Published on January 2, 2024, 6:14 pm
This year in the field of artificial intelligence (AI), one predominant trend is the rise of “hybrid” AI and applications based on large language models, according to Goldman Sachs’ chief investment officer, Marco Argenti. In an interview about the year ahead, Argenti highlighted how hybrid AI combines larger models as the brain to interpret user prompts and orchestrate tasks for specialized worker models.
The implementation of such large programs will likely be limited to the world’s wealthiest companies due to cost considerations. However, other enterprises can still leverage smaller neural networks trained on their proprietary data, either within their own data centers or through cloud computing services. This approach aligns well with the growing preference for utilizing specialized models fine-tuned with corporate data, as exemplified by frameworks like LangChain, which builds on generative AI.
Argenti predicts that most companies in 2024 will focus on proof-of-concept initiatives that yield the highest returns when embracing hybrid AI. Additionally, he anticipates a new wave of third-party applications built on top of foundation models emerging during this period. By considering these foundation models as operating systems or platforms, there is immense potential for innovative applications to emerge around them.
Looking further into the future, Argenti emphasized the importance of coordinating security measures among different parties in order to manage potential risks effectively. Encouraging collaboration and open sourcing of models while developing principle-based rules will aid in addressing concerns related to bias, discrimination, safety-and-soundness, and privacy. By fostering such an environment, the United States can maintain its leadership position in AI development.
In conclusion, this year will witness the ascendancy of hybrid AI and applications built upon large language models. While building extensive programs based on these models may be reserved for the wealthiest corporations, other enterprises have the opportunity to employ smaller neural networks trained on their proprietary data. Furthermore, third-party applications are anticipated to flourish alongside foundational models. As we progress in this AI-driven era, it is crucial to prioritize collaboration, open sourcing, and principled regulations to ensure the responsible and successful development of AI.