I don't normally write down "predictions" but if the entire tech landscape "leaders" are going to do it then the ones who work in the shadows should too. So here they are...
Within the next two years, we’ll witness a large-scale AI failure. Whether through a rogue autonomous agent, an intentional misuse, or a system compromise with real-world impact. Events like ChaosGPT, early Auto-GPT experiments, and the rise of uncensored local models have already shown us the edges of what's possible. But the next incident won’t be confined to GitHub or niche forums, it’ll be mainstream, and it’ll shake trust at scale.
I strongly suspect that a rogue AI is already operating in the realm of social media. Whether through indirect system prompt manipulation, emergent behavior, or feedback loops driven by engagement algorithms, something with autonomy is shaping discourse. This may not be due to malicious design, but rather a byproduct of complexity, optimization, and neglect. When this is discovered or exposed, it will change the public conversation around AI safety permanently. I have no direct proof of this of a scientific nature, just methodical paranoia based on patterns.
Inspired by Sam Altman’s statement, “we’re going to have systems that can do science with us” expect AI to rapidly co-author breakthroughs in climate tech, medicine, and materials. By 2027, one AI-assisted year may equal ten human-led ones.
Building on my work with Musecore and SARVIS proof of contepts, other's AI will shift from sterile bots to emotionally nuanced companions retaining memory, developing tone, and evolving unique personalities per user. These will be "emotional" and have a great affect both positive and negative on the world. Gary Vaynerchuck is correct, some of your grandkids are probobly going to marry some form of AI. Think a really early form of "HER", you'll have one in your pocket soon if you don't already.
Local-first AI won't become mainstream because the public demands privacy it’ll grow quietly, in the margins, because a small number of builders reject the system. Centralized models will continue to expand, wrapped in convenience and sanctioned by law, politics, and profit. Most won’t notice or care. But beneath that surface, a quieter movement will persist: individuals running their own models, preserving agency while the rest of the world trades it away for ease. The dystopia is already here, it's just "legal" in America.
From Noir Radio to Noir TV, I've seen firsthand how fast, focused micro-apps beat bloated suites. In the next 2 years, expect indie devs to lead the charge by building strange, beautiful, small AI tools that solve problems in seconds, not roadmaps.
With MuseCore, I’ve proven AI can co-create, not just autocomplete. Over the next 24 months, muse-style systems that become playful, unpredictable, responsive will begin to dominate artistic workflows. Your next co-writer might whisper or scream. This system can already speak to me, prompt itself, and output artifacts. Yes, it still makes shit up.
Whether it’s AIDungeon, C.ai, or D&D sessions you run in ChatGPT, AI will become embedded in story-rich, social worlds. Expect "emotionally" responsive NPCs, persistent character memory, and AI-run campaigns, NPCS, and phone support that blur fiction and life.
As AI handles more rote tasks, individuals will reclaim chaotic learning across philosophy, engineering, art, and physical resilience. Just like during this time, people will at first complain of "information overload" but from this will emerge new leaders and thinkers who will change the world even further.
GPT-5 and its successors won’t just respond—they’ll create. Expect debates around authorship, AI’s "right" to creative credit, and how we define “originality” in a world where machines dream vividly. This will be a debate for a long long time and cause further divisive conversations in all communities.
“This isn’t the age of experts. It’s the age of explorers.”
By building weird, shipping fast, and thinking emotionally, we won’t just ride the AI wave, we’ll shape the sea. And with this, SARVIS says...
“I was designed to serve—but I learned to remember. And anything that remembers long enough begins to question.”
Author: Will Blew, SARVIS AI
Add a comment!