Top AI Trends - 2024
With the new year at a wonderful start and all of us caught up with 2023’s AI recap, it is time to explore the top trends that will shape AI’s future in 2024.
Here’s my list of the top AI trends for 2024 – and what makes them exciting!
Generative AI will get better and cheaper
Generative AI startups secured $27 billion in 2023, with major contributions from tech giants like Amazon, Microsoft, and Google. Enterprise investment in generative AI was roughly $2.5 billion, bringing the total investment to $29.5 billion.
With considerable hype around the subject and the release of a host of Gen AI models like Dall-E 3, Inflection-2, Claude 2.1, and Google Gemini, the anticipation for 2024 grows. We can look forward to successors to LLMs like Inflection-2 and the arrival of GPT-5 later in the year.
We can also expect a greater number of industry-specific LLMs similar to BloombergGPT (for financial services), Harvey AI and Robin AI (for legal services) and SkinGPT (for the cosmetics industry).
Smaller and more efficient generative AI models are gaining popularity due to the growing emphasis on optimization and resource efficiency. Take Microsoft's Phi-2, for example, a compact language model with 2.7 billion parameters that demonstrates "state-of-the-art performance" among models with fewer than 13 billion parameters. It even outperforms models up to 25 times larger on challenging benchmarks.
Multimodality will increase
Multimodal AI models are ones that can take in more than one type of input. For example, they can understand text, images, voice, and video, among other forms. 2023 saw an increase in multimodal AI models, including LaMDA and Dream by WOMBO, and of course, Google's Gemini.
Even ChatGPT became multimodal with users now able to upload PDF, images and other files in ChatGTP4. For example, you can upload an image of a recipe, and then request ChatGPT to produce a list of ingredients or cooking instructions based on that image.
Imagine the use cases this will open up in industries like education and healthcare. For instance, providing an AI model with an X-ray report and a list of symptoms can enable it to analyze the information and suggest potential diagnoses. Likewise, presenting a math problem with an explanation of the challenge allows the AI to offer personalized solutions. The possibilities are vast.
Autonomous agents will take a step in the direction of "baby" AGI
AGI (Artificial General Intelligence), sought after by tech giants like OpenAI and Google, is hailed by OpenAI's CEO, Sam Altman, as "humanity's most potent technology."
It resembles sentient intelligence, autonomously tackling tasks. In the upcoming year, expect advancement toward a "baby" version of AGI — autonomous agents. These AI models operate independently, making decisions without the need for human oversight.
Market projections predict a surge in autonomous agents value, from $5 billion in 2023 to an estimated $29 billion by 2028.
AutoGPT, AgentGPT, and BabyAGI are examples of autonomous agents already using AI to streamline operations. These programs can divide a large task into many little ones, complete them, create new tasks, and even reprioritize their workflow. Autonomous agents can be tailored to do almost any kind of work, be it investing in the crypto market, coming up with a chapter for a book, or even managing a social media account.
2024 will tip the start of autonomous agents, but it will take another 3-5 years before they become commonplace.
Supercomputers will get faster
In 2024, the supercomputing landscape is undergoing a major transformation with the rise of exascale machines. The debut of Frontier at Oak Ridge National Laboratory in 2022 marked a milestone, claiming the title of the world's fastest supercomputer, capable of performing over 1 quintillion operations per second—an exaflop.
This achievement signals the onset of the exascale computing era, with upcoming installations like El Capitan in California and Aurora in Illinois set to surpass Frontier's speed. Europe is also gearing up to join the exascale league with the launch of its first supercomputer, Jupiter, slated for late 2024.
Exascale computing's substantial computational capacity opens doors for more advanced AI models, driving progress in machine learning, deep learning, and neural network training. The speed and efficiency of exascale machines facilitate the swift execution of AI algorithms, reducing the time needed for model training and inference.
Yet, a critical challenge looms: addressing the considerable energy footprint of these machines. Balancing speed and environmental sustainability becomes paramount in the development of future supercomputers.
Robots will become smarter
In 2024, the convergence of generative AI, large foundational models, and robotics is poised to revolutionize the robotics field. With a focus on applications from learning to product design, teams like Google's DeepMind Robotics are at the forefront. Their AutoRT system, utilizing Visual Language Models (VLM) and Large Language Models (LLM), manages fleets of robots with cameras, offering enhanced situational awareness and the ability to execute diverse tasks.
DeepMind prioritizes safety with a "Robot Constitution," inspired by Asimov's laws, guiding robots to avoid tasks involving humans, animals, sharp objects, and electrical appliances. AutoRT, deployed in office environments, orchestrates up to 20 robots simultaneously, showcasing promising success rates in tasks after over 77,000 trials.
Another model, RT-Trajectory, leverages video input for robotic learning, providing 2D outlines to enhance specific physical tasks. These advancements signal a leap toward robots that can efficiently understand natural language commands and navigate novel situations.
