The year 2024 has marked a significant upsurge in artificial intelligence systems, which are infiltrating nearly every aspect of daily life, from the mundane to the groundbreaking. Since AI technologies continue to advance, the vocabulary surrounding them is evolving, reflecting our growing understanding of these advancements, and the constant need to keep up with all the new terms.

Here is a glimpse into ten of the most pivotal AI terms that have not only defined discussions in technology, policy, and everyday life this year but are also transforming industries and society at large and continuing to shape the future of artificial intelligence, which is likely steering the course for years to come.

1. Generative AI

The term generative AI has exploded in popularity, referring to AI systems that create new content across various mediums, including music, art, and text. Tools like DALL-E for image creation and Jukebox for music composition exemplify how generative AI democratises creativity, allowing individuals to produce content with minimal traditional skills required. This rise in generative AI's artistic capabilities sparks debates about originality, copyright, emotional depth, and the intentionality of human creators, alongside discussions on its role in creative industries.

2. Multimodal AI

AI is evolving beyond text-based applications to encompass multimodal AI, which integrates and interprets various forms of data. Models like Google's Gemini can read images, listen to audio, analyse video, and respond based on this combined input. This capability is particularly revolutionary in sectors like healthcare, where AI can interpret medical imaging while cross-referencing patient histories from text-based records.

3. Large Language Models (LLMs)

Large Language Models, like Google's PaLM and OpenAI's GPT-4, are at the forefront of AI conversation, capable of generating human-like responses, creating images, videos, and texts, translating languages, and assisting in coding. Their ability to mimic human conversational styles has made them indispensable tools in customer service, education, and content creation. Their impact is so significant that understanding LLMs is akin to understanding the internet in the early 2000s— a fundamental shift in how we interact with data-driven machines.

4. Small Language Models (SLMs)

While LLMs grab headlines, SLMs, like DistilBERT, offer practical solutions for devices with less computational power, making AI more accessible and reducing cloud computing dependency. They are particularly useful for real-time processing or limited-resource applications, enabling developers to create efficient, cost-effective AI solutions across various industries.

5. AI alignment

AI advancement necessitates alignment with human values and ethics to ensure safety, transparency, and beneficial use, thereby preventing biased or harmful decisions. The discussion on AI alignment is thus gaining momentum, particularly in governance and finance, as developers and policymakers strive to establish responsible AI development guidelines.

6. Synthetic data

Data privacy laws and the scarcity of certain datasets have pushed the development of synthetic data—artificial data created by generative AI models trained on real-world data samples. Such datasets mimic real data's statistical properties, which is particularly useful for training AI models in sensitive areas like healthcare or finance without compromising individual privacy. They can also improve AI systems by addressing biases in real-world data, but ethical concerns remain regarding its quality and reliability for AI development.

7. AI safety

AI safety is a critical area of focus that aims to understand the long-term implications and risks associated with advanced AI systems, including the potential for unintended consequences. This field encompasses a variety of concerns, from preventing accidents and misuse to addressing ethical implications and societal values, mitigating risks such as bias, lack of transparency, and existential threats posed by superintelligent systems.

In 2023, the Center for AI Safety (CAIS) released a statement advocating for global prioritisation of mitigating extinction risks from AI, comparable to other societal-scale risks like pandemics and nuclear threats. Subsequently, the USA and UK have collaborative efforts through their AI Safety Institutes to work seamlessly with each other, partnering on research, safety evaluations, and guidance for AI safety.

8. AI washing

Jailbreaking is a form of bandwagon jumping, similar to greenwashing, where projects or products exaggerate their environmental benefits to attract eco-conscious consumers. In the rush to capitalise on AI's allure, companies are engaging in AI washing, exaggerating or misrepresenting AI capabilities in their products, causing a consumer awareness movement for transparency and accountability in AI marketing.

This phenomenon has sparked a consumer awareness movement, urging for more transparency and accountability in AI marketing, which has led to calls for stricter regulations to prevent deceptive practices, as it risks damaging reputations and consumer trust.

9. Jailbreaking

Jailbreaking refers to the manipulation or bypassing of AI systems' built-in restrictions, allowing users to elicit responses or actions that the system is designed to prevent. This practice can lead to significant misuse, as evidenced by instances where chatbots provide inappropriate or harmful responses. The term highlights the ongoing cat-and-mouse game between AI developers, who aim to secure their systems, and those looking to exploit vulnerabilities for malicious purposes.

10. Anthropomorphism

Anthropomorphism involves attributing human-like characteristics to AI systems, which can lead to misconceptions about their capabilities and emotions. This tendency can result in users overestimating the understanding and emotional intelligence of AI systems, leading to unrealistic expectations about their performance and decision-making abilities.

Additionally, this phenomenon influences how people interact with AI, fostering a sense of trust or companionship that may not be warranted, complicating ethical considerations surrounding AI deployment in sensitive areas such as healthcare, education, and customer service.

The broader impact

AI challenges our notions of creativity and originality. The integration of AI in education, for instance, through LLMs, could redefine how we learn and teach. Meanwhile, the push for AI ethics and alignment invites everyone to consider what kind of future we want to build with AI.

Understanding tech terms thus goes beyond jargon; it involves recognising their societal impact and participating in discussions about AI's role. This understanding is crucial for educators, artists, and policymakers to discuss ethical and practical AI evolution.

Md Abdul Malek is a graduate student at the University of California, Los Angeles. He is reachable at abmalek@ucla.edu.