The AI Future 5-10 Years From Now: Scientific Discovery and Robot Society as Envisioned by Sam Altman
In this article, we will organize and explain in an easy-to-understand manner the "medium-term (5-10 years) future vision of AI" that OpenAI CEO Sam Altman discussed in a podcast. From its application in the software domain to the automation of new scientific discoveries, applications in the physical world, and the impact and challenges for society, we will explain these concepts with concrete examples while citing Altman's insightful remarks.
1. Software and Productivity Improvement
AI is first going through a phase of significantly improving software productivity.
1-1. Coding and Chat
Coding Assistance: GPT-3 series models have already demonstrated code generation capabilities "on par with world-class competitive programmers," and there are cases where programmer productivity has tripled.
Chat-based Interfaces: "Chat agents" that retrieve information through dialogue are being utilized in a wide range of tasks, such as help desks and document searches.
"AI is already beginning to acquire reasoning capabilities on par with PhD-level experts" (Sam Altman)
1-2. Next-Generation Workflows
Virtual Employees
: It is expected that "virtual employees" that fully automate routine tasks will emerge, suggesting a future where AI autonomously performs tasks in collaborative environments like Google Docs.Revamping Social Experiences
: New SNS and entertainment experiences utilizing AI will be born, and the construction of services that "deliver optimized feeds with a single prompt" will progress.
2. Automation of Scientific Discovery
Altman emphasizes "automated scientific discovery" as the area with the greatest impact.
2-1. Utilizing Reasoning Capabilities
Having reached a level where "models can perform domain-specific PhD-equivalent reasoning," AI has the potential to combine existing scientific knowledge and autonomously generate new hypotheses.
2-2. Initial Application Examples and Expected Areas
Biology: There are cases where biologists have achieved a "Fundamental Leap," and AI is beginning to automate experimental design and data analysis.
Astronomy/High-Energy Physics: In fields that handle vast amounts of observational data, there is a shortage of PhD holders, and it is expected that high-speed AI screening will drive new discoveries.
"The first field where AI autonomously makes a new discovery might be astronomy, where the amount of data is vast" (Sam Altman)
3. AI Application to the Physical World
AI will move beyond the digital realm and advance into the "physical world," such as in robotics and autonomous driving.
3-1. Autonomous Driving
Regarding autonomous driving in standard vehicles, the view is that it is "nearly solved in terms of software," and it is expected that AI drivers will become widespread in fields such as walking, cargo transport, and logistics.
3-2. Humanoid Robots
Challenges: Mechanical body durability and precision were bottlenecks, but due to improvements in AI control performance, Mr. Altman says that "phenomenally excellent humanoids" will be realized within 5 to 10 years.
Social Impact: It is unknown whether robots walking in the city will be accepted by people, but their use in industry and nursing care settings is expected.
"Even if you have a perfect 'brain,' you cannot put it to practical use without an accurate 'body.' Now, that is being solved rapidly." (Sam Altman)
4. Impact on Society and Challenges
The improvement of AI capabilities brings not only benefits but also challenges in terms of employment and ethics.
4-1. Impact on Employment
With the automation of routine tasks such as customer support and market research, it is expected to follow the historical pattern where "many jobs disappear and new ones are created." Mr. Altman also shows an optimistic side, stating that "humanity has always found new roles."
4-2. Safety and Risk Management
Concerns about Weaponization: The risk of autonomous robots and AI being misused cannot be ignored.
Regulations and Social Systems: Because the speed of technological evolution is fast, legal systems and educational systems may not be able to keep up, and feelings of unfairness in value distribution may increase.
"Even if we achieve superintelligence, it is meaningless if society cannot utilize it well." (Sam Altman)
5. Future Technological Development and Outlook
Finally, we summarize the "inevitable technological path" and "uncertain parts" that Mr. Altman talks about.
5-1. Inevitable Technological Evolution
Advancement of Large-Scale Models: The spread of self-supervised learning that achieves high-precision inference with less data.
Construction of AI Platforms: The "AI Factory" concept that seamlessly integrates with various devices and services to unify the user experience.
5-2. Unresolved Social Issues
Distribution of value: Creating a mechanism to fairly distribute the abundant wealth generated by AI.
Cultural adaptation: It will take time for people to accept coexistence with AI and integrate the concept of an "AI companion" into their daily lives.
"Technically, we can see 'what needs to be done,' but there is still insufficient discussion on 'how society should be'" (Sam Altman)
In this article, we have organized medium-term AI evolution scenarios and their social impact, centered on Mr. Altman's statements. From software automation to scientific discovery, robotics, and institutional design, we provide a bird's-eye view of the future of AI from a multifaceted perspective, presenting the challenges that companies, researchers, and policymakers must each address.