Computex 2024 may as well be called AI-tex by this point, with new chips from Intel, AMD and Qualcomm driving AI performance forward and huge software upgrades taking advantage of that extra headroom (such as Nvidia's Project G-Assist).
But there is one key goal that everybody is sprinting towards — fully autonomous agent AI that is capable of helping you in all of your day-to-day tasks and take action on your behalf, alongside always learning and improving itself.
And after speaking to AMD’s Senior Technical Marketing Manager, Donny Woligroski, we know the industry is predicting this will become a reality in “3-5 years.”
What does an AI Agent look like?
AI Agents are already sort of a thing, but in very limited ways. For example, email spam filters are agents acting on your behalf, or in-game AI that takes action to assist or compete against you.
What we’re talking about here is something deeper. Something more akin to to an assistant like you see in the film “Her” (without copying Scarlet Johannson’s voice of course), which can learn to assist you with all your computing tasks.
“If I have a laptop, and I want to project wirelessly to the TV right now, how do I do it? Well, I probably Google how to do that, and I'll probably get some instructions. Then I go into the system settings, set these things and then turn it on.” Woligroski explained.
But with a fully-fledged AI assistant, it's a different story. “The idea would be to ask the laptop 'please could you project wirelessly to my television that's in front of me?' And it would do it."
So far, we’ve got some great multi-modal options like GPT-4o and Gemini Ultra, but they all stop just short of being fully agent-like like this. That is the next evolution, and the entire industry is “working towards the end goal” in Woligroski’s own words.
Let’s get in formation
“I've never seen the industry move together in lockstep. Like it's really, really out of the box,” Woligroski commented. “That means that everybody sees the potential, and everyone's gonna move toward that potential.”
It shows that everybody has spotted something big here, and every company is marching towards this AI promise of giving your own hardware agency to be the best assistant possible for you. That’s going to be huge for accessibility to tech too, or as Woligroski puts it, “it removes barriers between a non-technophile and what they want to do.”
Playing the waiting game
The Rabbit R1 promised us something like this with the Large Action Model (LAM), and let’s be honest, it hasn’t turned out to be anywhere near that promise (yet at least). But Donny did have something to say about this when asking him the simple question: when?
“The most exciting thing about this whole thing is that 10 years ago, if you watched the Star Trek episode, where Captain Picard talked to a computer and says “postulate this thing for me,” and it did, that was pure fantasy!” Woligroski said, getting good and nerdy with me.
But now with LLMs being very much a thing now, and multi-modal AI becoming the talk of the town, we’re moving at a breakneck pace towards that being a thing.
“So when is it going to happen? I mean, I think as little as three to five years if this pace keeps up? I think the hard part is for all the providers to catch up with the functionality,” Donny predicted.
Mark your calendars, in around half a decade, we can see full-blown AI agents really become a thing.
More from Tom's Guide
- Computex 2024 LIVE — Testing Asus ROG Ally X, Nvidia Project G-Assist and AMD claims full AI Agent possible “in 3-5 years”
- It looks like Google is backpedalling on AI Overviews in Search results — just days after defending them
- ChatGPT went down twice in massive outage — here's what OpenAI says
