Tesla bull and Grok investor Pierre Ferragu has shared his thoughts on the true nature of large language models (LLMs) amid ongoing debates in the AI space.
What Happened: Over the weekend, Carlos E. Perez, co-founder of Intuition Machine, took to X, formerly Twitter, questioning the capabilities of LLMs.
In his post, Perez noted that while LLMs can tackle complex problems, they often falter on seemingly simple logical steps.
His post spotlighted a study titled “Procedural Knowledge in Pretraining Drives Reasoning in Large Language Models,” which found that LLMs’ reasoning abilities are significantly influenced by programming code logic.
The study used EK-FAC influence functions to identify the specific training data that most impact the model’s output for a given query.
See Also: SpaceX’s Starlink Direct-To-Cell Service Gets Commercial License From FCC: Here’s What It Means
The research discovered a stark contrast in how LLMs handle factual and reasoning questions. LLMs often used a retrieval-based approach for factual questions.
However, for reasoning questions, LLMs consistently relied on documents demonstrating procedures—algorithms, formulas, and, importantly, code—for solving similar problems.
Sharing his post, Ferragu, an analyst at New Street Research, stated, “My left brain: LLM are digital gods. My right brain: LLM are glorified digital imitating monkeys. Time will tell and the truth is likely right in-between.”
Subscribe to the Benzinga Tech Trends newsletter to get all the latest tech developments delivered to your inbox.
Why It Matters: The debate on LLMs’ capabilities is not new. Earlier this year, a software engineer at Alphabet Inc.'s Google expressed concerns that OpenAI, the parent company of ChatGPT, had hindered the progress of AGI by 5 to 10 years.
Salesforce CEO Marc Benioff also warned that the world was nearing the “upper limits” of LLMs like OpenAI’s ChatGPT.
He predicted that the future of AI would focus on autonomous agents capable of performing tasks independently, rather than relying on LLMs for advancements.
Tony Fadell, the co-creator of the iPod, also expressed concerns about LLMs.
Previously, Nvidia CEO Jensen Huang said that humans will eventually work with AI agents and AI employees. Nvidia has also partnered with Accenture to deploy AI agents in businesses.
Microsoft Corporation also announced plans to let companies create their own autonomous agents, following Salesforce’s launch of Agentforce in September 2024.
OpenAI is also reportedly planning to launch a new AI agent, “Operator,” in January.
Check out more of Benzinga's Consumer Tech coverage by following this link.
Read Next:
Disclaimer: This content was partially produced with the help of Benzinga Neuro and was reviewed and published by Benzinga editors.
© 2024 Benzinga.com. Benzinga does not provide investment advice. All rights reserved.
Comments
Trade confidently with insights and alerts from analyst ratings, free reports and breaking news that affects the stocks you care about.
