AI Trends for 2025

[@iamhammadnasir]

1. Agentic AI 2. Inference Time Compute 3. Very Large Models 4. Very Small Models 5. More Advanced Use Cases 6. Near Infinite Memory 7. Human-in-the-loop Augmentation

[@ankitkumar-ff6ks]

Domain Specific Models

[@Leocesar3D-mx6rm]

- Mass customization. - Creation of new business models. - Inference Time Tuning. - No need to retrain the model to get better performance. - Advanced agents. - Automation of complex processes. - Model library: - By size (in billions of parameters). - Domain Specific Models. - Consolidation of Use Cases: from the simplest to the most complex. - Human or team cooperation with generative AI. - Adjustable memory: remember or forget via configuration.

[@anishnatekar7283]

Simulation Generation or Generative digital twins as the one showcased in the "Genesis" model. Agents need an environment to interact and learn to perform complex real world tasks for that an environment will need to be generated.

[@mohammedmusaib7725]

8. Industry specific Model, with more regulations and security guidelines added...

[@roadofsuccess4299]

So Sir in my opinion, there will be a lot of boom in ai agentic rags as compared to all of the others. It's because they will be fully autonomous and can be capable of doing anything in more better and effective way as compared to human and at a very cheap cost. So there will be a major layoff for a large number of people who are at call centers, data entries , basic web developers, as you also knew open ai launched their o5 model which outperformed in maths and coding with a remarkable benchmarks. So things going to be very interesting in the future. Let's see :)

[@user-mj2lm5fh1j]

One trend is none of theses companies will be profitable

[@jonatasorto]

When was this recorded? Before or after o3?

[@AiLife-f4l]

Agentic AI in healthcare management should become a trend! It could optimize workload management in large hospitals, like AIIMS New Delhi.

[@janbodnar7815]

data analysis via prompts or even better verbal instructions, deep search (searching completely redefined, new kinds of "browsers"), AI assistants popping everywhere, integration of AI into existing applications continues, importance of Local AI increases, holograms (this comes probably later)

[@asifdotpy]

Excellent explanation Sir @Martin Keen.

[@hdl75248]

I would like to know when AI is "thinking" during inference, what exactly goes on in the background.

[@Raymond_Cooper]

Thanks for posting ❤

[@leventkarabey4138]

Corporate specific AIs. Maybe already included within Very small Models. AI models trained with company specific data, closed to outside world. Working as an assistant at first and probably promoted being an employee later on. Then CEO :)

[@Researcher-c9c]

8. “Almost there” AGI

[@elmoreglidingclub3030]

9. Improved teaching at the university level. It’s where I teach, and AI will need to be taught through the lens of organizations (broadly) and fields of study. Lectures will need to be replaced by case studies and problem solving using AI tools as the best pedagogical practice. I am implementing these at my university and the demand is tremendous.

[@AjaySingh-ey7gt]

Very Good Video.Best Of luck.

[@amcluesent]

We won't know who set fire to the sky

[@frun]

8. Change of an architecture of ANN on the fly as in quantum fields (particle creation/destruction)

[@BizAutomation4U]

Great video Martin. I think 2025 will see the battle between closed source vs open source (LLAMA 4, etc..) really start to crystalize. How, well it would be great to hear your thoughts on this, perhaps from a future video, but in my view, I think enterprise apps (like the one built by my c ompany who I shall not shamlessly plug) will favor the open source because of how we can integrate and fine-tune vs the frontier models. I'd also be curious what you think of the rise of AI in government (Elon & Trump sitting in a tree... yada). Keep up the great work !

[@amritbro]

I think federative training and proper prompt engineering skill will also take their spot on 2025 so i will put them on trend number 8 and 9 respectively.

[@nodavood]

Mathematician/Theoretical Physicist AI: Combining an LLM agent with a symbolic math checker agent guided by RL to prove new theories.

[@sitedev]

Tighter integration with external resources/tools (better function calling). Eg Anthropic’s MCP

[@TsviGirsh]

8) Video generation/editing

[@owenmorgan805]

AI in surprising places, with seeming ornaments being AI endpoints

[@blaisepascal3905]

Matmul free models?

[@box4soumendu4ever]

Thank you very much Sir. 🥰🥰🙏🏼🙏🏼❤❤🕊🕊🪻

[@artbybigvee]

Thank you.

[@adamgibbons4262]

The AIs need a theory of mind like humans have. I really feel like self reflection should be a bigger part of how AIs operate. We don’t read a book or learn something a grasp it in a single instance we need to think new concepts over to fully understand them, AI needs to do something similar with safe guards in place to ensure it doesn’t unravel into an entropic mess

[@CrunchyMind]

Actually personalized assistants. Full knowledge of your personality with realtime communication. Maybe not next year, but also integration with IoT devices, like actual helpful integration not the shit we have right now

[@russwilson2182]

Better solutions built to ask questions on structured data.

