In a recent podcast, experts from venture capital firm Andreessen Horowitz (a16z) ignited a thought-provoking discussion. Participating artist Yoko Li gave an expansive, poetic account of an AI Agent. She described it as a common-sense, multi-step big language model (LLM) with an interactive decision tree. This definition highlights the complexities and evolving nature of AI technology, but it underscores a significant point: there remains no universally accepted definition of what an AI Agent truly is.
The discussion delved into many aspects of AI Agents, such as their emerging presence in the workforce and how to mitigate negative impacts. Andreessen Horowitz is pushing the idea that every white-collar job is going to have an AI copilot in short order. The ambitious firm is currently in the process of raising a $20 billion dollar megafund. This fund will begin a concerted effort to make transformative investments in the burgeoning AI sector.
Throughout the course of the four-day podcast, we struggled to pin down what exactly AI Agents are. Appenzeller emphasized the need to have an AI Agent that works well. He said it has to be “something very near AGI,” or Artificial General Intelligence. He stressed that an AI Agent has to last for decades. It also needs to proactively solve issues on its own to build its trustworthiness.
“It needs to persist over long periods of time,” – Guido Appenzeller
“It needs to work independently on problems,” – Guido Appenzeller
The panelists expressed that getting to this level of functionality has turned out to be an unexpectedly difficult task. Connect with Artisan CEO Jaspar Carmichael-Jack, whose company recently raised $25 million. He stressed what a difficult task it is to have AI Agent technology operate robustly and reliably. Artisan brakes on the viral trend to “stop hiring humans.” They’re still hiring humans, though, since today’s AI tech isn’t perfect and still can’t do it all.
Unfortunately, the fight over the definition of AI Agents doesn’t stop with venture capitalists. It’s not just the effect on tech that makes it such a big deal. Startups are jumping at the opportunity to label their product AI Agent for all kinds of strategic reasons, whether marketing or pricing. This trend has helped fuel increased head-scratching over what exactly is an AI Agent.
Maryland DOT’s chief of staff, Matt Bornstein, expressed this concern on the panel with the comment, “I’m just not sure that even is sort-of theoretically possible.” His comments indicate skepticism about whether current AI technologies can truly replace human creativity and critical thinking in most job roles.
The enthusiasm generated by AI Agents is further exacerbated by public excitement, as well as marketing speak, which dilutes the public discourse on this topic. Even amidst these challenges, titans of industry like Geoffrey Hinton are betting on a bright future for the AI tech. Yann LeCun, the foremost AI expert, foresees the emergence of a new architectural paradigm. He’s betting on this breakthrough occurring within the next five years. Such innovations would be key to creating AI Agents that could one day replace human laborers on many jobs.
OpenAI, creators of ChatGPT, and Anysphere, developers of AI-powered therapy, are two of the companies focused on creating AI-backed solutions in the firm’s portfolio. Their participation is a clear indicator of the firm’s ambition to direct the future course of chatter around all in artificial intelligence.