Yet in technology’s rapidly changing world, especially in artificial intelligence, much confusion remains about what an AI agent is and what it can do. Industry experts from the venture capital firm Andreessen Horowitz (a16z) have expressed skepticism about the technology. They particularly challenge its potential to substitute for human labor in the labor market. In a recent conversation, Guido Appenzeller, one of the foremost experts in the AI industry, provided his perspective. He noted that for AI software to be widely able to take on government jobs, it would have to be at least close to AGI.
Appenzeller wrote the steps necessary for creating AGI. He reiterated that any AI system would need to be able to endure over time and address grand challenge type problems on its own. He noted, “it needs to persist over long periods of time,” and “it needs to work independently on problems.” He noted that major technical hurdles remain. These have some significant challenges such as developing persistent long-term memory and overcoming the cost of deploying these capabilities. In addition to this, Appenzeller raised the problem of “stomping out hallucinations,” which is when AI systems produce outputs that are wrong or make no sense.
The very definition of an AI agent is hotly debated. As Yoko Li defined one AI agent, it’s “a reasoning tool.” This agent is a multi-step large language model (LLM) agent with dynamic decision tree branching. There is a lack of consensus on what exactly an AI agent is. Matt Bornstein conveyed his doubts about the feasibility of a future where humans are entirely replaced by AI, stating, “I’m just not sure that even is kind-of theoretically possible.”
For all the bewildering hype around AI agents, a16z is doubling down on this space. The firm has been a vocal and active supporter of multiple AI companies, including OpenAI and Anysphere. It’s apparently raising a record-breaking $20 billion megafund to turbocharge its AI investments. This extensive fiscal support is a testament to the faith that every knowledge worker will soon have an AI-based co-pilot. Both Appenzeller and Li admitted that these copilot systems “don’t work today.”
Against this backdrop of AI uncertainty, dozens of new startups are trying to cash in on the hype and excitement of AI agents. For instance, Jaspar Carmichael-Jack’s company, Artisan, launched a viral ad campaign promoting the idea of “stop hiring humans,” despite continuing to hire human workers. This open contradiction begs the question of how serious the impending human replacement rhetoric really is. In doing so, it further fuels confusion in the industry. In particular, Bornstein observed that this rhetoric is one of the biggest culprits in creating confusion over what AI can (and cannot) do.
In conversation, Appenzeller observed a trend of nascent startups marketing their products as agents. They’re doing this mostly to capitalize on the current, raging hype for AI. He remarked, “The simplest thing that I’ve heard being called an agent is basically just a clever prompt on top of some kind of knowledge base.” Taken together, this observation underscores a key flaw. Some startups may be puffing out their chests about their products’ abilities to attract investment and attention.
As discussions about AI agents continue to evolve, experts remain cautious about making definitive predictions regarding their future impact on the workforce.
Leave a Reply