Written by: Techub News Compilation

In a recent podcast episode of "Core Memory," OpenAI's two co-founders, Sam Altman and Greg Brockman, rarely participated in an in-depth interview together. This conversation took place against the backdrop of OpenAI's internal strategic adjustments, ongoing external competition, and legal disputes. The two key figures not only reviewed the company's ten-year journey but also articulated their thoughts on the development of AI technology, safe deployment, societal impact, and future visions. Below is a summary of the core content from this conversation.

1. A Decade of Storms: From "Too Late" to Industry Leader

Reflecting on the founding of OpenAI, Sam Altman and Greg Brockman admitted that they initially worried it was "too late," believing that organizations like DeepMind were far ahead. After a dinner in 2015, the two reached a consensus on their drive back to the city: despite the immense challenges, the goal of creating beneficial AGI (Artificial General Intelligence) for all humanity was too important to postpone.

Regarding their ten years of collaboration, Greg Brockman described their relationship as "built in the trenches." In an environment filled with drama, power struggles, and immense pressure, having a trustworthy partner with a shared background is crucial. Their work style consists of "continuous synchronization," with short calls possibly occurring five times a day, and this close collaboration continues to this day.

Their roles are complementary: Sam Altman excels at proposing grand visions and connecting different ideas while always focusing on the ultimate goal; Greg Brockman, on the other hand, emphasizes focus and often asks, "Is this the most important thing?" pushing the company to concentrate on the most critical tasks. A notable example is their discussion on "safety". Early on, discussions about safety in AI often fell into a specific framework that sometimes had power dynamics. Greg Brockman insisted that OpenAI must find a different way to talk about and practice safety, which ultimately gave rise to the idea of "iterative deployment"—gradually rolling out products to the world under controlled risk conditions, allowing society to evolve with them rather than keeping them secret.

Greg Brockman emphasized that one of OpenAI's core principles is "resilience." Ensuring that AGI benefits all humanity cannot rely solely on a single technological solution (such as thought chain monitoring) but requires a multifaceted and multi-layered societal response. This requires technology to interact with the real world, allowing people to participate early in the technological development to understand and shape its impact.

2. From Technical Demonstration to Value Perception: Why ChatGPT Became a Turning Point

Before the launch of ChatGPT, OpenAI attempted to herald the arrival of the AI era through well-crafted blog posts, winning video game competitions, and even creating robots capable of solving Rubik's cubes with one hand. However, Sam Altman admitted these efforts had little effect, "nobody cared at all."

Everything changed with the release of ChatGPT. Although it was far from OpenAI's most advanced achievement from a purely technical perspective, it allowed ordinary people to personally experience and use AI for the first time, thus gaining value. Sam Altman believes it was this perceptible experience that collectively "updated" the world's perception and began to take AI seriously. The follow-up code generation models (such as Codex) then brought about a second cognitive leap, showing how AI can directly empower creativity.

Greg Brockman used a friend's story as an example: a friend simultaneously inputted the description of an app his sister wanted into Codex while she was describing it, and a few hours later, a complete application that matched the description was generated. When the sister saw the finished product, she exclaimed, "What is this? Who made it?" The friend replied, "You did." This "aha moment" is key to the democratization of AI.

However, both founders pointed out that current AI capabilities still exhibit a "jagged frontier"—performing impressively in certain areas (like solving mathematical problems) while still falling short in others (like soulful writing). Sam Altman explained that this is largely due to the difficulty of evaluation: in mathematics and science, judging right from wrong is relatively straightforward; whereas in open-ended tasks like creative writing, defining standards for "good" and providing feedback signals is much more complex. Nonetheless, as AI becomes smarter, it will also be better able to provide such evaluative signals, pushing the boundaries of capability further.

A recurring pattern is that whenever OpenAI releases a powerful new model (like GPT-4), there are always voices saying, "That's enough, no need to be smarter, just cheaper will do." But months later, when looking back at old versions, people would feel, "What a terrible thing at that time." Sam Altman believes that the world’s expectations and possibilities for AI capabilities are being rapidly refreshed on a quarterly basis.

3. Looking to the Future: Personal AGI, Agents, and Computational Equity

OpenAI's thoughts on future development focus on several core directions:

1. From "Computer Work" to "Personal AGI"

Greg Brockman explained the company's product strategy focus. He believes the industry is clearly shifting towards the "Agent" era. Future AI will no longer be a tool requiring users to repeatedly explain and coax, but a partner that understands the complete context, knows the user's life, and acts on behalf of the user. He calls it "Personal AGI."

This requires AI to access users' computers and browsers, and learn about their interpersonal relationships, schedules, preferences, etc., within the scope permitted by users. For example, if AI knows you like a particular musician and discovers they are performing in town soon with discounted tickets, it can proactively purchase them for you. This experience will fundamentally change the nature of human-computer interaction.

At the same time, AI will be deeply integrated into work scenarios, taking over tedious "computer work" (Greg emphasized that this term is more fitting than "knowledge work"). OpenAI is working to expand Codex's capability from software engineers to all professionals who need to operate computers.

