https://www.reuters.com/commentary/breakingviews/open-source-spectre-haunts-ai-feast-2026-05-28/
Next couple of years will be interesting. I am looking forward to see DeepSeekās impact in general because not only it is open sourced, but it is significantly more efficient in both data processing and hardware use. I believe, unless ābig boysā manage to rewrite their agents to replace current brute force model to similar to DeepSeekās smart processing they will be in a deep trouble. The hardware cost to run their LLMs alone is unsustainable in a long run. There will be a period of shareholders hanging on and dumping good money after bad before the collapse comes.
To me what was interesting about the article was moving the whole process from a paid service on the net with all the associated security issues (as well as corollary environmental issues) to a local instance of a curated LLM(s) residing on a computer isolated from the net in your office. It was demoed in this video Building a āSecond Brainā: Opportunities, Risks, and Implications for AI Adoption in Singapore
AI is not going away, you just have to be adaptable and learn how to use it as the tool that it is. THe limits will be dependent on your imagination and creativity.
I 100% agree with that.
Offerings may change from current chatbot style interfaces to more specialised ones (forced by financial constraints) for general public. The other part is that conversation centres around consumer end AI, everyone forgets that it is in widespread use in scientific and medical research and pretty much everywhere else out of the general publicās view. I believe those are the sectors we will see biggest benefits coming from but also biggest changes in what type of AI infrastructure is used. It may lead to the fragmentation of the field as well, whether that will be good or bad is to be seen.
āAIā also incorporates many different kinds of algorithms, with many different applications. Many scientific and medical applications use purpose-built machine learning tools. The people throwing billions of dollars at giant data centers would love for you to think that LLMs are just a more capable version of the same thing, but they very much arenāt.
2 Likes