwork, however retain human judgment for making selections.

At all times confirm sources, as AI-generated citations will be deceptive. Don't belief conclusions blindly, however apply crucial considering and cross-check info with respected sources. For prime-stakes matters—resembling well being, justice and democracy—complement AI findings with professional enter.
c
Regardless of prolific advertising and marketing that tries to inform us in any other case, generative AI nonetheless has loads of limitations. People who can creatively synthesize info, problem assumptions and suppose critically will stay in demand—AI can't exchange them simply but.

Offered by The Dialog

This text is republished from The Dialog below a Inventive Commons license. Learn the unique article.

Quotation: Viewpoint: OpenAI's new 'deep analysis' agent remains to be only a fallible software—not a human-level professional (2025, February 12) retrieved 12 February 2025 from https://techxplore.com/information/2025-02-viewpoint-openai-deep-agent-fallible.html This doc is topic to copyright. Aside from any truthful dealing for the aim of personal examine or analysis, no half could also be reproduced with out the written permission. The content material is offered for info functions solely.

Discover additional

OpenAI tailors model of ChatGPT for US authorities shares