- OpenAI COO Brad Lightcap defined AI agents on a recent episode of the company’s podcast.
- Agents should perform tasks “autonomously,” even ones they’ve never seen before, he said.
- This week, OpenAI unveiled its latest agent, which works with ChatGPT to execute tasks.
There’s not a whole lot of agreement among AI leaders about how to define all the new things they are developing.
No one really agrees on what artificial general intelligence is, for instance. Microsoft CEO Satya Nadella recently called it “nonsensical benchmark hacking.”
There’s also debate about what exactly an AI agent is.
OpenAI Chief Operating Officer Brad Lightcap offered his thoughts on OpenAI’s podcast this week.
“It has to be a system that can be reliably handed complex work, that it can take on autonomously, and execute at a high level of proficiency, where it hasn’t seen that work before,” he said. The “hasn’t seen” part is “critical,” Lightcap added.
On Thursday, OpenAI unveiled its latest contribution to the space: ChatGPT Agent. The new agent “can now do work for you using its own computer, handling complex tasks from start to finish,” OpenAI said in its announcement. It combines the work of Operator to engage with websites, Deep Research to synthesize information, and ChatGPT’s conversational ability, the company said.
Companies are going all out to integrate agents into their work. Major tech firms from Intuit to Salesforce are rolling them out at light speed. Startups, too, are unveiling new agents.
The race to adopt AI has left workers worried about whether they’ll soon be replaced by a machine. However, at least in Lightcap’s view, humans are necessary at every level.
Agents aren’t just entities that are “trained to copy,” he said. They leverage their reasoning capacities to solve problems.
Still, it’s only a “teammate,” he said.
Read the full article here
