AI: three predictions for 2024

In recent weeks, I've read numerous articles about 'AI predictions for 2024,' many featured in famous publications, others here on LinkedIn. However, I couldn't help but notice a trend: many of these forecasts seem almost too evident to be labeled as 'predictions.' Statements like 'AI will increasingly influence various industries in 2024,' 'a growing number of jobs will require AI proficiency,' and 'multimodal AI solutions will become more widespread' are indeed likely, but they're so predictable that calling them 'predictions' seems a bit of a stretch, at least from my personal perspective.

Based on this, I've decided to share my own predictions for AI in 2024. Please note that these are purely personal conjectures. Like anyone else, I'm “extremely open” to the possibility of being mistaken. In fact, as someone who enjoys making predictions, I'm probably wrong more often than most people. To limit the risks, I've chosen to make only three top predictions, so at worst, I may be wrong three times! ??

Nevertheless, I'm eager to hear your perspectives on these predictions. Even more, I would like to know your own predictions connected to AI for 2024. Thanks in advance for your feedback, sharing, and insights!

?

Prediction 1: Beyond Human Language: The Dawn of AI Worlds Models

Reflecting on the history of Generative AI, it's clear that LLMs like ChatGPT and Bard have been initially predominantly focused on text processing. As groundbreaking as they are, their scope has been largely confined to linguistic tasks (they are correctly named “Large Language Models”). As AI evolves to integrate multimodal domains – encompassing text, images, audio, video, and physical and scientific interactions – the term 'LLM' (Large Language Model) seems increasingly inadequate

I expect that we will soon observe a pivotal shift in both terminology and AI solutions, that will reshape our understanding of AI. This transformation, which I firmly believe will define 2024, moves us from the realm of Large Language Models (LLMs) to an era dominated by what I propose to call "Worlds Models". This shift is more than a mere linguistic update; it represents a profound conceptual revolution, signaling AI's progression into a more integrated and globally nuanced era.

Worlds Models will not just be adept at processing a variety of data inputs – they transcend traditional boundaries. Imagine an AI model designed for drug discovery, analyzing molecular structures. This model isn't just a language model; it's venturing into the realm of biological linguistics. In chemistry you may use the chemical components as “words” for the WM, and this extend to all the reals of science. Here, the term WM becomes much more fitting, encapsulating the model's broader capabilities. The potential applications of WMs are incredibly exciting. In healthcare, WMs will revolutionize personalized medicine by interpreting complex genetic data. In environmental science, they predict climate patterns by analyzing vast arrays of satellite imagery and sensor data. And in robotics, WMs enable machines to understand and interact with the world with an unprecedented level of sophistication, combining visual data, sounds, physical perception, and even non-human sensors like infrared, radars, thermic-scanners, and other IoT world data, for real-time decision-making. In short, our world will become the language of WM.

It is worth to notice that the advent of Worlds Models necessitates unprecedented interdisciplinary collaboration. Integrating expertise from fields such as linguistics, sociology, environmental science, and more is crucial to ensure that WMs are not just technically proficient but also contextually aware and sensitive to the nuances of various domains. This collaborative approach will be instrumental in developing WMs that are not only innovative but also responsible and reflective of the diverse world they are designed to serve. Transitioning to WMs will not be without its challenges. Issues like data privacy, computational resource demands, and inherent biases in AI models are significant hurdles. Tackling these challenges will require an increased collective effort from technologists, ethicists, and policymakers. WMs hold the promise of a more interconnected and intelligent world, but they also bring forth critical questions about data governance and AI ethics that we must confront directly.

?

Prediction 2: From Generative Models to Personalized Active Companions

As we progress into 2024, the landscape of Artificial Intelligence may evolve from being predominantly generative to becoming highly personalized and proactive in its approach. Generative Pre-trained (GPT) models, which have been the hallmark of recent AI advancements, especially in 2023, are now being integrated with user-specific data, leading to a new era of personalized AI experiences.

One of the significant strides in AI development is the emergence of agent-based models. These models are designed to interact with various interfaces such as your company assets, ticket buying platforms, and transportation apps, providing a more integrated and interactive experience. Although these models showed only tentative progress in 2023, they may demonstrate more convincing applications in 2024. For instance, tedious tasks like submitting insurance claims could be simplified using these agent-based AIs, showcasing their practical utility in real-life scenarios.