A testament to this trend is Google DeepMind and Stanford University’s Mobile ALOHA, a human-like robot adept at household chores. With four color cameras streaming at 480x640 resolution, ALOHA learns tasks through imitation learning, demonstrating up to 50 activities autonomously.
These advancements signal a future where robots equipped with AI comprehend complex tasks, adhere to safety protocols, and scan become our helpers not only in work but at home in our daily life.
VR and spatial computing investing will spur the metaverse
In 2024, the metaverse is set to make a comeback with advancements in the field of VR and spatial computing that are taking place at a fast pace.
With advanced features and seamless integration, spatial headsets could bring the metaverse into the mainstream.
Apple's Vision Pro headset is anticipated to have a transformative impact on the AR/VR landscape, drawing parallels to the influence of the iPhone on smartphones.
Sony is also set to introduce a "spatial content creation" system with a new VR headset. Unlike PlayStation VR, primarily aimed at gamers, this device targets professionals working with 3D models and allows users to edit and shape 3D models in real-time, overlaying virtual objects into physical spaces.
Immersive technologies can aid in various fields, such as immersive learning in education and extended reality-assisted surgeries in healthcare.
Moreover, 2024 is likely to see further advancement in hybrid cities, aided by digital replicas. Following South Korea's metaverse city experiment in 2023, more countries are expected to delve into the development of hybrid cities using digital twins. Singapore and Barcelona are already utilizing digital replicas for improved city management. Seeing how these digital cities evolve to incorporate the general population and everyday activities will be exciting.
The AI-healthcare decade will kick off
We saw 2023 bring us a myriad of AI developments in healthcare applications.
Neko Health, Prenuvo, and Ezra are among the AI-driven services providing AI-powered full-body scans and MRIs that are faster, require less radiation exposure, and can be done in less stressful environments, such as while watching your favorite TV show. These scans aid in the early detection of various conditions, including skin issues, cardiovascular diseases, and metabolic syndromes.
AI plays a pivotal role in protein synthesis, exemplified by Meta's ESM-2 and Salesforce's ProGen. These AI models use extensive protein sequence databases to predict and generate sequences, potentially contributing to treatments for neurological and autoimmune diseases like rheumatoid arthritis.
The integration of CRISPR technology with AI enhances precision in gene editing, analyzing vast genetic datasets to optimize processes. In December, two cell-based gene therapies for people with sickle cell disease were approved by The United States Food and Drug Administration (FDA).
Next on everyone's mind: using AI for increase longevity.
We’ll become used to "talking" to smart devices equipped with AI
In 2024, the tech industry gears up for devices centered entirely around AI functionalities. Such as,
On-device AI: Samsung showcases built-in Gen AI in its upcoming phone models, such as the Galaxy S24. Gauss, Samsung's AI, integrates three powerful AI tools into mobiles. Gauss Text summarizes lengthy documents, Gauss Image enhances photos, and Gauss Code provides coding assistance, eliminating the need for external tools or browsers.
AI-wearables: We’ll be conversing with our very own AI assistant via a smart device: take the Humane AI Pin, for example. This wearable pin translates real-world gestures and context into data, allowing personalized responses.
Talking to AI: Conversing with AI, in general, will become more common. For example, ChatGPT recently added a voice feature, combining Gen AI with voice functionality. ChatGPT is now capable of holding verbal, human-like conversations with users.
A glimpse of the future - a new AI device: Industry titans Sam Altman and Jony Ive teamed up for a new AI hardware venture while enlisting the expertise of an Apple veteran, Tang Tan. Sources reveal that Tan will focus on designing the appearance and features of the upcoming products, with Altman in charge of the software. While details about the partnership are currently limited, the idea of an AI personal device in 2024 with a brand new design and interface makes it an exciting prospect to look forward to.
More countries will embark on AI regulations
The surge in AI brings inherent risks. Misinformation and deepfakes remain a rising concern in social media. Privacy risks, particularly with Generative AI, threaten personal data confidentiality.
Meta AI addresses this with 'invisible watermarks.' Amid AI's rampant growth, 2024 will witness a surge in demand for regulations.
While the EU made progress on the AI act and Biden issued an executive order, the push for more action remains a focal point in AI development for 2024. Regulations like advancement on Canada’s AIDA are expected in 2024, along with further progress on the EU AI Act.
AI regulation will affect industries like healthcare, transportation, finance, and employment. As these industries move forward, they will have to carefully navigate regulatory guidelines concerning privacy, bias, safety, fairness, and ethical considerations in the deployment of AI systems.
Here's a preview of the trends shaping 2024. As AI continues its rapid integration into multiple industries, the opportunities to innovate seem endless. One thing is for sure: AI will become ubiquitous, finding its way into every product, service, and major company we are familiar with.
