At the recent Dreamforce, a premier global technology conference held in San Francisco, Figure founder Brett Adcock gave an in-depth interview. This marked his first major public remarks following the launch of the new Figure 03.

Standing beside Figure's first- to third-generation robots, Brett Adcock stated that the greatest challenge facing the robotics industry today is not manufacturing, but achieving true intelligence that enables robots to work autonomously in the physical world.

To address this, Figure has set a goal for humanoid robots to complete general-purpose tasks via voice commands by next year, even when operating in unfamiliar environments.

While many companies are currently anxious about mass production, Adcock pointed out that manufacturing scale is not the core issue. He argued that if the problem of robotic intelligence can be solved, producing one million units would happen rapidly.

However, acknowledging current technological limitations, Adcock admitted that Figure has not yet fully solved general-purpose robotics technology, and he shared some of his perspectives on the matter.

Goal: Make robots truly intelligent by next year

When discussing the mass manufacturing of robots, Brett Adcock stated that manufacturing is currently merely 'the 50th issue,' not the core challenge. While many in the industry compete on who can produce more robots, most cannot operate without remote control. The core problem lies in endowing robots with true intelligence, enabling them to work autonomously in the physical world.

In a conversation with Salesforce CEO Marc Benioff, he sharply pointed out that many robot demonstrations on the market are pre-programmed performances, such as dancing, where many movements require manual remote operation. No company has yet managed to build one million robots and have each one operate autonomously.

Humanoid robots possess numerous degrees of freedom, meaning the number of possible states and achievable positions is astronomically large. It is virtually impossible to program every scenario to enable them to perform general-purpose tasks.

Brett Adcock's solution is to use neural networks to learn these state representations. Since the human brain can achieve this, we need to equip robots' neural networks with similar learning capabilities.

"The Figure 03 released last week completed its task demonstrations autonomously," as shown in the video. Tasks such as watering plants and washing dishes—approximately a dozen application scenarios—were highlighted. He emphasized that achieving this from R&D to demonstration took less than 30 days.

In practical applications, Figure places great emphasis on the operational reliability of its robots. Since April this year, they have been working on BMW's production line for 10 hours a day, using this approach to verify whether the robots can truly function in real-world scenarios.

Regarding when robots will achieve true intelligence, Brett Adcock made a bold prediction: "I expect that by next year, humanoid robots will be able to complete general tasks via voice commands—for example, if placed in an environment they have never visited before, they will autonomously accomplish everything you assign them."

Even though this has not yet been achieved, he still believes that Figure is at least one to two years ahead of other robotics companies globally. "If we master general-purpose robot technology now, mass-producing one million units would actually be a very quick process."

In terms of robot manufacturing, Figure previously introduced its BotQ factory during the launch of the Figure 03. It is one of the few exclusive mass-production facilities for humanoid robots in the industry, with its initial design featuring a first-generation production line with an annual capacity of 12,000 units.

Humanoid robots will ultimately become the absolute mainstream

Last month, Figure just completed a Series C funding round exceeding $1 billion, reaching a post-money valuation of $39 billion (approximately RMB 270 billion), breaking the global valuation record for humanoid robot companies.

When discussing this topic, Brett Adcock remained confident, stating that if issues such as general-purpose robot intelligence are resolved, Figure's value could reach tens of trillions of dollars.

And with a wide variety of robotic forms emerging in the market, he stated that humanoid robots will ultimately occupy an absolutely dominant position, while other forms will account for a very small share. There are three key reasons for this:

• Mass production requires a single product form factor, which significantly reduces costs;
• Training data for neural networks comes from humans; when training Figure's humanoid robots, they can directly utilize human behavioral data, an advantage not available to other robotic forms;
• Future technological iterations will drive down robot costs, potentially even leading to robots building robots. Moreover, humanoid robots can adapt to almost all scenarios, whereas other forms struggle to achieve this versatility.

Brett Adcock also pointed out that what Figure truly aims to do is develop civilian robots to enter homes and workplaces, solving difficulties people face in their daily lives and work—this is the direction they genuinely want to invest in.

Therefore, their latest release, the Figure 03, focuses heavily on these capabilities. In home settings, it can help fold laundry and accompany pets; in commercial environments, it can deliver packages and even serve as a hotel front desk receptionist.

However, Brett Adcock noted that robots currently cannot yet reach the level where "placing a robot at home and instructing it to clean the house perfectly before 5 PM" is feasible. But he revealed that Figure has already identified a clear path forward, and the company is working diligently to accelerate solutions for general-purpose robotics technology.

In the most critical issue of intelligence within robotics, let us see what answers Figure will provide in the future.