2 ms·

I agree that generative AI doesn't have any direct parallels, but > but no technology has ever been able to do what these models do. That's not new. Steam pow

I agree that generative AI doesn't have any direct parallels, but > but no technology has ever been able to do what these models do. That's not new. Steam power, the first computers, _especially_ electricity did things no previous technology could.

pixl97 3y ago
>Steam power, the first computers, _especially_ electricity did things no previous technology could. And it massively changed the world.... so, um, buckle up, I guess.
- dukeyukey 3y ago
  I absolutely agree it'll change the world. But in a meta sense, we've been through this before. Even in living memory, my grandparents were middle-aged before computers became a big thing and at least one of them grew up in an unelectrified Irish village.
  - pixl97 3y ago
    Rate of change is a significant factor. Electrification is an example of that, it takes a massive amount of time to put the infrastructure in. Then even when you got a power pole to your house, you had to run lights and it was unlikely you had any other appliances that used power. Those were slowly and expensively added over time. AI may work from the opposite direction. It uses the infrastructure that already exists, we have power poles, and network lines, and all kinds of computing devices gathering information and feeding it back to datacenters. This sets up a 'fertile data environment'. One potential outcome here is we see an increase AI compute power over the next few years with only a moderate increase in AI intelligence, then by change in algorithm that compute power is now far more efficient in creating intelligence. This could manifest itself as "AI is a great helper for a programmer" to "Actually, why do we have so many human programmers at all" in an exceptionally short time frame. It is the disruptive timeframe that is the largest risk to society. If for some reason we go from "Corporations need people" to "Corporations don't need people" in a very short time frame, we'll Moloch ourselves in to high unemployment and social instability quickly.

pixl97 3y ago

>Steam power, the first computers, _especially_ electricity did things no previous technology could. And it massively changed the world.... so, um, buckle up, I guess.

dukeyukey 3y ago
I absolutely agree it'll change the world. But in a meta sense, we've been through this before. Even in living memory, my grandparents were middle-aged before computers became a big thing and at least one of them grew up in an unelectrified Irish village.
- pixl97 3y ago
  Rate of change is a significant factor. Electrification is an example of that, it takes a massive amount of time to put the infrastructure in. Then even when you got a power pole to your house, you had to run lights and it was unlikely you had any other appliances that used power. Those were slowly and expensively added over time. AI may work from the opposite direction. It uses the infrastructure that already exists, we have power poles, and network lines, and all kinds of computing devices gathering information and feeding it back to datacenters. This sets up a 'fertile data environment'. One potential outcome here is we see an increase AI compute power over the next few years with only a moderate increase in AI intelligence, then by change in algorithm that compute power is now far more efficient in creating intelligence. This could manifest itself as "AI is a great helper for a programmer" to "Actually, why do we have so many human programmers at all" in an exceptionally short time frame. It is the disruptive timeframe that is the largest risk to society. If for some reason we go from "Corporations need people" to "Corporations don't need people" in a very short time frame, we'll Moloch ourselves in to high unemployment and social instability quickly.