Mark Zuckerberg Says AI Agent Tech Advancing More Slowly Than Expected

Meta CEO Mark Zuckerberg has admitted that AI agent technology is advancing more slowly than the company expected. He made the comment during an internal town hall with staff.

Zuckerberg said the journey to superintelligence is on track but will take time and hard work. He acknowledged the competitive nature of the AI landscape and the challenges involved.

Meta is investing heavily in AI development. The company is pouring tens of billions of dollars into talent and infrastructure to build advanced models.

However progress on AI agents has not moved as fast as hoped. Zuckerberg pointed out that building reliable AI agents requires significant effort and time.

The company is also adjusting some of its internal AI initiatives. Chief Technology Officer Andrew Bosworth said the AI training program that used employee keystrokes will now be opt in only.

This change came after a leak and backlash from employees who were uncomfortable with their activity being recorded. Meta had paused the program last month.

Bosworth acknowledged that the rollout had damaged morale and trust within the company. He noted that the program generated more useful data than expected but the mandatory approach caused problems.

The comments reflect the tension at Meta as it pushes aggressively into AI. The company is trying to balance speed with employee concerns and trust.

Meta recently gave engineers the option to leave its Applied AI task force after reassigning thousands of them. Some employees called the move the undraft.

Zuckerberg remains optimistic about Meta's AI future. He said the company expects to see some benefits within the next three to six months.

The AI race continues to intensify. Companies like Meta are investing heavily but also learning that moving fast comes with challenges.