In 2023, we saw AI break new grounds, evolving from novelty to necessity. From GPT-4 to LLaMA-2 to Gemini, AI's trajectory has been nothing short of revolutionary. Its rapid adoption, however, has not been without its share of concerns. Questions about ethics, data privacy, and the future of work persist as AI becomes increasingly ubiquitous.
Looking ahead to 2024, we're on the cusp of AI redefining industries and lifestyles yet again. To get a better look at those very trends, we tapped into the collective expertise of Appen’s executives and thought leaders to provide a glimpse of what’s to come as AI progresses in the coming year.
From Generalists to Specialists. The Theme of the Year Is Provenance.
Current Large Language Models (LLMs) have demonstrated a wide range of capabilities in natural language processing and understanding, performing tasks such as text generation, language comprehension, image creation, and more. However, these capabilities are classified as general skills applicable to all industries. In 2024, as AI continues to evolve to become increasingly specialized, the demand for AI will shift from generalists to specialists.
We expect to see a trend towards traceability to the highly skilled humans with specialized skillsets in industry verticals that contribute to these models. Their expert knowledge will be essential for the development of reliable AI solutions at both consumer and enterprise levels. It’s also likely that we’ll witness the emergence of new revenue models, such as IP royalty streams, for human intelligence embodied in AI implementations. Appen VP of AI Strategy, Mike Shwe suggests, for example, that we can “expect to see LLMs capture the expertise of IT specialists, in high-leverage, time-consuming tasks such as data migration and software integration.”
2023 marked a year of increased attention for AI regulation and safety; this trend towards human attribution aligns with the industry’s stride towards controlled development of Artificial General Intelligence (AGI). It’s a movement that champions human oversight in an increasingly automated world, ensuring, as machines get smarter at computing and making decisions, there are always humans in-the-loop, ensuring outputs remain ethical and harmless.
Knowledge Workers Will Evolve into Freelance AI Agents.
The job market is transforming significantly and how AI will impact that change is something hard for any of us to imagine. In 2023, we witnessed industry-wide disruptions and economic shifts. From tech to banking to the frontlines of healthcare, highly skilled workers are making career pivots.
According to the A.Team Knowledge Worker Survey, 74% of knowledge workers say that layoffs last year made freelance work more attractive. In 2024, we will see more and more knowledge workers seeking flexible roles as AI trainers, now that their skills are essential to training these systems.
This evolution heralds a broader change in the work landscape, underscoring the need for adaptability and a new definition of professional necessity. AI enables knowledge workers to obtain financial stability via multiple sources of income, and as more highly skilled knowledge workers become AI specialists, AI systems have the potential to become more powerful and grow in complexity. This expertise and adaptability will drive the refinement of AI technology, ushering in an era where AI solutions, fueled by expert specialists, are tailored to meet the most demanding industry needs.
Enterprises Will Stop Fearing AI, Thanks to Risk Management Frameworks.
So far, massive corporations, which tend to have more to risk, have been hesitant to integrate AI into their systems and processes. This is expected to change this year. The NIST AI Risk Management Framework, launched in early 2023, set the stage for widespread adoption by the private sector.
With regulation and normalization of AI, we’re already seeing its impact leading to more informed and comfortable interactions with the technology. According to a Microsoft & IDC Study on AI integration in companies:
- 71% of companies are already using AI.
- 92% of AI deployments are taking 12 months or less.
- 52% report that a lack of skilled workers is their biggest barrier to implement and scale AI.
As more companies integrate these risk management frameworks such as risk assessment, documentation, and continuous monitoring into their AI strategies, we can expect a surge in AI innovation, grounded in safety and propelled by trust. This alignment between innovation and risk mitigation promises, not only to advance AI's capabilities, but also to solidify its role as a driver of enterprise success.
The leading adopters of LLMs have already implemented company-wide review processes that customer-facing LLM deployments must follow before going into production. But there’s a potential conflict of interest in these internal review processes. In 2024, “You’ll see the emergence of third-party evaluation and auditing practices for LLMs, similar to what we experience with financial accounting and security reviews,” predicts Shwe. He adds, “Government advisories and ultimately regulations will develop concomitantly with these third-party services.”
There Will Be an Increased Focus on Multimodal AI Systems.
We’ve already seen great optimism in the capabilities of large language models, which have, indeed, demonstrated more ability than anyone expected. What will come next is an intense race to develop multimodal AI systems.
Hyperscalers and startups alike are in a technical race to lead the next revolution in large language model AI. These sophisticated systems, equipped with multi-sensory capabilities, stand at the threshold of a new era where AI begins to perceive the world resembling human experience—seeing, touching, and smelling.
The implications of this technological leap are profound. Large Multimodal Models (LMMs) are language models that integrate multiple data types, such as images, text, language, audio, and more. This technology is set to tackle challenges for specific real-world problems, such as assisting the visually impaired community—and that’s just the tip of the iceberg. Multimodal AI systems are setting new standards and redefining a variety of industries:
- In Automotive, manufacturers leverage technologies like image, video, LIDAR, and language data to detect driver fatigue, distraction, and attention loss.
- In Insurance, technologies are used to understand a large corpus of claims documents alongside image and video data, including claims adjuster notes, police records, images of damage, etc.
The potential is boundless, but so are the challenges. As AI's sensory horizon expands, the necessity for human involvement increases with it. It's the synergy between human oversight and AI's multi-sensory analysis that will ensure these systems stay relevant and attuned to the complexities of real-world applications.
The Future is Transformational.
Looking into 2024, our journey reveals pivotal transformations. From knowledge workers redefining their roles within the AI ecosystem to executive boards closing the AI skill gap, we are envisioning a dynamic yet revolutionary year ahead.
It’s important to note, all these evolutionary breakthroughs are useless if humans are not involved. At Appen, we place an emphasis on the role of humans and diverse representation in ensuring AI evolves to be harmless and equitable for all. It's a stance that celebrates the collaboration between human intellect and artificial intuition, reminding us that human ingenuity is critical in steering the course of AI towards a more ethical and sustainable future.