‘It’s slop’: OpenAI co-founder Andrej Karpathy pours cold water on agentic AI hype – so your jobs are safe, at least for now
With concerns rising about the impact of agentic AI on the labor market, OpenAI co-founder Andrej Karpathy thinks agents are nowhere near ready to begin stealing jobs.
In a recent appearance on the Dwarkesh Podcast, Karpathy described the output of AI agents as “slop” and suggested the latest iteration of the technology isn’t quite living up to the hype.
“They’re cognitively lacking and it’s just not working,” he told host Dwarkesh Patel. “It will take about a decade to work through all of those issues.”
Karpathy added that, in their current form, AI agents “just don’t work” and aren’t quite the silver bullet for productivity and efficiency that major industry providers have framed them as.
“I feel like the industry is making too big of a jump and is trying to pretend like this is amazing, and it’s not. It’s slop”.
Big tech goes all-in on AI agents
While Karpathy’s comments appear to pour cold water on the potential of AI agents, leading industry stakeholders have been banging the drum for the technology repeatedly over the last year.
Microsoft, for example, said it expects 1.3 billion AI agents to be operational by 2028, while Salesforce has essentially pinned future success on the potential of its own agentic AI service, Agentforce.
Launched last year, CEO Marc Benioff hailed the arrival of agents as the next big step in the evolutionary path of generative AI.
A key factor behind the appeal here is that agents can carry out tasks on behalf of human workers. They’re autonomous bots that can pick up the slack anywhere, and any time.
By contrast, AI assistants, which represented the first iteration of the tech during the early days of the generative AI ‘boom’, act as a supplementary tool for individual employees.
With this in mind, leaders across the industry have touted the potential for agents to streamline efficiency across a range of business functions, from customer support to software engineering and even in manual trades.
At the company’s annual Dreamforce conference earlier this month, Salesforce once again doubled down on agentic AI in a rallying call for customers to jump on the bandwagon.
But even while hailing the technology, adoption rates at the CRM giant show it’s struggling to achieve significant uptake.
Too much hype is causing serious issues
Karpathy isn’t the first in the industry to question the hype surrounding AI agents.
Indeed, Capgemini research shows IT leaders are still largely skeptical about the technology and are keeping their cards close to their chest on any investment in this domain.
Earlier this year, analysis from Gartner suggested that a concerning number of so-called agentic AI products were just repackaged software or RPA tools or basic chatbots.
“Agent washing,” as the consultancy described the trend, has become a serious issue as vendors look to capitalize on enterprise interest in the technology.
Longer term, the outlook isn’t exactly positive, according to Gartner, with 40% of agentic AI projects expected to be cancelled within the next two years.
This is largely due to poor returns on investment and the fact that most current use-cases are immature, which aligns closely with Karpathy’s comments regarding immediate benefits.
Previous research from Gartner on specific use-cases for agentic AI also show the technology has fallen flat, particularly in customer service and support roles.
Analysis from the consultancy in June noted that while workers in these positions were among the first on the AI chopping block, many are now backtracking on the trend.
Around half of companies that expected to “significantly reduce” customer service workforces due to AI agents are now projected to abandon plans by 2027.
Make sure to follow ITPro on Google News to keep tabs on all our latest news, analysis, and reviews.