One issue is that ChatGPT and other generative AI tools don’t really know you. They can remember a few of your previous prompts, but these are quickly forgotten. Imagine a tool that really knows you – what you like, how you work, what you know and don’t know, what you’ve asked for in the past, and what you are likely to need in the future. An AI companion will know which app can solve a problem you have, then download and configure it for you. It will help you draft a business proposal, write an email, share a selfie, analyze your data, propose a training course, schedule a meeting, buy a train ticket, or even suggest a lifestyle change. These “AI companions will be designed to understand individual user preferences and behaviors. Unlike previous AI tools, which had a limited understanding of user context, these AI companions will be capable of offering personalized assistance. They promise to be more than just assistants; they aim to be proactive companions in managing various aspects of our lives.

Clippy didn’t survive and Siri, Alexa, and Cortana may have been sometimes disappointing. But these tools were powered by old AI technology. Generative AI companions will be able to predict what you want, find what you need, and generate new information to help you to live and work better. Want to plan a long weekend in Bergamo, Italy? It will suggest hotels to fit your budget, restaurants that serve food you like, and activities that are compatible with your lifestyle. Then, if you approve, it will scour the Internet for the best deals, negotiate pricing directly, and handle the booking for you, organizing your entire weekend based on your preferences and guidance.

This shift towards personalization, combined with the proactive agency of AI systems, will mark a new era where AI is not just a tool for generative tasks but a personalized active companion aiding in various aspects of life. While challenges like privacy concerns remain, the benefits offered by these advanced AI systems may lead to their widespread adoption in 2024.

?

Prediction 3: the year of Deep Fakes

My third prediction for 2024 revolves more around a pressing concern I hold rather than a mere forecast. The year 2024 might be marked as the year of the Deep Fake, a technology that's been in the backdrop but is poised to surge to the forefront during the next year due to its improved accessibility, quality, and potential consequences.

The tools required for creating deep fakes are becoming increasingly available and affordable. Generative AI, which is highly effective in this domain, is a key factor. For instance, AI can now generate emails in any style with remarkable accuracy. It's becoming challenging to discern whether a message is genuinely from a person or an AI-generated facade. This uncertainty has even led some in tech circles to resort to writing in lowercase without punctuation, as a means to distinguish human authorship.

The real advancements in 2024, however, are expected in the realms of audio and video. Voice cloning technologies, like those from many start-up companies, have reached a point where they can replicate voices with astonishing precision. This advancement threatens the reliability of voice-based security measures and voice mails. You could easily use this technology to clone your own voice, allowing you to narrate written content effortlessly. Yet, this convenience comes with a downside – in the wrong hands, this could lead to misrepresentation.

The video domain is even more transformative. Combining cloned voices with video will lead to highly convincing fake footage of individuals saying or doing things they never did. These videos are becoming more realistic, accessible, and inexpensive to produce. Consider the havoc that could be wreaked by a fabricated video showing a prominent CEO in a compromising situation or an executive making damaging admissions about their products. This could escalate corporate espionage and blackmail.

The impact extends to the political arena. In 2024, key global elections are scheduled in countries like the United States, India, Indonesia, Mexico, Russia, and the UK. I’m highly concerned about the impact deep fake videos may have on influencing these elections. The widespread distribution of deep-fake videos could create uncertainty about the authenticity of information, potentially swaying election outcomes.

The repercussions are vast, ranging from election fraud and corporate blackmail to market manipulation and cybersecurity incidents involving deep fakes. A key to discerning the authenticity of content might lie in small details, such as the appearance of ears and fingers. As a response, a specialized industry might emerge, focused on developing AI tools to detect content created by other AI technologies.

I have chosen to share the three AI predictions for 2024 above, as I believe they encompass broad and potentially life-impacting developments for many of us. There are numerous other advancements that AI will bring in the new year. These include mainstream applications integrating AI, a move beyond the Search and Generative dichotomy, significant progress in simultaneous translation, and a variety of industry-specific and function-specific new applications of AI.

I will post links to some of the most interesting predictions I've found on the Internet, covering these areas and many others, in the comment section of this LinkedIn article. Additionally, for those more passionate about AI technologies, I'll share articles discussing intriguing technology predictions, such as LinWindow, AILF, InterAI, NextLM, AI FullReasoning, QuanTransformer, and more.

I look forward to reading your predictions and comments, either in the comment section of this article or wherever you prefer.

Wishing you a wonderful end to the year and an even better 2024. Always remember that, when it comes to predictions, "The best way to predict the future is to create it." ??