2. Strategic Focus and Product Selection

To concentrate resources to achieve the above vision, OpenAI recently made some strategic adjustments. Greg Brockman revealed that the video generation model Sora is the "most obvious" project to be adjusted. The reason is that the model technology Sora relies on is not on the same "technology tree" as the core GPT path, and its application scenarios (creative expression) do not fully align with the company's current focus on "computer work" and "personal AGI." He emphasized that the Sora team's work is excellent and that the technology may be used for other applications in the future, but it is not a priority at this stage.

The company's current core priorities are: 1) Building an outstanding agent platform; 2) Applying agent capabilities to "computer work"; 3) Developing "Personal AGI." These three will ultimately merge into a unified AGI that helps users solve problems.

3. Prosperity and Inequality: Core Challenges of the AI Era

Regarding concerns that AI may exacerbate social inequality, Sam Altman candidly stated this is at the heart of OpenAI's founding mission: how to ensure the most powerful technology ever can truly benefit all humanity.

He painted two possible futures: one where the "floor" (i.e., the basic living standard in society) is significantly raised, with everyone ten times wealthier than now, but at the same time, those adept at using AI and computational resources could become trillionaires, increasing inequality. The other scenario has a smaller increase in the "floor" (e.g., everyone becoming twice as wealthy), but a decrease in inequality.

Sam Altman and Greg Brockman personally lean toward the first scenario, believing that significant overall prosperity is more desirable. However, they also understand the emotional contradictions people feel—when tools have compound growth effects, the fear of inequality is real.

Greg Brockman proposed a third pathway: the democratization and cost reduction of computational resources. If everyone can obtain powerful computational capabilities (running AI agents) affordably or even for free, similar to using smartphones, then anyone with the desire and will to learn and create will have unprecedented opportunities for development. This would be the ultimate version of the "American Dream." Therefore, ensuring ample and affordable computational resources is key to preventing the extreme polarization of inequality.

Sam Altman added that OpenAI's decision to launch ChatGPT rather than keeping it a secret stems from the idea of letting the technology benefit more people. Meanwhile, the non-profit foundation of OpenAI holds a substantial stake in the company (about 25%-30%), and its value (which could reach hundreds of billions if the company succeeds) will all be used to "benefit the world," which is also a safeguard of the company's structure for its mission.

4. Competition, Safety, and Geopolitics

When discussing competitors like Anthropic, Sam Altman acknowledged their excellent execution in some areas, such as recognizing earlier the need to adapt models to "chaotic real code bases," which prompted OpenAI to accelerate its focus in related fields. He believes healthy competition enhances execution levels on both sides. Currently, he believes OpenAI's products lead in many aspects.

Regarding the recent discussions about "some AI models being too powerful to be widely released" (implicitly referring to remarks from companies like Anthropic), Sam Altman expressed caution. He believes there are indeed genuine safety concerns (like cybersecurity), but "fear marketing" may also be used as an excuse to keep technology controlled within a small circle. OpenAI's approach is to establish a "preparatory framework," devising a tiered deployment plan (like a trusted access program), allowing the world to contact, understand, and co-create this technology as much as possible while ensuring safety.

On the hardware and geopolitical competition front, both founders expressed serious concern about America's lagging status in tangible areas like robotics and chip manufacturing. Sam Altman believes without a disruptive combination like "AGI + robotics," it will be challenging for the United States to catch up quickly. He envisions that if there were a powerful "robot instruction model," like Codex, capable of directing robots to configure factories or mine resources, the game rules would change. OpenAI is actively exploring the field of robotics as they recognize it is key to expanding AI capability into the physical world.

5. Lawsuits, Disputes, and Commitment to Mission

Inevitably, the topic turned to OpenAI’s legal disputes with Elon Musk. Greg Brockman detailed the key points of the negotiation breakdown: everyone agreed that OpenAI needed to transform into a for-profit company to gain resources, but fundamental disagreements emerged over equity and control. Elon Musk demanded a majority stake and to serve as CEO to ensure "everyone knows he is in charge," even insisting on "absolute control."

Greg Brockman stated that it was this last point—“regardless of who that person is, no one person should control the entire future”—that touched the bottom line of OpenAI's mission, leading to the negotiation breakdown. For many years, OpenAI has not publicly rebutted the other party's narrative, but this lawsuit will force the company to "tell the truth" and clarify its true motivations and stance to the world. He believes this provides a chance for the public to understand what OpenAI is truly fighting for.

Finally, when discussing the increasingly intense "dramatic" disputes in the AI field and the recent extreme events targeting Sam Altman personally, the two founders appeared frustrated yet resolute. Sam Altman admitted the incidents are discouraging and frightening, but he praised Greg Brockman’s steadfast focus on the mission under pressure. They called for AGI to be a collective project completed by all humanity and not to become a battleground for individual or single ideological competition.

Sam Altman concluded that OpenAI thinks daily about how to deliver technology that enables everyone to thrive and gives individuals more control over their futures. They believe that although the process will have twists and turns, it will eventually lead to a better world. Anyone who shares this goal is a partner they are happy to work with.

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。