Building AI Agents To Work for You at COMPUTEX
A new phase of artificial intelligence is moving beyond chatbots. Developers at GTC Taipei are teaching people to build AI agents that work for them, acting as personal assistants and performing tasks. But how much control are you willing to give them?
REPORTER:
Agentic AI: It’s the biggest buzz word in tech right now.
I'm here at one of the world’s largest events for AI – COMPUTEX – stepping inside a live demo on OpenClaw.
People here say you don’t need to be a computer programmer or a software engineer. You just need to start building.
REPORTER:
Lobsters... not just a sea creature. This is also what open-source AI agent Open Claw has adopted as its mascot.
This has led to a trend called “raising lobsters” - training these AI agents to act as personal assistants and perform tasks.
REPORTER:
So, what exactly is OpenClaw?
Unlike traditional chatbots, like ChatGPT, that wait for you to type a question, AI agents move on their own. They work in the background, interacting with different apps on your computer.
They can help you summarize your meetings, send messages, book flights, and more.
Raymond Lo (DEVELOPER ADVOCATE MANAGER, NVIDIA):
You can actually remote access your browser and do tasks for you.
Think about booking your flights.
Maybe find a flight from Taiwan to Hong Kong.
And the date is from June 6 to 10.
Same thing, talk to it like a human.
If you look at the browser here, it’s actually going to the website, entering all the information you need.
Now, I'm NOT telling you to install OpenClaw.
What I am saying is: this is just the beginning. There will be more agents like this, showing up in your work and life very soon.
REPORTER:
Some companies are already experimenting with the technology and working to tackle emerging security concerns over these platforms’ access to accounts, credit cards and personal data.
Mark Heaps (DIRECTOR OF DEVELOPER COMMUNITY, NVIDIA):
We've introduced a new open-source project called Open Shell as a part of Nemo Claw, where you can write policies that are applied to your agent. So even if the agent has the skill of web search or operator on a browser, you might write a policy that says never click the buy button.
REPORTER:
OpenClaw says these safeguards can reduce risks in the short-term.
But here’s the catch:
To make it a true personal AI assistant, you don’t just give it tasks.
You give it access—to everything.
So the real question becomes: How much control are you willing to give it?
Ryan Wu and Lily LaMattina in Taipei for TaiwanPlus.
