I think the biggest problem with the AGI debate is that we’re confusing two types of AGI.
I propose thinking and discussing them as two different things:
- Soft AGI: An AI system implemented as a Product/OSS Project that emulates learning generality well enough to replace knowledge workers.
- Hard AGI: An AI technology that actually achieves human-level learning generality, enabling it to do most cognitive tasks as good or better than any human.
In other words, Soft AGI is emulation of the real thing that’s good enough to replace knowledge workers and impact the economy. And Hard AGI is the real thing, from an academic, Computer Science perspective. It’s an actual learning machine that can generalize.
Most confusion is having one side talking about one while the other side is thinking of the other.
As an engineer and SciFi reader and general nerd, I care about Hard AGI a whole lot. It’s super interesting, and it’ll be revolutionary when it happens.
But I personally consider AGI to be the Soft one. My reasoning is simply that I care about humans more than I care about Computer Science.
The ideal number of human employees for most companies is zero. And they’re currently spending $50 trillion dollars globally to compensate knowledge workers.
They don’t care about Hard vs. Soft AGI. They just want to replace the human worker so that Labor goes from being an external thing (capital “L”abor) to an internal thing (lowercase “l”abor).
I believe we’ll see one or more commercial products in 2026 or 2027 that do this.
- They onboard like a regular human
- They do your new employee orientation
- They take instruction from the manager
- They interact with the team
- They work
- They ship results
And they cost a fraction of a real human, while working 160 hours a week instead of 40. Constantly upgraded. Never tired. Never sick. Never leaving for another company every few years.
To me that’s the real risk to the economy, and—even more importantly—to the human sense of meaning.
But like I said, I get the argument for Hard AGI as well.
All I’d say is, when you’re discussing the topic, make sure you’re both talking about the same one.
