The 4 Biggest AI Stories of 2024 and one key prediction for 2025

By all criteria 2024. was the biggest year for artificial intelligence to date – at least as far as the commercialization of the technology is concerned.

The Large Language Model (LLM) boom sparked by the launch of ChatGPT in late 2022 has shown no sign of slowing down, with numerous new LLMs introduced not only by OpenAI and stalwart tech giants such as Microsoft, Meta and Google, but also from numerous other startups and individual developers.

Reports of a slowdown in AI research have proven, if not unfounded, then certainly exaggerated for now.

Additionally, new technologies have begun to emerge beyond the Transformer architecture that underpins most major LLMs, such as Liquid AI’s liquid-based models.

Finally, firms have begun to embrace a fully “agentic” approach to AI—developing specific AI-powered bots, applications, and workflows that can work on specific problems independently or with less human guidance than typical LLM chatbots .

Distilling the year’s news to a top 14, much less a top 10 or top 4, was a nerve-wracking effort. But I went ahead and tried, albeit cheating slightly, by combining several stories into larger threads. In my eyes, here’s what will have the biggest impact after this year’s rollout:

1. OpenAI has expanded far beyond ChatGPT

The company perhaps most responsible for ushering in the era of the AI ​​generation hasn’t missed a beat this year, despite intensifying competition from startups and legacy technologies, even from its own investor and partner Microsoft.

model o1: OpenAI released its first new family of large general-purpose models outside of the GPT series, the series o1 ‘reasoning’.which spends more time processing complex prompts resulting in higher accuracy. It is particularly effective in science, coding and thinking tasks.

o3 Model: It followed up September’s o1 model with a blockbuster year-end announcement more advanced o3 model. While this won’t be publicly available or even to third parties until early 2025, it shows that OpenAI isn’t resting on its laurels.

ChatGPT Search: This feature, originally launched as a stand-alone, invite-only product called SearchGPT before being collapsed into ChatGPT proper, it allows for more real-time web retrieval within ChatGPT and improved search results presentation, improving its usefulness for topical queries and going head-to-head with Google, Bing and newcomer Perplexity.

Canvas: Introduced in October, Canvas extends ChatGPT’s interface beyond that of a conversational one to a workstation-like window that can dynamically update content at the user’s request, such as editing a document or coding project. Of course, it was hard not to see it as a reaction to, or at least a comparable characteristic of Anthropic’s Artifacts announced a few months ago.

Sora: After almost a year of teasing us with your strictly guarded video generator model, OpenAI in early December finally released Sora to the massesquickly eliciting a wide range of reactions as it sought to differentiate itself in a highly competitive AI video space with a unique and well-thought-out interface and scripting feature.

2. Open source AI has taken off

Lama 3 and 3.1: Meta entered Llama 3 in Aprilsetting a new performance standard in open source AI, then quickly followed it up with Llama 3.1 in July with 405 billion parameters. The Llama 3.1 versions were used to power Meta AI, the company’s assistant integrated into platforms such as WhatsApp, Messenger, Instagram and Facebook, aiming to become the most widely used AI assistant.

Lama 3.3: Released in December 2024, Llama 3.3 delivered performance comparable to larger models but at a fraction of the computational cost, making it more affordable for enterprise applications.

Meanwhile, Chinese models like Alibaba’s Qwen-2.5 family and DeepSeek’s new V2.5 and R1-Lite preview came seemingly out of nowhere to top some of the benchmark charts, and Nvidia itself has gone beyond supplying graphics cards and software architectures to launch its own powerful open source Model Nemotron-70B.

Nous Research, a small team in San Francisco aiming to offer more custom and less restrictive AI models as open source, also debuted several cool new ideas.

And let’s not forget France Mistralwhich has rapidly expanded its own open source and proprietary AI offerings.

3. Google’s Gemini series has become a serious contender for the best available

In the comeback story of the year, Google’s Gemini series of AI models, once derided for their weird image generation and criticized for being too “woke”, have roared back with new, more powerful versions now leading third-party performance benchmarks and is becoming increasingly attractive to developers and businesses.

Google presented Gemini 2.0 Flasha multimodal AI model that supports streaming video analysis and can see and instruct what you’re doing on your screen and follow it with Gemini 2.0 Flash Thinking which competes with OpenAI’s o1 and o3 reasoning models.

4. Agentic AI swept the enterprise

Over the course of the year, “agent” AI went from being a fad to a real series of major product announcements and initiatives from leading enterprise software vendors. Take for example:

Salesforce’s Agentforce 2.0: Salesforce introduced Agentforce 2.0 a few days ago, an advanced AI agent program to improve reasoning, integration and personalization features in its CRM and sales propositions, as well as Relaxationgreatly enhancing enterprise productivity tools.

Joule on SAP: SAP converts its Joule chatbot to AI agent powered by large open source language models (LLM), driving innovation and efficiency in corporate settings.

The Google Astra project: As part of the Gemini 2.0 initiative, Google launched Project Astra, an AI assistant designed to provide real-time contextual responses by leveraging Google’s suite of services aimed at improving user productivity and decision-making.

My big prediction for 2025: AI-generated content will reign supreme

Based on these achievements, 2025 is poised to witness the proliferation of AI-generated content in business and consumer domains, especially as everyone from OpenAI to Meta, Google, Microsoft, Apple, even Elon Musk’s XAI now has AI image generators built into their offerings.

This extension will streamline content creation, improve personalization and drive efficiency across sectors.

Additionally, we anticipate the initial large-scale deployment of large-scale language models (LLM) and generative, AI-driven robotics in both commercial and consumer settings, revolutionizing automation and human-robot interactions.

That’s it in the last #AIBeat newsletter for 2024. Thank you for reading, writing, subscribing, sharing, commenting and being here with us. We look forward to sharing more and hearing more from all of you in 2025.

Happy Holidays and a Happy New Year from all of us at VentureBeat to you and your loved ones.