Another whirlwind year in AI is coming to an end, marked by nothing less than pure AI madness. Last year, when AIM made bold AI predictions for 2024, we were pretty aligned with what unfolded. From calling 2024 the year of small language models and open-source AI models levelling up with the closed-source ones, our predictions have been on point.
En route to 2025 and going by the wave, the AI predictions for the coming year are sure to cause an impact at a larger scale. It is likely that physical AI will appeal at a consumer/personal level, with humanoids probably making their way into people’s homes.
Wild? Absolutely. Plausible? Maybe. So, let’s dive deep into some of the wildest AI predictions for 2025.
Agentic AI
No points for guessing, but going by the trend in the second half of this year, agentic AI will continue to grow massively in 2025. All major big-tech companies, including, Oracle, Microsoft, Google, and Salesforce, have launched their agentic suite of products, addressing verticals such as HR, finance, operations, supply and more. A couple of months ago, Microsoft announced the ability to create autonomous agents with Copilot Studio.
Copilot is the UI for AI, and with Copilot Studio, customers can easily create, manage, and connect agents to Copilot.
— Satya Nadella (@satyanadella) October 21, 2024
Today we announced new autonomous agent capabilities across Copilot Studio and Dynamics 365 to help scale the impact of every individual, team, and business…
Interestingly, Gartner named agentic AI ‘the emerging trend for the year 2025’, indicating that autonomously operating agents would have a significant impact across enterprises.
AI in Healthcare
While this may not seem like a new use case, AI applications in the healthcare sector are witnessing an increasing adoption. Not just in healthcare, where medical copilots can assist healthcare practitioners with analysing and summarising reports, discovery and protein research are also seeing huge growth.
Recently, HCG partnered with Accenture, a leading technology company, to use advanced AI to expedite drug discovery and other developments.
Big day today – @nvidia is open-sourcing the BioNeMo Framework, a toolkit of programming resources, libraries, and AI models designed for drug discovery. This release equips academic labs and biotech companies with advanced tools for protein design, small molecule generation, and… pic.twitter.com/4QKpU54co6
— Vega Shah (@dr_alphalyrae) November 19, 2024
Recently, Google DeepMind open-sourced its revolutionary AlphaFold 3 model, making its training weights accessible to researchers for non-commercial use. This has further opened up the possibility of experimenting with drug discovery.
Personal Use of Humanoids
We dubbed 2024 ‘the year of robotics’ as the field saw a surge in research, development, and advancements, including progress with humanoid robots.
Figure, Agility Robotics, Boston Dynamics, and Sanctuary AI, are just a few US companies dedicated to building commercial humanoids. Many of them are already being implemented in automobile companies. However, like in the movie I, Robot, it is possible that humanoids will make their way into our homes too.
Many Chinese tech companies are also speeding through humanoid developments.
Meanwhile, Tesla promoted its humanoid Optimus at the We, Robot event as a humanoid that can assist people with their chores and be part of one’s daily life. While fully functional humanoids might still take time to become reality, the year 2025 will be an important one for this.
In 2025, I look forward to seeing major advancements in developing intelligent systems for humanoid robots—AI pilots that are far more capable than today’s state of the art.
— Dr Singularity (@Dr_Singularity) November 27, 2024
A good benchmark of their potential would be a robot entering an unfamiliar home and competently… pic.twitter.com/CT7eg5W0Or
AI to 3D Systems
Generative AI in 3D modelling is another area that promises to explode in the coming year. Companies such as Common Sense Machines and World Labs are already converting images to 3D.
AI innovation follows a predictable path: text → images → audio → video → 3D.
— Allie K. Miller (@alliekmiller) November 24, 2024
Today, 3D generative AI feels underwhelming, but history suggests that’s temporary. By 2025, expect breakthroughs to reshape industries from gaming to design via world models.
The next dimension… pic.twitter.com/grDtvfzJ7i
Unlike generating images or videos from text, using generative AI to convert images to 3D models is considered highly complex owing to the multifaceted nature of 3D data. This requires understanding and manipulating intricate geometries, textures, and lighting.
Now, with consumer brands already implementing AI to churn out advertisements, 3D will add value to this space. Considering how many companies are experimenting in this segment, 2025 will witness a surge in 3D model companies.
Edge Computing
Edge computing is becoming more critical than ever before for running LLMs on phones. Interestingly, this year saw major phone makers, such as Samsung and Apple, bring intelligence to their phones. Qualcomm and Apple are integrating AI capabilities into their chipsets, allowing robust edge computing for LLMs.
Theres Payton, former White House CIO, said that by 2025, edge computing would become even more widespread, particularly as AI and IoT expand.
https://t.co/a8N6F5DOzI
— Albert Jiang (@AlbertQJiang) October 16, 2024
Two edge models out with impressive capabilities. High time to have a silicon intelligence on your laptop or your phone :) pic.twitter.com/j0ax9l7gXM
China to Lead the AI Race
Co-founder and CEO of Hugging Face Clement Delangue, known for his year-end AI predictions, also made one for 2025. One of his predictions resonates with our suggestion last year that China has a chance to be at the forefront of the AI race.
China is consolidating the world's top AI talent. In 2019 China had 29% of the world's top AI reseachers. Now it has 47% and the U.S. has 18%. To reverse these trends, we need to expand skilled immigraton ASAP. pic.twitter.com/lLqzLoWU1w
— Sam Peak (@SpeakSamuel) December 1, 2024
While the country seems to be in a league of its own, its AI models have been outperforming on the leaderboard. Recently, a powerful open-source model Qwen 2.5 has been clearly leading the AI agents race. Similarly, Baidu has also been making significant developments in the AI field.
