If you listen to Google input/output, Spring update for OpenAIor Microsoft build This month, you may have heard the term artificial intelligence agent used a lot in the last month. They’re quickly becoming the next big thing in tech, but what exactly are they? Why is everyone suddenly talking about them?
Google CEO Sundar Pichai describes an artificial intelligence system that can Return a pair of shoes on your behalf On stage at Google I/O. At Microsoft, the company announced Copilot, an artificial intelligence system that can act independently like virtual workers. At the same time, OpenAI launched GPT-4 Omni, an artificial intelligence system that can see, listen and speak. Previously, OpenAI CEO Sam Altman told MIT Technology Helpful agents have the best potential of the technology. These types of systems are the new benchmark that all AI companies are trying to achieve, but that’s easier said than done.
Simply put, an AI agent is just an AI model that does something independently. just like jarvis iron ManTar comes from room installer, or HAL 9000 from space odyssey. They go one step further than just creating a response like the chatbots we’re familiar with – there are actions. First, Google, Microsoft, and OpenAI are trying to develop agents that can handle digital operations. This means they are teaching the AI agent to use various APIs on the computer. Ideally, they can push buttons, make decisions, autonomously monitor channels and send requests.
“I agree that the future is intelligent,” said Alexander Kvamme, founder and CEO of Echo AI. His company builds artificial intelligence agents that analyze businesses’ conversations with customers and provide insights on how to improve the experience. “The industry has been talking about this for years, but it hasn’t happened yet. It’s a really hard problem.
Kvamme said a true agent system would need to make dozens or hundreds of decisions independently, something that would be difficult to automate. For example, to return a pair of shoes, as Google’s Pichai explains, an AI agent might have to scan your email for a receipt, pull your order number and address, fill out a return form, and complete various tasks on your behalf. kind of operation. During this process, there are many decisions that you don’t even think about, but you make subconsciously.
As we have seen, large language models (LLMs) are not perfect even in controlled environments. Altman’s favorite new thing is calling ChatGPT “Incredibly stupid,” He’s not entirely wrong. When you ask LL.M.s to work independently on the open web, they’re prone to making mistakes. But that’s exactly what countless startups, including Echo AI, and major companies like Google, OpenAI, and Microsoft are working on.
If you can build an agent digitally, there’s not much of a barrier to creating an agent that also works with the physical world. You just program the robot with the task. Then you really get into the stuff of science fiction, as AI agents offer the potential to assign tasks to robots, such as “take orders for that table” or “install all the shingles on this roof.” We’re still a long way from that goal, but the first step is teaching artificial intelligence agents to complete simple digital tasks.
There’s an oft-discussed issue in the field of artificial intelligence agents: Make sure you’re not designing the agent to perform a task also Excellent. If you set up an agency to return shoes, you have to make sure that it doesn’t return all your shoes, or possibly all the items you receive in your Gmail inbox. As silly as it sounds, a small group of artificial intelligence researchers worry that too determined an artificial intelligence could spell doom for human civilization. I guess that’s a legitimate concern when you’re constructing science fiction stuff.
On the other hand, optimists like Echo AI believe the technology will empower people. This divide in the AI community is stark, but optimists see the liberating effects of AI agents as comparable to those of personal computers.
“I firmly believe that a lot of work [agents] We’re solving jobs that humans don’t want to do,” Kvemme said. “There are higher value uses for their time in their lives. But again, they have to adapt.
Another use case for AI agents is self-driving cars. Tesla and Waymo are currently the leaders in this technology, with cars using artificial intelligence to navigate city streets and highways. Although self-driving technology is a niche field, it is a fairly developed field of artificial intelligence agents, and we are already seeing artificial intelligence operating in the real world.
So what could get us into a future where artificial intelligence can return shoes? First, the underlying AI models may have to get better and more accurate. This means that updates to ChatGPT, Gemini, and Copilot may occur before a fully functional proxy system is available. AI chatbots still have huge hurdles to overcome hallucination problem, many researchers have not found a solution. But the proxy system itself also needs updating. Currently, OpenAI’s GPT store is the most mature effort to develop a proxy network, but even that is not very advanced yet.
While advanced AI agents are certainly not here yet, this is the goal of many AI companies large and small today. This could make artificial intelligence more useful in our daily lives. Although it sounds like science fiction, we’ve spent billions of dollars making agents a reality in our lifetimes. However, it’s a tall order for AI companies that have been struggling to get chatbots to reliably answer basic questions.
