Credit: VentureBeat/ChatGPT
The 4 biggest AI stories of 2024 and a key prediction for 2025.
Join our daily and weekday newsletters to receive the latest updates on AI and exclusive content. Learn More
2024 will be remembered as the year of artificial intelligence, at least in terms of commercialization.
Despite the fact that the large language model (LLM), spurred by ChatGPT’s launch in 2022, has not slowed down, a number of new LLMs have been introduced by OpenAI, Microsoft, Meta, Google, and other tech giants.
Reports that AI research was slowing down were, if not unfounded then certainly overstated.
New technologies, beyond the Transformer architecture, which underpins many big LLMs, began to emerge, such as Liquid AI’s Liquid Foundation Models.
Lastly, firms began to fully adopt the “agenttic” approach to AI – developing specific AI-powered applications, workflows, and bots that can work independently or with less human oversight than the typical back-and forth of LLM chatbots.
It was difficult to narrow down the top 14 stories of the year, let alone a top 10, or even a top 4. I have tried to do this, even if it was a little cheating by combining a few stories into larger themes. Here’s what I think will have the biggest impact in the future:
OpenAI expanded beyond ChatGPT.
Despite increased competition from newbies, legacy tech and even Microsoft, the company that is credited with kicking off gen AI didn’t miss out on a beat in 2018.
The o1 Model (): OpenAI released the first of its new large general-purpose models, beyond its GPT family, called the “reasoning” series. This series allocates more processing time to complex prompts and results in higher accuracy. It is particularly effective at science, coding, and reasoning tasks.
The o3 model was announced at the end of the year, following a huge announcement in September about an even more advanced o3 model. OpenAI isn’t resting on their laurels. This model won’t be available to the public or even third parties until early 2025. ChatGPT Search (19459034) : Originally launched as a standalone product for invitation only, SearchGPT, this feature allows more real-time retrieval of web information within ChatGPT, and a refined display of search results. This feature is a direct competitor to Google, Bing, and the newcomer Perplexity. Canvas: Introduced in October of this year, Canvas extends the ChatGPT user interface beyond a conversational pane to a workstation like pane which can update content dynamically on the user’s command, such as editing documents or coding projects. It was impossible to miss the comparison to Anthropic’s announcement of Artifacts, which had been made several months earlier.
Sora () : After nearly a full year of teasers with its closely guarded model of video generator, OpenAI finally launched Sora in early December, provoking a variety of reactions. It sought to differentiate itself in the fiercely competitive AI video market with a unique interface and storyboarding feature.
2. Open-source AI took flight
Llama 3: Meta introduced Llama 3 (a new standard in performance for open-source AI) in April and followed it quickly with Llama 3.1.1 in July, which had 405 billion parameters. Meta AI’s assistant, which is integrated across platforms such as WhatsApp, Messenger and Instagram, was powered by versions of Llama 3.1. The company aimed to make it the most widely used AI Assistant. Llama 3.0: Released in December of 2024, Llama 3.0 delivered performance comparable to large models at a fractional cost, making it easier for enterprise applications.
Chinese models like Alibaba’s Qwen 2.5 family and DeepSeek’s new V2.5, R1-Lite Preview and DeepSeek’s new R1-Lite Preview, appeared out of nowhere and topped some of the benchmark chart. Nvidia launched its own powerful, open-source Nemotron 70B model. Nous Research, an open-source company in San Francisco that aims to provide more customized and less restrictive AI models, has also introduced some cool new ideas.
Let’s not forget France Mistral, who has rapidly expanded their own open-source AI offerings.
3. Google’s Gemini Series became a serious contender in the race for the best AI model available
In arguably the most impressive comeback of the year, Google’s Gemini AI models, which were once ridiculed for their strange image generation and criticized for appearing too “woke”, have returned with new and more powerful versions. These are now topping the performance benchmark charts of third-party companies and are becoming increasingly attractive to developers and businesses.
Google released Gemini 2.0 Flash – a multimodal AI that can analyze streaming video and see what you are doing on your screen. It was followed by Gemini 2.0 Flash Thinking, which competes with OpenAI o1 and o3 thinking models.
4. Agentic AI has swept the enterprise.
As this year progressed, “agentic” AI became a real product announcement and initiative by leading enterprise software vendors. Take, for example,
Salesforce’s Agentforce2.0. Salesforce released Agentforce 2.0 a couple of days ago. It is an advanced AI agent that will enhance reasoning, integration, and customization features in its CRM and Sales offerings, as well Slack. This will improve enterprise productivity tools. SAP’s Joule (19459034) : SAP transformed its Joule chatbot to an AI agent powered with open-source large-language models (LLMs), enabling innovation and efficiency within enterprise settings.
Google Project Astra: As part the Gemini 2.0 initiative Google launched Project Astra. It is an AI assistant that provides real-time contextual responses using Google’s suite services. The goal is to improve user productivity and make better decisions.
My big prediction for the year 2025: AI generated content will be king
As a result of these advances, 2025 will see the proliferation and adoption of AI-generated images across all business and consumer domains. This is especially true as everyone from OpenAI, Meta, Google, Microsoft and Apple to Elon Musk’s xAI has AI image generators integrated into their offerings. This expansion will streamline the creation of content, enhance personalization and drive efficiency across various sectors.
We also anticipate the first large-scale deployments, both in commercial and consumer settings of large language models and generative AI powered robotics, which will revolutionize automation and human-robot interaction.
This is the last #AIBeat 2024 newsletter. Thank you for reading, writing and subscribing. We appreciate your comments, sharing and sharing with others. We look forward to hearing from you in 2025.
From all of us at VentureBeat, we wish you and your family a happy holiday season and a prosperous New Year.
VB Daily provides daily insights on business use-cases
Want to impress your boss? VB Daily can help. We provide you with the inside scoop about what companies are doing to maximize ROI.
Read our privacy policy
Thank you for subscribing. Click here to view more VB Newsletters.
An error occured.
