Sign up for our daily and weekly newsletters to get the latest news and exclusive content on leading AI developments.
By any measure, 2023 was an extraordinary year for AI. Large Language Models (LLMs) and their chatbot applications were the main attractions, but there were notable advancements in image, video, and voice generation. These technologies have created new business models and use cases, including digital humans becoming common as influencers and newscasters.
Importantly, 2023 saw many people start using AI intentionally in their daily work. Rapid innovation in AI has sparked predictions for the future, from friendly home robots to achieving artificial general intelligence (AGI) within a decade. However, progress can be unpredictable, and challenges may delay some of these forecasts.
As AI becomes more integrated into our daily lives, what’s next?
Physical Robots Could Arrive Soon
While digital advancements continue to amaze us, physical AI, particularly robotics, isn’t far behind. LLMs, especially when combined with image recognition via cameras, could give robots the “brain” they need to better understand and respond to their environment.
In a report, Nvidia’s VP of robots and edge computing, Deepu Talla, mentioned that LLMs will help robots better understand human instructions, learn from each other, and comprehend their surroundings.
One way to enhance robot performance is through multiple models. MIT’s Improbable AI Lab has created a framework using three different foundation models tuned for language, vision, and action tasks. These models together improve robots’ decision-making processes.
Yet, making robots practical for everyday use still has hurdles. To address these, Stanford University developed a new AI system called Mobile ALOHA. This system enables robots to autonomously perform complex tasks like cooking, opening cabinets, and using faucets.
An ImageNet Moment for Robotics
Jack Clark pointed out in his ImportAI newsletter that robots might be approaching their “ImageNet moment”—a turning point where the cost of learning robot behaviors decreases, as does the data needed for these behaviors. ImageNet is a large dataset of labeled images that advanced computer vision and deep learning. In 2012, a deep learning model drastically improved image classification, marking the start of modern AI. Clark believes we could be nearing a similar breakthrough for robots, envisioning a future where they assist in homes, hospitals, factories, and stores, effortlessly managing household chores.
The Pace of AI Advancement is Breathtaking
Many such pivotal moments could be on the horizon. Nvidia CEO Jensen Huang recently stated that AGI might be achieved within five years. Jim Fan, senior research scientist at Nvidia, compared the past year’s AI progress to leaping from the Stone Age to the Space Age.
McKinsey estimates that generative AI will add more than $4 trillion annually to the global economy. UBS predicts the AI market will grow from $2.2 billion in 2022 to $225 billion by 2027, a 152% compound annual growth rate. Enthusiasm for AI’s potential to enhance our lives is high, with Bill Gates noting that AI will supercharge innovation. A New York Times article cited David Luan, CEO of AI start-up Adept, highlighting the inevitability of AI’s rapid advancement.
Given these developments, it’s no surprise that generative AI is at the peak of inflated expectations according to the Gartner Hype Cycle for Emerging Technologies.
Is AI Progress Inevitable?
As we celebrate AI’s achievements in 2023, we must also consider the potential challenges ahead. The momentum behind AI is unparalleled, similar to the Internet boom of the dot-com era—and we remember how that turned out.
Might a similar situation occur with AI in 2024? A Fortune article suggests there might be a retrenchment as investors realize some companies lack viable business models and large firms find computing costs outweigh benefits.
This aligns with Amara’s Law, which states that we tend to overestimate a technology’s short-term impact but underestimate its long-term effects. Past AI advancements led to “AI winters,” periods of stagnation due to unmet promises, occurring from 1974 to 1980 and again from 1987 to 1993.
Not All Rainbows and Unicorns
Currently, we’re enjoying an “AI summer,” but there’s a risk of another winter. Challenges include computing costs, energy use in AI training, and sustainability issues. Additionally, data bias, security, copyright infringement, and AI “hallucinations” pose significant concerns.
The biggest worry is the potential existential threat from AI. While some see AGI as a path to abundance, others fear it could lead to humanity’s destruction. A survey of over 2,700 AI researchers showed significant concern about advanced AI leading to human extinction.
A Balanced Perspective
These issues serve as a brake on AI enthusiasm. However, momentum continues with predictions for further advances in 2024. The New York Times highlights the rapid technological improvements enabling new media forms, human-like reasoning, and advanced robots.
Ethan Mollick, in his blog, anticipates that AI development will continue accelerating before eventually slowing due to technical, economic, or legal limits.
The coming year in AI promises dramatic changes, potentially improving our quality of life with advancements like new life-saving drugs. While the highest expectations may not be met in 2024, leading to some market disappointment, it’s the nature of hype cycles, and hopefully, it won’t cause another AI winter.