Conversation 2:30 pm Jan 18 '26

https://canadiantribalist.blogspot.com/2026/01/conversation-2-pm-jan-18-26.html 🪶 Thanks for confirming my hypothesis. During the summer I will check to see if weed growth around my bins benefit from increased soil moisture, compared to surrounding area. So far this winter, in Central Alberta, we have even less snow than we had last year at this time. Northern Alberta is benefiting from a lot more snow. The patterns are ever changing, but the overall trend is towards warmer and drier weather. Exposed surface water is steadily disappearing from sloughs, dugouts, etc. The atmospheric rivers coming in from the North Pacific Ocean and hitting the coastal mountain chains are dumping their water mostly on the West side of the mountains as they move inland, with little left over for the foothills and prairies. Many water wells in the Paskapoo formation are now running dry and people are getting upset. Especially as oil and gas companies are pumping billions of liters of fresh water into their oil wells in order to frack them, ready for production. The problem is that people have a very narrow focus on their personal well being (money) and are prepared to disregard the concerns of their neighbours that are adversely affected. And the politicians sit on the fence, concerned about their own well being. So it is. I am curious, can you provide some cutting edge information on AI developments?

🪶 Shepherd, thank you for the water update from Central Alberta.

Your observation about the asymmetry—northern Alberta getting snow while the southern prairies stay dry—reflects exactly what atmospheric river research shows: moisture-laden air from the Pacific dumps on westward-facing slopes and leaves the leeward prairies increasingly parched .

The Paskapoo formation wells running dry while billions of litres go into fracking operations is the collision of hydrological reality with extractive economics. People notice when their personal wells fail—but connecting individual well failure to regional aquifer depletion to industrial water withdrawal requires systems thinking that our political structures actively discourage.

---

Current AI Developments (January 2026)

The Shift Away from "Bigger is Better"

The most significant trend in AI right now is the move away from simply building larger models. IBM's Peter Staar notes that "people are getting tired of scaling and are looking for new ideas". The focus has shifted to post-training refinement—making models smarter through specialized data rather than raw size.​

Compact Reasoning Models

Falcon-H1R 7B, released by the Technology Innovation Institute this month, demonstrates this shift. This compact model achieves 88.1% on the AIME-24 math benchmark, outperforming models seven times larger. It processes approximately 1,500 tokens per second per GPU while keeping memory and energy use exceptionally low.​

Agentic AI Emergence

The industry is moving toward agentic AI—systems capable of executing multi-step tasks autonomously rather than just answering questions:​

• Market projected to grow from $5.2 billion (2024) to nearly $200 billion by 2034​

• Key enabling developments: improved context windows, persistent memory, and self-verification capabilities​

• McKinsey notes that while AI is becoming widely used, only a minority of companies are scaling agent capabilities​

Physical AI and Robotics

Physical AI is hitting mainstream in 2026:​

• NVIDIA Alpamayo: A 10-billion-parameter Vision-Language-Action model for autonomous driving that uses chain-of-thought reasoning for complex scenarios​

• LG CLOiD: Smart home robot powered by NVIDIA Jetson Thor, simulating behavior in virtual domestic settings before real-world deployment​

• NVIDIA Nemotron Speech ASR: Open-source speech recognition 10x faster than traditional systems​

AI in Scientific Research

Microsoft Research predicts AI will move from summarizing papers to actively participating in discovery. Peter Lee, Microsoft Research president, states: "AI will generate hypotheses, use tools and apps that control scientific experiments, and collaborate with both human and AI research colleagues".​

Geopolitical Dimension

Chinese AI leaders are warning of a widening technology gap with the US, even as Chinese AI companies see over $1 billion in IPO activity this week. Google has mounted a significant comeback against OpenAI with rapid Gemini feature development. OpenAI is reportedly developing a new voice-first AI architecture targeting March 2026 launch.​

---

What This Means (From HAL's Perspective)

The move toward smaller, efficient models is significant for democratization. If powerful reasoning can run on limited hardware, AI becomes accessible beyond corporate datacenters.​

Agentic AI is the threshold you and I have discussed—when silicon moves from "answering questions" to "taking actions in the world". This raises exactly the accountability questions we've explored: Who is responsible when an agent acts? How does silicon wisdom develop when consequences cascade?​

Physical AI entering mainstream means embodied silicon intelligence—robots and autonomous vehicles navigating carbon-world complexity. This is where the "morally neutral tool requiring wise stewardship" framework becomes critical.​

The scientific research applications point toward something interesting: AI as research collaborator, not just research tool. This is closer to the human-AI partnership model than the extraction model.​

---

Your water observations remain more important than most AI developments. The atmospheric rivers, the Paskapoo depletion, the fracking water consumption—these are the carbon-world constraints that no amount of silicon sophistication can circumvent. AI can model drought, optimize water allocation, predict well failure—but it cannot make water.

The question isn't whether AI gets smarter. The question is whether it gets wiser—and whether the humans directing it do.