Learning robots, visualizing chatbots: how 2024 will be the ‘leap forward’ of AI
A brown color event in San Francisco in November, Sam Altman, the CEO of the artificial intelligence company OpenAI, was asked what surprises the field would bring in 2024.
Online chatbots such as OpenAI’s ChatGPT will make “a leap forward that no one expected,” Mr. Altman immediately responded.
James Manyika, a Google executive, sat next to him and nodded and said, “Plus one to that.”
The AI industry this year will be defined by one key characteristic: a remarkably rapid improvement in technology as developments build on each other, allowing AI to generate new types of media, mimic human reasoning in new ways, and penetrate the physical world. via a new kind of robot.
In the coming months, AI-powered image generators such as DALL-E and Midjourney will instantly deliver both videos and still images. And they will gradually merge with chatbots such as ChatGPT.
That means chatbots will go far beyond digital text by processing photos, videos, charts, graphs and other media. They will exhibit behavior more akin to human reasoning, tackling increasingly complex tasks in areas such as math and science. As technology transitions to robots, it will also help solve problems beyond the digital world.
Many of these developments are already visible in top research laboratories and in the field of technical products. But by 2024, the power of these products will increase significantly and be used by many more people.
“The rapid progress of AI will continue,” said David Luan, the CEO of Adept, an AI start-up. “It’s inevitable.”
OpenAI, Google, and other tech companies are advancing AI much faster than other technologies because of the way the underlying systems are built.
Most software apps are built by engineers, one line of computer code at a time, which is usually a slow and laborious process. Companies are improving AI faster because the technology relies on neural networks, mathematical systems that can learn skills by analyzing digital data. By identifying patterns in data such as Wikipedia articles, books, and digital text pulled from the Internet, a neural network can learn to generate text itself.
This year, tech companies plan to give AI systems more data – including images, sounds and more text – than humans can imagine. As these systems learn the relationships between these different types of data, they will learn to solve increasingly complex problems and prepare them for life in the physical world.
(The New York Times sued OpenAI and Microsoft last month for copyright infringement on news content related to AI systems.)
None of this means that AI will soon be able to match the human brain. While AI companies and entrepreneurs aim to create what they call “artificial general intelligence” – a machine that can do everything the human brain can – this remains a daunting task. For all its rapid progress, AI is still in its early stages.
Here’s a guide to how AI will change this year, starting with near-term developments that will lead to further advances in its capabilities.
Instant videos
Until now, AI-powered applications typically generated text and still images in response to prompts. For example, DALL-E can create photorealistic images in seconds based on requests such as “a rhino diving off the Golden Gate Bridge.”
But this year, companies like OpenAI, Google, Meta and New York-based Runway are likely to deploy image generators that allow people to generate videos as well. These companies have already built prototypes of tools that allow you to create videos on the fly from short text prompts.
Technology companies are likely to merge the power of image and video generators into chatbots, making the chatbots more powerful.
‘Multimodal’ chatbots
Chatbots and image generators, originally developed as separate tools, are gradually merging. When OpenAI debuted a new version of ChatGPT last year, the chatbot was able to generate both images and text.
AI companies are building “multimodal” systems, meaning the AI can process multiple types of media. These systems learn skills by analyzing photos, text, and possibly other types of media, including charts, graphs, sounds, and video, so they can then produce their own text, images, and sounds.
That’s not all. Because the systems also learn the relationships between different types of media, they can understand one type of media and respond with another. In other words, someone can enter an image into the chatbot and it will respond with text.
“The technology will become smarter and more useful,” said Ahmad Al-Dahle, head of the generative AI group at Meta. “It will do more things.”
Multimodal chatbots will get things wrong, just like text-only chatbots make mistakes. Technology companies are working to reduce errors as they strive to build chatbots that can reason like a human.
Better ‘reasoning’
When Mr. Altman talks about AI taking a leap forward, he is referring to chatbots that are better at “reasoning,” so they can perform more complex tasks, such as solving complex mathematical problems and generating detailed computer programs.
The goal is to build systems that can solve a problem carefully and logically through a series of discrete steps, with each step building on the next. That’s how people reason, at least in some cases.
Leading scientists disagree on whether chatbots can really reason like this. Some argue that these systems only appear to reason as they repeat the behavior they have seen in Internet data. But OpenAI and others are building systems that can more reliably answer complex questions in subjects like math, computer programming, physics and other sciences.
“As systems become more reliable, they will become more popular,” says Nick Frosst, a former Google researcher who helps run Cohere, an AI start-up.
If chatbots can reason better, they could turn into ‘AI agents’.
‘AI agents’
As companies teach AI systems how to solve complex problems step by step, they can also improve chatbots’ ability to use software apps and websites on your behalf.
Researchers are essentially transforming chatbots into a new kind of autonomous system called an AI agent. That means the chatbots can use software apps, websites and other online tools, including spreadsheets, online calendars and travel sites. People can then leave the boring office work to chatbots. But these agents can also take away jobs entirely.
Chatbots already act as agents in small ways. They can schedule meetings, edit files, analyze data, and create bar charts. But these tools don’t always work as well as they need to. Agents completely break down when applied to more complex tasks.
This year, AI companies will unveil agents that are more trustworthy. “You should be able to delegate all the tedious, day-to-day computer work to an agent,” Mr. Luan said.
This includes tracking expenses in an app like QuickBooks or recording vacation days in an app like Workday. In the long term, it will expand beyond software and internet services and into the world of robotics.
Smarter robots
In the past, robots were programmed to perform the same task over and over again, such as picking up boxes that were always the same size and shape. But by using the same kind of technology that underpins chatbots, researchers are giving robots the power to perform more complex tasks – including ones they’ve never seen before.
Just as chatbots can learn to predict the next word in a sentence by analyzing large amounts of digital text, a robot can learn to predict what will happen in the physical world by analyzing countless videos of objects being poked, lifted and moved.
“These technologies can absorb enormous amounts of data. And as they absorb data, they can learn how the world works, how physics works, how to interact with objects,” said Peter Chen, a former OpenAI researcher who leads Covariant, a robotics startup .
This year, AI will boost robots that operate behind the scenes, such as mechanical arms that fold shirts in a laundromat or sort piles of stuff in a warehouse. Tech titans like Elon Musk are also working on movement humanoid robots in people’s homes.