The forthcoming significant advancement in AI may draw inspiration from the legendary physicist Albert Einstein.
"Large language models [LLMs] do make errors, and they always will. However, the frequency of these mistakes is expected to decrease over time. Eventually, the error rate will be low enough that LLMs can be applied in fields like medicine and law," stated Jonathan Ross, founder and CEO of Groq, on Yahoo Finance’s Opening Bid podcast (watch above; listen below).
Ross suggested that the capability of AI models to generate original ideas is an impending development that could occur within his lifetime. At present, LLMs primarily select the most likely responses to inquiries, which impedes their potential to unearth novel concepts.
"If it's merely the most apparent answer, it won't represent quality writing. It won't yield good science or foster innovative creations. It won't discover a new drug. That's our next goal: to enable innovation," Ross elaborated.
During his time at Google (GOOG, GOOGL), Ross engineered the custom chips used by the tech giant for AI model training.
Now, Ross has a unique perspective on the future evolution of LLMs in the coming years.
Explore more insights from the Opening Bid podcast.
What to expect from the impending 'decade of reckoning' for the bond market
Insights from Goldman Sachs’ tech chief about human oversight of AI agents
Guidance for investors during challenging market conditions
Groq develops what they refer to as language processing units (LPUs), designed to enhance the speed and efficiency of large language models, surpassing the capabilities of Nvidia's (NVDA) GPUs focused on LLM training.
By optimizing performance, Groq's chips can guide LLMs toward becoming inventive rather than solely relying on reasoning, unlike current chatbots such as ChatGPT.
The latest funding round for Groq occurred in August 2024, where it secured $640 million from investors like BlackRock (BLK) and Cisco (CSCO).
At that time, the company was valued at $2.8 billion, representing a small portion of Nvidia’s over $3 trillion market cap. Its current valuation stands at $3.5 billion, as per Yahoo Finance's private market