Industry executives and experts share their predictions for 2024. Read them in this 16th annual VMblog.com series exclusive.
3 high-impact AI trends to follow in 2024
By Madhukar Kumar, CMO, SingleStore
As witnessed by the end-of-year tug of war between OpenAI and its board, 2023 has been the year of AI advancements and newfound digital transformation. But as organizations work to rapidly embrace generative AI tools and the possibilities these solutions unlock, a new question emerges: What does the next frontier for AI innovation hold?
It’s the million (or billion) dollar question. In the wake of generative AI’s breakout year, the spotlight turns toward how organizations can keep pace with rapid AI innovations and benefit from the smarter technologies at their disposal. As the CMO of a company firmly at the forefront of today’s generative AI revolution, here are the top three AI questions I’m looking to answer in 2024 – and the trends I’m keeping a close eye on to do so.
1. Is real-time data a passing fad?
Not in the slightest. Throughout 2023, companies across industries faced increased pressure to generate and make actionable real-time data at the enterprise level. While for some, the emphasis on real-time data may feel like it bubbled up out of nowhere, the call for rich, trusted and useful information has been a long time coming – and will only grow more important for companies in the years to come.
The rise of real-time data is largely a response to the increased pressure organizations face to swiftly respond to queries, often within milliseconds.
The ability to do so for both customers and employees hinges on a data plane capable of comprehending, analyzing and contextualizing information in real time. Real-time data integrations will prove essential for the majority of generative AI applications and services we’ll see at market in 2024, with most technologies adopting a Retrieval Augmented Generation (RAG) pattern to support. Vectors will also rise as a fundamental requirement across all AI databases. And we’ll see more organizations adopt hybrid searches (e.g., combining keyword matching with vectors) as a standard practice – particularly those dealing with extensive datasets aimed at developing real-time AI applications.
Additionally, as regulatory efforts intensify across AI spaces and data sources, we’ll see a surge in advancements concerning RAG data management and retrieval patterns. This innovation may manifest in enhanced controls and visibility over data provenance fed to AI tools, along with advancements in data access, governance, audit and security. Each of these factors must scale alongside the massive amounts of data required to popularize AI solutions – it’s an element of the AI world we can’t underestimate or leave for later.
2. Will Large Language Models (LLMs) grow in sophistication?
Yes, and in more ways than one. In 2024, I expect we’ll see the rise of more sophisticated LLMs, particularly in the open-source community – and most likely released by large, resource-rich companies like Meta and Microsoft. What is also more likely to show up is the use of an ensemble of LLMs to perform more complex tasks and use cases.
The projection starts with the reality that LLM technologies are decreasing in physical size, opening the door to broader applications. We’re seeing quick improvements of the traditional GPUs found in our laptops and desktops, such as Apple’s M3 chip (which now features dynamic caching), and the upcoming launch of Intel’s Meteor Lake Chip. As LLMs become more pocket-size, we’ll see a diverse spectrum of companies integrate the technology into products like wearable devices, like Meta’s Ray-Ban glasses and watches.
More compact LLMs will naturally usher in substantial enhancements in both response quality and speed. For instance, more seamless integration of LLMs across wearables will enable these devices to support question-answering capabilities akin to what we already experience with ChatGPT online. Companies will achieve this dynamic through a network of LLMs designed to handle the most intricate use cases, incorporating contextual data through both audio and video inputs. As LLMs grow more mainstream, expect a proliferation of specialized LLMs tailored to specific industries and use cases, mirroring initiatives like BloombergGPT (an LLM trained on a wide range of financial data). We’ll also likely see increased opportunities for LLMs to assist humans in core tasks, such as content generation or customer support, rather than deploying fully autonomous systems.
Interestingly, with sophistication will come mass market appeal. The evolving sophistication and proliferation of LLMs open the door to an emergence of commercial LLM solutions, designed for mini-AGI frameworks and to help users independently perform complex tasks without the need for multiple interactions. The AGI systems will also be most likely driven by an agent-oriented architecture, in which each agent will have its own custom instructions, custom knowledge (data) and custom tools. We will also see a lot more rapid innovation around frameworks to orchestrate autonomous tasks among agents.
3. What does this mean for AI startups?
Unfortunately, the answer to this question is a bit harder to predict. What I do anticipate is a world in which several major players in the AI space incorporate more diverse and nuanced functionalities into their 2024 products, alongside the core capabilities they’re already working on. This evolution could make some of the niche and specialized apps developed by startups in 2023 redundant.
But I don’t think that means the end of companies that have quickly emerged in the generative AI boom. Rather, the ongoing success of these startups will rest on the quality of their finely tuned responses, built on top of more sophisticated LLMs. For example, the precision of responses capable from startups like MidJourney might stand out in quality and usefulness when compared to a platform like OpenAI’s DALL-E. In the face of larger corporations building out their own generative AI capabilities, the answer for 2023’s startups may be to go even more niche in 2024.
I also expect we’ll see an emergence of startups in the realm of AI and mixed reality. Mixed reality merges the physical and digital worlds, offering immersive experiences by overlaying virtual objects onto real environments. It’s an exciting frontier for organizations, and the future of mixed reality will be built on the back of AI-based tools capable of incorporating robotics, graphics, machine learning (ML) and other cutting-edge technologies.
From more sophisticated LLMs to the rise of real-time data, today’s AI landscape continues to evolve at a remarkable pace. In the wake of generative AI’s rapid ascent in 2023, it’s time to look ahead to the strategies organizations must adopt to fully harness the innovations poised to define the AI realm in 2024.
##
ABOUT THE AUTHOR
Madhukar is the CMO of SingleStore, as well as a developer, turned growth marketer and an expert in Product-Led Growth (PLG) with 18+ years of experience leading product management and marketing teams. He has successfully implemented PLG at Nutanix, Redis and DevRev, and is a guest lecturer at Duke University on PLG and New Product Development.
