The 4 biggest AI stories from 2024 and key predictions for 2025

By all accounts, 2024 was the biggest year for artificial intelligence – especially when it comes to marketing technology.

The development of major languages ​​(LLM) encouraged by the launch of ChatGPT at the end of 2022 has not shown a slow decline, with many new LLMs launched not only by OpenAI but also by technology giants such as Microsoft, Meta and Google, as well as many other startups . and individual producers.

Reports of a decline in AI research have proven to be, if not unfounded, still overblown at this point.

In addition, new technologies beyond the Transformer architecture that support the largest LLMs began to appear, such as. Liquid AI’s Liquid Foundation Models.

And finally, companies began to embrace the “assistant” approach to AI – creating AI-driven bots, applications, and workflows that can solve unique problems on their own, or with less human leadership than the usual back-and-forth. for LLM chatbots. .

Narrowing the year’s stories down to the top 14, let alone the top 10 or top 4, was a struggle. But I’ve gone ahead and tried, even if I’m cheating a little by combining several stories into one big topic. In my eyes, here are the things that will make the biggest changes from this year:

1. OpenAI grew far and wide beyond ChatGPT

The company that undoubtedly started the era of gen AI did not miss out this year, despite the growing competition from new startups and technological innovation, even its investor and partner Microsoft.

o1 example: OpenAI released its first family of major models beyond its GPT series, list o1 “discussion”.which gives more time to solve problems, which makes it more accurate. It is especially useful in science, writing, and reasoning.

o3 Example: It followed the o1 model from September with a year-end announcement of even world class o3 example. While this won’t be available publicly or publicly until early 2025, it shows that OpenAI isn’t resting on its laurels.

Search for ChatGPT: This feature, first discovered as a call-only stand-alone product is called SearchGPT Before falling into ChatGPT properly, it helps to make online information available in real-time within ChatGPT and better display search results, improve its relevance for recent queries and go head to head against Google, Bing, and the new Perplexity. .

Cloth: Released in October, Cloth expands ChatGPT’s interface beyond the conversational interface to a pane-like workspace that can quickly change content based on the user’s needs, such as editing a document or writing project. Of course, it was hard not to see how it happened, or how it looked like Anthropic’s Artifacts announced a few months ago.

Sora: After almost a year of teasing us with its well-protected video generator, OpenAI in early December finally launched Sora to the publicquickly ordering different systems as they want to differentiate themselves in the hot video AI competition with unique and well-thought-out features and story features.

2. Open source AI took off

Llama 3 and 3.1: Meta is introduced Day 3 in Aprilsetting a new standard for performance in open AI, then followed up with Llama 3.1 in July with 405 billion units. The Llama 3.1 version was used to power Meta AI, the company’s assistant integrated into platforms such as WhatsApp, Messenger, Instagram, and Facebook, with the aim of becoming the ultimate AI assistant.

Rule 3.3: Released in December 2024, Llama 3.3 offered the same functionality as the main models but at a lower cost, making it more accessible to businesses.

Meanwhile, Chinese brands such as Alibaba’s Qwen-2.5 family and New DeepSeek V2.5 and R1-Lite Preview it looked like crap at the top of the benchmark charts, and Nvidia itself continued to offer graphics cards and software designs to support its open-source, powerful system. Nemotron-70B model.

Nous Research, a small outfit in San Francisco that wants to provide more information standard and limited AI modes as open source, as well as the first appearance of several nice new things thoughts.

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

3. Google’s Gemini index turned out to be the most controversial of the best available

In the comeback story of the year, Google Gemini series of AI models that were previously ridiculed for the strange generations of images and criticized for “waking up” too much, came back screaming with new, more powerful models that are now at the top of the third-party charts . and is attracting more and more manufacturers and businesses.

Google was introduced Gemini 2.0 Flasha type of multimodal AI that helps analyze streaming videos and can see and advise what you’re doing on your screen, and track it Gemini 2.0 Flash Thinking which competes with OpenAI’s o1 and o3 models.

4. Agent AI swept the business

As the year went on, “agent” AI went from being a buzzworld to a real list of big business announcements and startups and top software vendors. Take an example:

Salesforce’s Agentforce 2.0: Salesforce unveiled Agentforce 2.0 a few days ago, advanced AI software to help improve thinking, integration, and visibility on all of its CRM and sales, as well Lazinessimproving the infrastructure of the business.

SAP and Joule: SAP changed its Joule chatbot to AI assistant powered by open source versions of major languages ​​(LLMs), driving innovation and creativity in business.

Google Project Astra: As part of the Gemini 2.0 project, Google launched Project Astra, an AI assistant designed to provide real-time solutions, with the help of Google, in order to improve user productivity and decision-making.

My biggest prediction for 2025: AI-generated content will dominate

Based on this, 2025 is about to witness an increase in AI products for businesses and consumers, especially as everyone from OpenAI to Meta, Google, Microsoft, Apple, even. Elon Musk’s xAI now has AI image generators bound to their contributions.

This development will improve production, increase customization, and improve the efficiency of various departments.

In addition, we anticipate the deployment of large-scale language learning models (LLMs) and AI-powered robotics in business and consumer environments, revolutionizing automation and human interaction.

That’s all for the final #AIBeat issue of 2024. Thanks for reading, writing, subscribing, sharing, commenting, and joining us. We look forward to sharing more and hearing more from you all in 2025.

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