"The Smart Newcomer Never Stops Working... This 'Rookie' Is Intimidating [Click e-Industry]"

The Computing Race Ignited by GPT-6
"AI Now Works Independently for Hours"

Beyond Basic Q&A: AI Handles Real-World Tasks
Computing for Long-Term Reasoning Grows in Importance

Securities industry experts have forecast that, with the launch of OpenAI's new artificial intelligence (AI) model, 'GPT-6 Astra', there will be a significant increase in the amount of time AI is able to independently perform work, and the competitive landscape of the AI industry will change dramatically.

A newly hired "rookie" who recently joined the company is rapidly expanding their scope of work. There is a growing expectation that artificial intelligence (AI), which operates independently without human intervention, will establish itself as a new workforce. The photo is an image representing a new employee. Photo by Getty Images

On September 7, Lee Youngjin, a researcher at Samsung Securities, stated, "The most important change with GPT-6 Astra is the extension of its time horizon, allowing autonomous completion of tasks without human intervention," and added, "The competitive benchmarks in the AI market will expand beyond simply having a model to how much actual work can be processed."

Expansion of Autonomous Operation Time... Execution Environment Design Determines Performance

With the launch of OpenAI's new artificial intelligence (AI) model 'GPT-6 Astra,' it is anticipated that the time AI independently performs tasks without human intervention will significantly increase. Getty Images

Astra has enhanced its ability to use computers and tools for extended periods, perform complex multi-step tasks, and utilize multiple agents simultaneously. The capacity for autonomous execution—going beyond just accuracy on individual problems to independently completing tasks over several hours—has become a key factor in driving genuine productivity differences.

Lee explained, "Astra has become the first to reach the 'Critical' level in cybersecurity under OpenAI's safety assessment framework," noting, "This advancement is not just about greater knowledge of vulnerabilities, but is based on improved ability to explore actual systems, identify new vulnerabilities, and solve problems."

He also emphasized the importance of the execution environment, based on findings from the ARC-AGI-3 evaluation. Lee said, "Even with the same model, a performance gap of about 37 percentage points was observed depending on the approach to state management and context retention," adding, "In future AI competition, not only the core capabilities of the model, but also how memory, tools, and execution environments are designed to ensure stable long-term operation will be essential factors."

AI Accelerating AI Research... Ongoing Increase in Computing Infrastructure Demand

With AI now actively deployed in research work to develop next-generation AI, the acceleration of research is becoming increasingly evident. According to OpenAI's research organization, agents are performing a workload equivalent to roughly 3.1 days for each one day of human researcher labor. Individual researchers operate multiple agents in parallel, delegating coding, debugging, experiment execution, and result analysis to them.

Lee noted, "A flywheel structure is forming, where improvements in research throughput and productivity lead to further enhancement of AI performance," and added, "As human time gradually ceases to be a bottleneck, computing becomes increasingly important as the primary constraint determining research speed."

It is analyzed that, even if improvements in model efficiency reduce the number of tokens and cost required for individual tasks, the overall demand for inference and computing could surge explosively, as the range of work that can be entrusted to AI expands and the use of parallel agents increases.

AP Yonhap News

'Value Assessment Shifting to Cost per Task'

The software industry is also facing a structural transformation. As agents begin to move seamlessly across multiple software platforms and carry out entire business processes, companies are tasked with evolving their products from simple function providers into the core execution environments for agents.

Lee said, "The value metric for AI is already shifting from price-per-token to cost-per-task," and added, "Going forward, metrics such as work completion rate, the cost to finish a task, and the amount of human labor time that can be saved will be the key indicators for evaluating AI's economic viability, rather than just the price per token."

This content was produced with the assistance of AI translation services.