Amjad Masad, CEO of Replit, argues that the economy does not require true artificial general intelligence (AGI) to realize significant transformations. During an episode of the a16z podcast published on March 7, 2024, Masad asserted that “functional AGI” is already capable of automating substantial portions of today’s workforce. He emphasized that while the pursuit of true AGI remains a focus in Silicon Valley, practical advancements in functional AGI are both achievable and economically beneficial.
Masad defined functional AGI as systems that can learn from real-world data without needing human-like consciousness or reasoning. He stated, “We can get to functional AGI,” highlighting that such systems are sufficient for automating significant labor across multiple sectors. “We’ll target every sector of the economy, and you can automate a big part of labor that way,” he added. His perspective indicates a shift in focus from theoretical advancements to practical applications that can yield immediate economic benefits.
Despite the enthusiasm surrounding true AGI, Masad expressed skepticism about whether humanity will reach this level of intelligence, which would mimic human cognitive abilities across various fields. He noted that while true AGI could potentially elevate civilization, he is “bearish on true AGI breakthrough” due to the current utility and economic value of the advancements already being made.
Masad also warned that the industry might be caught in a “local maximum trap,” where companies optimize existing models rather than innovating. This could hinder progress toward a genuine breakthrough in AGI. “Maybe the general problem is actually not within our lifetimes,” he commented, reflecting on the complexities involved in achieving general intelligence.
The debate surrounding the viability of AGI is intensifying, especially as major AI labs like OpenAI, Google, Meta, and Microsoft continue to pursue this ambitious goal. Some experts have begun to question whether current large language models (LLMs) can evolve into true AGI. Gary Marcus, a prominent AI researcher, stated in an August blog post, “Nobody with intellectual integrity should still believe that pure scaling will get us to AGI.” His comments suggest a growing concern among industry leaders about the feasibility of achieving AGI in the near future.
The release of OpenAI’s GPT-5 in August 2024 did not meet the high expectations associated with AGI development, further fueling skepticism. OpenAI CEO Sam Altman acknowledged that while GPT-5 demonstrates advanced capabilities, it still lacks essential elements that many associate with true AGI.
Echoing these sentiments, Yann LeCun, Meta’s chief AI scientist, indicated that achieving AGI might still be “decades” away. He cautioned that increased data and computational power do not necessarily translate to smarter AI, emphasizing the complexity of scaling intelligent systems.
As discussions on the future of artificial intelligence evolve, Masad’s insights reflect a pragmatic approach to the potential of functional AGI. With its ability to transform the economy without reaching the theoretical pinnacle of AGI, this perspective could reshape how businesses and society engage with AI technologies in the coming years.
