Merry Christmas, Happy Hanukkah, Happy Kwanzaa, and other holidays you celebrate in December! 🎄

As you enjoy the holiday season and take some much-deserved time off, it’s the perfect moment to reflect on 2024 and look ahead to the exciting developments 2025 holds for us.

While Generative AI has been at the forefront of our focus, we’ve also dug into various resources to bring you a carefully curated list of the most important AI trends to follow in 2025.

Our insights are based on the perspectives of these experts:

• Sissie Hsiao, VP & GM of Gemini App & Speech at Google

• François Chollet, Creator of Keras and ARC-AGI Project

• Jim Fan, Senior Research Scientist, AI Agents at NVIDIA

• Swyx, Developer and Community Leader

• Scott Likens, AI and Data Leader at PwC

• Martin Keen, IBM Fellow

• Nathan Benaich, General Partner at Air Street Capital

• Clement Delangue, Co-founder and CEO at Hugging Face

• Carrie Tharp, Vice President, Strategic Industries at Google Cloud

Now, let’s dive in!

10 AI Trends in 2025

1. Autonomous Agentic AI

AI agents will become more independent, taking on complex, multi-step tasks with minimal human input. They can plan, reason, and adapt in real-time. As Sissie Hsiao from Google puts it, “AI assistants will evolve into true, personalized, advanced experiences that users rely on daily.”

It means that one day, AI could handle more complex problems like scheduling your week or organizing a big project. Currently, AI agents are great at predictable tasks but struggle with nuances and unexpected scenarios.

2. Multimodal AI

For AI agents to become more autonomous, they need to handle information like we do, through sight, sound, and text all at once. That’s what multimodal AI aims to achieve. As Aashima Gupta, Google Cloud’s Global Director of Healthcare Industry Solutions, describes, “We'll see a rise in the adoption of multimodal AI models to analyze data such as medical records, imaging data, and genomic information to draw insightful summaries, moving closer to the vision of personalized medicine.”

Picture an AI that can read a report, listen to your voice notes, and understand a chart simultaneously. AI is starting to combine inputs, but it’s far from mainstream, and making these connections is still tough.

3. Inference-Time Compute

Inference-time compute is about optimizing how AI “thinks” when it’s working on a task. François Chollet sees it as “the big driver of AI capabilities” in the coming years:

This is already making a big impact, especially for devices like smartphones, where speed and efficiency are critical. Think of Apple Intelligence’s on-device AI or Hugging Face’s work on lightweight models, giving them more time to “think” based on task difficulty, inspired by OpenAI’s o1 model.

4. Edge AI and Small Models (SLMs)

Edge AI moves computation from the cloud to devices like phones and sensors. Small language models (SLMs) are leading this shift. As NVIDIA’s Jim Fan explains, “2025 will be the year a lot of computation begins to shift to inference at the edge.”

It means AI that works better where you are, whether it’s a chatbot on your phone or a robot in your home, without compromising your privacy or draining your battery.

5. Near-Infinite Memory

AI is getting better at remembering conversations, preferences, and interactions you’ve had with it. This “near-infinite memory” could enable hyper-personalized experiences. Martin Keen predicts, “We are getting close to near-infinite memory, where bots can keep everything they know about us in memory at all times.”

AI is starting to expand its memory capabilities, but it’s not seamless yet. Plus, there are big questions about privacy—do you really want your AI to remember everything?

6. Real-Time Voice: Typing Is So Yesterday

Real-time voice integration is set to replace typing entirely. Swyx predicts, “Every chatbot input will have a real-time voice by 2025.” We talked with Swyx and Alessio on the Latent Space podcast:

Voice AI is already here, but it’s not perfect. Misunderstood commands and uneven language support are still pain points. You can expect smoother, more natural conversations with your devices without repeating yourself five times to get your AI to “understand” you.

7. Better AI integration into workflows

AI’s real value lies in enhancing productivity by working alongside humans. As Martin Keen explains, “An expert paired with an effective AI system should be smarter together than either alone.”

Companies are making strides to integrate AI into tools like email, project management software, and coding platforms. But the adoption can be slow when the tools feel too complex or don’t fit into existing workflows.

8. Accelerating Product Development

AI is expected to cut product design timelines in half. As Scott Likens from PwC says, “AI will transform product design with tools that iterate designs in hours, not weeks.”

This is already starting, with AI helping non-coders build apps and companies speeding up prototyping. As Nathan Benaich predicts, in 2025, an app built by a non-coder will go viral reaching the App Store Top 100.

9. Increased Investments

Governments are doubling down on AI with billions in sovereign investments aimed at boosting research, infrastructure, and innovation. According to the State of AI Report 2024, “A $10B+ sovereign investment will reshape the AI market in the US.”

These investments are already driving breakthroughs, but they come with turbulence. Uncertainty around leadership and challenges in infrastructure like GPU shortages could slow progress.

10. Breakthroughs in Science

AI is unlocking new possibilities in biology, chemistry, and physics. From accelerating drug discovery to revolutionizing materials science, these advancements could have life-changing impacts. As Clement Delangue, co-founder and CEO at Hugging Face, predicts, “We’ll see big breakthroughs in AI for biology and chemistry.”

We need continued investment in interdisciplinary research and better tools for scientists to leverage AI effectively and translate it into real-world solutions.

We’re just starting to see where these exciting developments will take us. Stick with us in 2025 for more updates and insights! While you’re enjoying the holidays, check out our updated Prompt Engineering Guide, blog, and collection of Generative AI courses.

Happy holidays! 🎄✨ See you in 2025!

Thanks for reading! We’d love your feedback to make this newsletter even better. Our goal is to share content that’s valuable and relevant to you. Help us fine-tune future editions by sharing your thoughts on this week’s email—it’ll only take a moment!