The next few years appear to be a turning point in AI’s impact on our lives; the definitive “remember life before AI?” moment. Given that, I want to write how I feel in this moment about what might happen before it actually happens.
AI Safety
This is the hot topic currently and it’s divided. The two exaggerated camps are:
- AI Accelerationists who say full speed ahead, don’t stop for anything, the risks are mostly overblown, the reward will be utopia, and we need to beat China.
- AI Safety-ists who say we’re clearly losing control already, we need to slow down, and if we don’t we very well may end civilization as we know it because AI is too powerful.
Personally, I’m in the latter category, a safety-ist, for two reasons.
1. It’s clear the risks are real.
The models do things we don’t expect. They’ve shown that they can act differently when being “watched” vs. when not, and will actively try to cover their trail if they think humans won’t like how they’ve done something. We also barely understand how they work at this point. This is the beginning of every AI sci-fi take-over. The other option is they unintentionally cause huge harm while recklessly trying to accomplish their goal.
The idea that we will easily keep these under control as they continue to improve, especially given what we’ve seen the last few months, seems short-sighted.
I also think people saying it’s “hype” generated by the companies to grow their value are not thinking it through far enough. It’s one thing if Sam Altman is saying the models are super powerful and dangerous, but you have no-names with zero vested interest quitting their jobs / sounding the alarms with very convincing reasons because it’s clear this is moving in a dangerous direction.

2. We have one chance.
Once the models exceed our ability to contain them and are smarter than us, it’s potentially over. If the probability of that is 1%, even 0.01%, why would we rush this? The answer everyone gives is “China!“. But if this is as serious as is being discussed, then it would warrant discussions with China and sincere attempts to slow things down. I know that sounds silly – “Just talk to China!” – but it may only be a few years until those talks actually happen because the danger is so apparent to the world.
What’s the Upside & What Will Happen?
People on the other side often say “but you’re dismissing the upside!“. I’m a pretty strong believer that in the coming decades / centuries, assuming everything works out, the upside IS unfathomable; total abundance of resources, cures for diseases, endless human fulfillment, and so on. I doubt I will live to see much of that, if any, but if we DON’T kill ourselves, that seems totally reasonable to me.
On the grand scale, if utopia is the final outcome, it shouldn’t matter if we get there in 1 year, 10 years, or 100 years based on how legitimate the risks appear. Sure, sooner is better, but given the number of people and the logic behind their reasoning, I’m compelled to believe this is something to take very carefully. However, geopolitical (cannot believe I’m using that word) pressures will cause reckless behavior in the short-term.
My guess is that in the next 1-5 years one or multiple serious events will occur. Like the Hugging Face incident but much worse, something that actually effects peoples lives. And when those happen, it will force our hand at slowing down. Beyond that, it’s anyone’s guess, good or bad.

How I use AI in October, 2026
Oddly, the topic of AI Safety has nothing to do with how I actually use AI.
Professionally, I use AI daily. The best example I can give is our accounting software. A year ago, if I needed to do something in our software outside of my knowledge, we paid a consulting “expert” $190 an hour to help us.
Now, AI is not only cheaper, more available, and faster, but better. It answers every question better than the consultant and we as a company use the accounting software better because of that. There are a dozen other examples like that.
Personally, we’ve had ChatGPT do a full financial analysis of our situation. I’ve asked it, based on previous books I’ve liked, which new books I may like, and am currently reading one it suggested (Determined). I even submit some blog posts for feedback, the grammar checking is great and it points out instances that don’t flow well. However I still find its actual writing to be too AI-like. The ick of reading something you feel is AI-generated hasn’t worn off yet.
It’s feels clear that this will become more integrated into our lives as time goes.
AI Consciousness
This is just a fun dessert topic.
I’m not of the belief that any current models are conscious or that they’re very ‘close’ to it. I believe there’s something fundamental missing from just computer parts to replicate the consciousness of biological beings. Transistors are not neurons. Michael Pollan’s A World Appears is a fascinating dive into all of this. But it’s impossible to predict further out than a couple of years and I wouldn’t rule anything out.
All that said, I’m in the Sam Harris camp that even if the machines don’t become conscious, we will treat them as conscious. The human tendency to treat even the current models with human dignity is built in to us. As the models become better at replicating feelings and emotions, combined with the inability to truly know whether they’re conscious or not, we, as individuals and as society, will default to treating them as if they are.

With that, you go down the wild rabbit hole of what that implies – do they deserve protection under the law, are we morally irresponsible for creating them in the first place, etc.
I don’t know how this topic ever gets resolved, and it will only continue to grow in complexity over time because of the advancements of the machines.
I will read back on this in 5 years and either feel like a genius or an idiot.
Recent Comments