Subscribe to our daily and weekly newsletters to receive the latest updates and exclusive content on industry-leading AI reporting. Learn more
All in all, 2024 was the biggest year yet for artificial intelligence – at least in terms of commercialization of the technology.
The Large Language Model (LLM) boom sparked by the launch of ChatGPT in late 2022 showed no signs of slowing down, with numerous new LLMs coming not only from OpenAI and established tech giants like Microsoft, Meta and Google, but also from numerous others Startups were introduced and individual developers.
Reports of a slowdown in AI research, while not unfounded, certainly proved exaggerated for now.
In addition, new technologies emerged that went beyond the Transformer architecture that underlies most large LLMs, such as: Liquid Foundation models from Liquid AI.
And finally, companies began to fully embrace the “agentic” approach to AI – developing specific AI-powered bots, applications, and workflows that work on specific problems independently or with less human responsibility than the typical back-and-forth of LLM chatbots can work.
It was an arduous task to narrow down the year’s news stories to the top 14, let alone the top 10 or top 4. But I tried, if cheated a bit, by combining several stories into larger themes. Here’s what I think will have the biggest impact this year:
1. OpenAI has expanded far beyond ChatGPT
The company arguably most responsible for ushering in the Gen AI era hasn’t slowed down this year, despite increasing competition from new entrants and legacy technologies, even from its own investor and partner Microsoft.
o1 model: OpenAI has released its first new family of large general-purpose models beyond the GPT series. The O1 “Argumentation” series.allowing more time to process complex prompts, resulting in greater accuracy. It is particularly effective for science, coding and reasoning tasks.
o3 model: It followed the o1 model from September with a blockbuster year-end announcement of a balanced balance sheet more advanced o3 model. Although this won’t be publicly available or even available to third parties until early 2025, it shows that OpenAI is not resting on its laurels.
ChatGPT Search: This feature was originally introduced as a standalone product called Invite-Only SearchGPT Before being integrated into ChatGPT proper, it enables better real-time retrieval of web information within ChatGPT and a more refined presentation of search results, improving utility for current queries and competing with Google, Bing and newcomer Perplexity.
canvas: Launched in October, canvas Extends the ChatGPT interface beyond that of a conversational interface to a workstation-like area that can dynamically update content at the user’s request, e.g. E.g. when editing a document or a coding project. Of course, it was hard not to see it as a reaction to that, or at least a comparable trait Anthropic’s Artifacts was announced several months earlier.
Sora: After nearly a year of teasing us about his closely guarded video generator model, OpenAI finally introduced Sora to the masses at the beginning of Decemberwhich quickly garnered 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 user interface and storyboarding feature.
2. Open source AI took off
Lama 3 and 3.1: Meta introduced Llama 3 in Aprilwhich set a new standard for performance in open source AI, quickly followed in July with Llama 3.1 with 405 billion parameters. Versions of Llama 3.1 were used to power Meta AI, the company’s assistant integrated into platforms such as WhatsApp, Messenger, Instagram and Facebook, with the goal of becoming the most widely used AI assistant.
Llama 3.3: Published in December 2024, Llama 3.3 delivered performance comparable to larger models but at a fraction of the computational cost, making it more accessible to enterprise applications.
There are now Chinese models like Alibaba’s Qwen 2.5 family And DeepSeeks new V2.5 And R1 Lite Preview seemed to come out of nowhere to the top of some benchmark charts, and Nvidia itself went beyond providing graphics cards and software architectures and released its own powerful open source device Nemotron 70B model.
Nous Research, a small company in San Francisco that wants to offer more personalized and less restrictive AI models as open source, several also debuted Cool new Ideas.
And let’s not forget France mistralwhich rapidly expanded its own open source and proprietary AI offerings.
3. Google’s Gemini series became a serious competitor for the best series available
In the comeback story of the year, Google’s Gemini line of AI models, once mocked for their strange image generations and criticized for being overly “woke,” came back with new, more powerful versions that now top third-party performance benchmark charts and are increasingly attractive for developers and companies.
Google introduced Gemini 2.0 Flasha multimodal AI model that supports streaming video analysis and can see and instruct what you are doing on your screen and then track it Gemini 2.0 Flash Thinking which competes with OpenAI’s o1 and o3 argumentation models.
4. Agentic AI took over the enterprise
Over the course of the year, “agentic” AI went from being a fad to a real series of major product announcements and initiatives from leading enterprise software providers. Take for example:
Agentforce 2.0 from Salesforce: Salesforce has introduced Agentforce 2.0 a few days ago an advanced AI agent program to improve reasoning, integration and customization capabilities in its CRM and sales offerings Loosesignificantly improving business productivity tools.
Joules from SAP: SAP has converted its Joule chatbot into one AI agent based on open source LLMs (Large Language Models).) and promotes innovation and efficiency in the corporate environment.
Google’s Project Astra: As part of the Gemini 2.0 initiative, Google launched Project Astra, an AI assistant designed to provide real-time, contextual answers using Google’s services, with the aim of improving user productivity and decision-making.
My big prediction for 2025: AI-generated content will take over
Building on these advances, 2025 is expected to see the proliferation of AI-generated content across business and consumer sectors, particularly across everything from OpenAI to Meta, Google, Microsoft and even Apple Elon Musk’s xAI now has AI image generators integrated into their offers.
This expansion will streamline content creation, improve personalization, and increase efficiency across various sectors.
Additionally, we expect the first large-scale deployment of large language models (LLMs) and generative AI-powered robotics in both commercial and consumer sectors, revolutionizing automation and human-robot interactions.
It’s all in the final #AIBeat newsletter for 2024. Thank you for reading, writing, subscribing, sharing, commenting and for being here with us. I look forward to sharing more with you and hearing more from you in 2025.
From all of us at VentureBeat, we wish you and your loved ones a happy holiday and a new year.
Source link