The year 2024 was an interesting one for Generative AI. It certainly wasn’t bigger than 2022 and 2023, when Generative AI went mainstream through the release of ChatGPT. On the other hand, it wasn’t a ‘nothing’ year as well. Sure, Generative AI is more widespread and prevalent than previous years. Helped through both Android and Apple deploying generative AI on mobile devices.
But, as I look back, I feel it was a year of consolidation. Consolidation in the big players, consolidation and confidence in the technology, and consolidation in regulation.
Simply, Generative AI Matures
Since the release of ChatGPT in late 2022, the business world has been transformed. Generative AI quickly became a force to be reckoned with. Adapt, or be forgotten. Industries were reshaped, peoples careers changed, or feared.
This year it was nearly impossible to avoid discussions about its influence—whether regarding ethical considerations, competitive performance, costs, or even the global GPU shortages caused by skyrocketing demand for AI training and deployment.
While 2024 did not deliver any paradigm-shifting innovations, it served as a litmus test for the capabilities and limitations of generative AI. Businesses and researchers spent the year exploring what these models could—and could not—achieve, creating a more nuanced understanding of their role in society.
The Open Source Surge
Open-source AI saw significant strides in 2024, narrowing the performance gap between freely available models and proprietary ones. Open-source models have proven indispensable to democratising access to cutting-edge technology, empowering startups, and fostering community-driven innovation.
One of the more notable releases of the year was Meta’s Llama 3.2, a colossal 405-billion-parameter model. By nearly all industry benchmarks, it matches the performance of OpenAI’s GPT-4o. Llama 3.2’s release reinforced the idea that open-source projects can achieve remarkable results, but scaling, fine-tuning, and deployment remain areas where closed-source giants hold distinct advantages.
This huge 405 billion parameter model requires large scale compute - that is simply unattainable for individuals and companies. Advancements in the smaller parameter models have been huge. Take, Llama 3.2 8B, outperforming the previous generation of Llama 2 70B, a model 62 billion parameters smaller outperforming the larger.
Smaller models lower the barrier to entry, enticing new players to the market and new opportunities - such as LLMs on edge.
The EU Leads on AI Regulation
In May 2024, the European Council formally approved the EU AI Act, a landmark piece of legislation that had been in development since 2021. As the world’s first comprehensive framework for regulating artificial intelligence, the AI Act underscores the EU’s commitment to ethical and transparent AI development.
The Act places stringent guardrails on AI systems deemed high-risk. For example, AI technologies used for social scoring—reminiscent of China’s controversial system—are outright banned. Meanwhile, other applications, such as biometric surveillance, must adhere to rigorous oversight and ethical guidelines. While critics argue that such regulations could stifle innovation, proponents maintain that these measures are crucial for protecting individual rights and fostering trust in AI systems.
The EU’s regulatory approach starkly contrasts with the United States and China, where industry-led initiatives and state-driven projects dominate, respectively. This divergence raises questions about the long-term impact of differing regulatory philosophies on global AI competitiveness.
This is reflected in Chinese company Alibaba being the largest corporate open source contributor on Hugging Face, and their flagship open source LLMs and vision LLMs, Qwen - family achieving outstanding results on public leaderboards.
The Premium Era of LLMs
One of the boldest business moves of 2024 came from OpenAI, which introduced a $200 USD monthly Pro subscription for its services. This shift—toward premium pricing for advanced AI capabilities—seemed to defy the prevailing trend of making AI cheaper and more accessible.
The Pro subscription offers a suite of advanced tools and resources, catering to enterprises and power users who require top-tier performance and features. This strategy reflects a broader industry trend: bundling multiple models and tools into cohesive packages, allowing companies to generate recurring revenue while delivering tailored solutions for specific use cases.
At the same time, smaller AI providers and open-source communities have worked to offer affordable or even free alternatives. These divergent strategies—premium services versus democratized access—illustrate the ongoing tension between monetization and accessibility in the AI industry.
Looking Ahead
As 2024 comes to an end, the AI industry is at a fascinating juncture. Ethical concerns persist, and the path to achieving significant performance improvements remains uncertain. The focus is no longer solely on when AI models will improve but increasingly on how they will improve—a shift underscored by Ilya Sutskever’s critique of current training paradigms.
Meanwhile, as frontier companies have already deployed advanced sector-specific solutions, 2025 is likely to see slower-moving organisations adopt AI technologies to remain competitive. With innovation and regulation advancing on separate fronts, the coming year promises both challenges and opportunities for stakeholders across the AI ecosystem. What lies ahead for AI in 2025? Will we see breakthroughs that redefine the industry, or will it be another year of consolidation and refinement? One thing is clear: the journey of artificial intelligence is far from over, and its implications will continue to shape our world for years to come.