As we get to the end of a momentous year for AI, we can go back and look at some of the biggest trends shaping up to disrupt business, and in all actuality, the rest of our lives as well.
Let’s take a look back at some words of wisdom from none other than Lex Fridman, who visited us by teleconference (or more accurately, recorded message) at this year‘s IIA event. Now, keep in mind, THIS WAS SIX MONTHS AGO, but some of what he said then has resonated throughout the year and we are making brilliant, sudden, exquisite progress in these areas!
Anyway, I was thinking of some of his remarks: the speech six months ago wasn’t a very long address, but he shared no less than five trends that he sees making waves in the tech world in 2024.
A Chatbot in Every Pot
It’s worth reusing this headline, in talking about the power of personalized LLMs.
There’s the emerging idea that, with the new hardware and neural net designs that we now have at our fingertips, we’ll be able to run decent LLMs on a normal endpoint device.
This was also first in the lineup of Fridman’s remarks, and since he made this prediction, we’ve actually been making a lot of progress on neural network types that can process information at the edge of a network, rather than sending it through the cloud.
This reverses a decade-long trend in centralization for data crunching, and it’s something that many of our top people are looking at closely. If the average person can carry around this kind of power with them wherever they go, that’s a game-changer.
AI and Translation
Fridman’s second focus was on the power of AI to break language barriers.
In the rest of the 21st century, we’re likely to see the painstaking work of personal language learning as pretty obsolete. People will basically be able to talk in a certain language, and render it in some other language, pretty effortlessly, with the capabilities of large language models in play.
It’s an interesting application that brings to mind the old story of the Tower of Babel, (with whatever you take from that tale) and represents one of many ways that AI is going to enhance human capability.
Re-Inventing Web Search
Web search, Fridman suggested, will also be transformed. “Google and Bing and every other search engine will rapidly evolve,” he said, talking about a new way of accessing the vast amounts of information on the global Internet.
It’s not hard to see the potential of this change. For those of us who were paying attention as the Internet matured, the first 20 years were the reign of keyword-related web search or web 1.0.
The new search will be much more intuitive and seamless, and the old paradigm will start to seem clunky and outdated.
Anyway, web search is another harbinger of the big changes that AI is going to bring, in this year and the years ahead.
Running With Humanoid Robots
Fridman also mentioned the rise of the robot – an interface that has a lot of potential to trick and astound humans, and to pass all manner of physical Turing tests.
“There are a lot of great companies working on this,” he said, encouraging people to look at the design of Atlas at Boston Dynamics.
So what is Atlas doing now?
As we quickly step up in interface design, robots are going beyond just moving and navigating physical spaces. They are learning to perceive what’s around them, to manipulate objects, and beyond that, to “feel” object attributes, to “think” about the objects that they come in contact with. You can see more at the BD lab page, with video of this kind of experimental work.
The Open Source Promise
“Open source is good for everyone,” Fridman said. “It’s a win-win.”
And certainly, there are a lot of us excited about the potential of open source design.
Along the way, we’re likely to have battles between walled garden licensed products and the open source model, just like we’ve had with so many other different kinds of technology tools. If open source wins out, it will foster more community collaboration and transparency about how we’re using these models.
That’s a little about what Lex Fridman had to tell us back toward the beginning of the year, and how those trends shook out as this year progressed. Keep an eye out in 2025 for much more new progress in exploring what we can do with rapidly evolving artificial intelligence.
