What is Deep Analysis, OpenAI’s new AI tool that provides analyst-grade stories? | Technology News

OpenAI unveiled Deep Analysis’s capabilities in an illustration video posted on YouTube. The corporate additionally privately showcased the AI tool to US lawmakers and different authorities officers in Washington DC final week, in line with The New York Instances.

The AI startup behind ChatGPT has been put in a precarious position by DeepSeek, which claims that its reasoning mannequin (R1) rivals OpenAI’s o1 mannequin in efficiency, whereas utilizing lesser superior graphics processing items (GPUs) and costing considerably lesser to develop.

This purported breakthrough in compute effectivity has led to a notion that OpenAI is shedding floor to Chinese language upstarts like DeepSeek within the AI arms race. In response to the thrill round DeepSeek, OpenAI CEO Sam Altman had stated that the corporate would “ship a lot better fashions” and can “pull up some releases.”

deepseek’s r1 is a formidable mannequin, significantly round what they’re in a position to ship for the value.

we are going to clearly ship a lot better fashions and in addition it’s legit invigorating to have a new competitor! we are going to pull up some releases.

— Sam Altman (@sama) January 28, 2025

Deep Analysis is the second AI agent launched by OpenAI in a single month. A couple of weeks in the past, it previewed an AI agent referred to as Operator that is able to looking out on the web and visiting web sites on a consumer’s behalf to buy groceries, make restaurant reservations, and so on.

What is Deep Analysis? How was it developed?

Whereas OpenAI’s Operator agent is constructed on high of its multimodal massive language mannequin (LLM) GPT-4o, Deep Analysis appears to be the primary AI tool to be powered by a version of the upcoming o3 model that has been optimised for net shopping and information evaluation.

In response to OpenAI, Deep Analysis builds on the reasoning capabilities of o1 to hold out complicated duties that require “intensive context and knowledge gathering from numerous on-line sources.”

Deep Analysis was educated on real-world duties throughout a variety of domains that demand the usage of a browser and Python tool. The coaching methodology employed was end-to-end reinforcement studying.

“By means of that coaching, it realized to plan and execute a multi-step trajectory to seek out the information it wants, backtracking and reacting to real-time info the place obligatory,” OpenAI stated in a weblog put up revealed on Sunday.

Tips on how to use Deep Analysis? What can it do?

In response to a consumer’s immediate, Deep Analysis will present a complete report after discovering, analysing, and synthesising content material equivalent to textual content, pictures, and PDFs from a number of on-line sources.

The AI-generated analysis stories will comprise synthesised graphs and images from websites, together with clear citations and a abstract of its pondering. Customers can even add particular information and spreadsheets so as to add context to their immediate.

OpenAI claimed Deep Analysis is particularly good at uncovering hard-to-find info that would normally require looking out by way of a number of web sites. For instance, a consumer might describe a scene from any TV present and ask it to dig by way of the online to seek out the precise episode the place it occurred.

“[Deep Research] can do complicated analysis duties that would possibly take an individual wherever from half-hour to 30 days,” Kevin Weil, OpenAI’s chief product officer, was quoted as saying.

However the firm has acknowledged that the AI tool can nonetheless get issues mistaken. “It might battle with distinguishing authoritative info from rumors, and at the moment reveals weak spot in confidence calibration, usually failing to convey uncertainty precisely,” OpenAI stated.

Deep Analysis is at the moment accessible by way of the online model of ChatGPT as a part of its Professional subscription package deal that prices $200 per 30 days (Rs 16,950 approx.). Customers have to pick the ‘deep analysis’ tab within the message composer earlier than coming into their prompts.

“As soon as it begins operating, a sidebar seems with a abstract of the steps taken and sources used […] You’ll get a notification as soon as the analysis is full. The ultimate output arrives as a report throughout the chat,” the corporate stated.

Deep Analysis is additionally very compute-intensive, which implies that the longer it takes to analysis a question, the extra processing energy it requires. It is at the moment not accessible within the UK and European Union (EU) area.

How did Deep Analysis carry out in benchmark exams?

OpenAI claimed that the optimised model of o3 powering Deep Analysis, achieved a new excessive rating of 26.6% accuracy in Humanity’s Final Examination, a not too long ago launched benchmark take a look at that exams AI fashions throughout a broad vary of topics on expert-level questions.

“In comparison with OpenAI o1, the biggest positive factors [of Deep Research] appeared in chemistry, humanities and social sciences, and arithmetic,” it stated.

Moreover, Deep Analysis’s underlying o3 mannequin topped one other public benchmark take a look at referred to as GAIA that evaluates AI fashions on their reasoning, multi-modal fluency, net shopping, and tool-use skills, with questions spanning three ranges of problem.

On what’s subsequent, OpenAI stated Deep Analysis shall be upgraded to embed pictures, information visualisations, and different analytic outputs in its AI-generated analysis stories. On the consumer finish, they are going to be capable to add extra specialised information sources to make Deep Analysis’s outputs “much more strong and personalised”.