[@amoldhamale3202]

I think you're contradicting yourself 6:09 onwards "chatbot+docs group also scored lower than when the bot was asked to solve cases alone and that is failing of AI and human augmentation" and then you go on to say that "expert paried with AI system should be smarter together" So how did the experiment with chatbot+docs fail to score higher than the chatbot alone?

[@agracian1]

Healthcare, particularly diagnostic imaging from multiple device types, should extend to preventative care, not be limited to symptom-based diagnosis.

[@savant_logics]

Trend # 8: Breakthrough in Coherence

[@d.d.z.]

The responsibility of this not be just hype is to find use cases.

[@muhammedcagrkartal9954]

real time thinking models should be worked to be achived

[@Iamamc]

8. Systematic Framework Establishment of a self conscious agent enhancing human-AI integration

[@raymobula]

Vision and sense making (look up the video where Gemini observes the screen and teaches a user how to use Blender). Personally I hope the lying and scheming gets addressed. Real System 2 thinking is achieved. As well as models based on values are created.

[@amritbro]

Yeah for sure 2025 is gonna be a year of agentic AI

[@user-xt9ji2ld5y]

Optimize energy efficiency for Ai

[@surajjaiswal1371]

I believe Multi-Modal AI will be a major trend in 2025, transforming how we interact with AI and driving the growth of AR gadgets. Robots will also see significant performance boosts. I'd love to see a Linux distro built for this-an efficient, practical OS with AI and local Multi-Modal Agents, free from gimmicks.

[@rajausamah7569]

Dedicated AI OS and chips

[@x_or_10]

Google has made great strides for generative games this year to GameNGen to Genie. Therefore, I substantial advancements in both the capabilities and commercialization of this technology in 2025.

[@abhilashanand5886]

8. more guardrails and frameworks for LLMs

[@henrycobb]

Thou shall not brute force the universe. The missing trend is robotic sloth where the machine doesn't need a human to tell it what things aren't worth worrying about at the current moment.

[@anant8449]

Automation of computer tasks using AI

[@ruvenwang6769]

Sustainability AI or Green AI

[@HshsHsjs-h1m]

8. More advanced ai chips and gpu's combined with cpu's. Gpu's built into the laptop

[@bradhatch8302]

LLMs using forecasting models as tools, and being better at interpreting quantitative information.

[@epic_miner]

❤❤❤

[@darshantawte7435]

Unless and until they find a nuclear reactor to powerup these GPU the energy costs and demands will keep soaring, also there are very few companies who actually use LLMs or need them for that matter. Most of them are still better with traditional ML algorithms.

[@christopherpetersen342]

8. Not sure if it will be 2025, but the current approach will create such diminishing returns given the insane and unsustainable resource inputs that a lot of the hype will finally start to be revealed as ... misleading. If 5 orders of magnitude more "parameters" doesn't make hallucinations go away, why should we believe there is a threshold with the current types of models that ever will? If folks are not willing to pony up billions of $$$ or tens of gigawatts to train and run this stuff, the entire cycle has to shift to prioritize efficiency, effectiveness, and sustainability over growth.

[@realstudy6083]

8)Video generation

[@ruvenwang6769]

Quantum AI

[@xrobot8052]

Distributed models

[@1010Analytics-io]

❤

[@Prakashmcc3]

🎉

[@data_wizard-v6w]

AI powered robots

[@mirzacickusic]

8: AI generated conten

[@nadeemchaudhry7623]

Multimodal customer service bots.

[@MalcumY]

Decentralized AI Agents 😂

[@patriciaetienne5070]

Digital Twins

[@artbybigvee]

I think you should be more assertive.

[@RK-fr4qf]

Open source

[@mentor1013]

Better Video Models

[@JohnBeeblebrox]

I found this presentation interesting but biased (naively utopic viewpoint). I believe there will be a pushback against "AI" when some of the negative features (potential or actual) start to impact peoples lives and society at large. Just my 2p worth...

[@messi-r7m7r]

-gf relationship simulation chat bot. -something fun around TikTok probably, brain rot "skizo poster" style.

[@aventurasenvideo]

Call centers running virtual AI voice agents that are extremely good in understanding the human and solving their problems while sounding empathetic.

[@abhijithts7500]

deep fakes

[@leninjacobregi4010]

Personalised clone ai bot

[@info-cv8tg]

agi

[@shahul41u]

AGI

[@Nendonen]

First 🎉

[@davidtheopgamer7181]

naaaaaaa hype

[@acedemiczoomer-je1de]

skibidi toilet

[@creatorcode96]

I am the first commenter

[@DefenderX]

8. AI Singularity

[@data_wizard-v6w]

Quantum AI

[@code0101-t9i]

Quantum AI