Today's large language models (LLMs) have impressive capabilities but must be revised. They struggle with hallucinations, lack real-world grounding, unreliable reasoning, opacity, and bias. These issues stem from their core architectures and training methodologies, making them unsafe and not environmentally sustainable. Investment should focus on new AI paradigms that address these issues, such as task-specific models, embodied agents, hybrid neuro-symbolic systems, and interpretable models. These approaches emphasize domain-specific expertise, real-world interaction, transparent reasoning, and reliable behavior, creating AI systems that are safer, more robust, and aligned with human values. The venture capital community must balance investments in current LLMs with bold bets on startups pioneering these innovative approaches. This diversified strategy will ensure AI evolves into a powerful, ethical, and beneficial societal tool. While today's LLMs have significant limitations, exploring new paradigms and investing in innovative AI approaches is essential for developing responsible, safe, and green AI technologies.
We firmly believe that AI systems must respect human rights, embrace diversity, and promote fairness. This principle guides us to scrutinize how AI technologies are designed and implemented, ensuring they promote equality and not perpetuate or exacerbate existing biases. The current generation of large language models (LLMs) has made significant strides in simulating human-like intelligence, but they are fundamentally flawed. Issues such as hallucinations, lack of grounding in real-world contexts, unreliable reasoning, opacity, and potential bias arise from their core architectures and training methodologies. These problems are not mere bugs but inherent limitations that question these models' safety, robustness, and true intelligence.
To address these critical issues, we must explore new AI paradigms that move beyond the current approach of training massive neural networks on vast datasets. Task-specific models, embodied agents, hybrid neuro-symbolic systems, and interpretable models represent promising alternatives. These approaches prioritize domain-specific expertise, real-world interaction, transparent reasoning, and reliable behavior, paving the way for AI systems that are more capable, safer, and more trustworthy.
The venture capital community plays a crucial role in this transition. We can drive significant breakthroughs by balancing investments in scaling up existing LLMs with bold bets on startups pioneering these innovative approaches. The future of AI lies not just in incremental improvements but in fundamentally rethinking and redesigning our AI systems to align with human values and societal needs.
As we chart the course of artificial intelligence in the coming years, a diversified investment strategy that includes both incremental scaling and revolutionary innovations will be essential. This approach will ensure that AI evolves into a powerful tool that benefits humanity, minimizes risks, and adheres to ethical standards, ultimately leading to more responsible and beneficial applications across various domains.
As a side note and in full transparency, I do use today’s LLMs and teach students and professionals how to use them as productive tools. However, part of the curriculum teaches the issues and limitations and how to mitigate them.
LISTEN NOW: LLMs Will Never Be Responsible, Safe or Green (18 MIN)