The artificial intelligence (AI) industry is rapidly evolving, with the next frontier being artificial general intelligence (AGI) — systems capable of human-level reasoning and task completion. Despite massive investments in startups like Anthropic (raising billions) and Mistral AI (achieving unicorn status), experts believe the true potential of AI remains untapped. Himanshu Tyagi, co-founder of Sentient and a professor at the Indian Institute of Science, asserts that the path to AGI lies in decentralized AI.
Decentralized AI: Unlocking the Data and Compute Bottleneck
Tyagi argues that AGI requires “completely new data on human strategies and specialized models trained on this data.” This data goes beyond what is available on the internet; it includes deep heuristics and strategies used by humans in complex tasks such as technical interviews, sales techniques, or brand design. Collecting such data through centralized silos would be of limited utility, he says. Instead, he advocates for “decentralized, open, and incentivized mechanisms” to gather valuable data. Similarly, model development must allow people to freely contribute trained models with specific skills and alignment, while providing compute resources at Google scale. “Decentralized model ownership with incentives and decentralized training solves these problems,” Tyagi stated.
Warning to Young Developers: Beware of Early Optimism
Tyagi cautions young developers about the “great initial optimism” that often accompanies building AI applications. The journey from proof-of-concept to a stable, scalable product is fraught with challenges, as large language models (LLMs) introduce errors such as hallucinations, factuality issues, and security vulnerabilities. Addressing these requires a new software layer and specialized model training — capabilities that early-stage teams may lack. His advice: “Sharply focus on their specific use case and rely on external offerings for resolving these issues.” Sentient Chat is designed to provide such services, offering AI search APIs, hosted models, agentic frameworks, and Trusted Execution Environment (TEE) libraries as accessible tools for agent builders. Notably, Sentient’s models are open-source, allowing developers to understand functionality and avoid vendor lock-in.
Sentient Chat: Disrupting Traditional Search
Sentient Chat aims to challenge traditional search engines by building a community-owned AI chatbot. Tyagi points out that Google has dominated search for decades but its model is limited to finding information on the internet, and its advertising revenue model makes it hard to transition. AI offers an opportunity to transcend this limitation: “We can simply get things done directly instead of gathering information first, analyzing it, and then taking action.” To realize this vision, Sentient Chat is building an ecosystem of AI agents powered by diverse data sources and contributions from a community of developers. “To realize this crazy future, we need many varied sources of indexed data and many builders to offer agents that take the final action,” Tyagi emphasized. The platform operates under community governance, ensuring transparency and incentivizing participation from data providers and agent builders. Tyagi also hinted at rapid expansion, noting that “there are much more than 15 agents coming on Sentient Chat!” This positions Sentient Chat as a dynamic, action-oriented alternative to traditional search, empowering users to accomplish tasks directly through AI agents.