> I find a lot of the AI skepticism to be totally unfalsifiable.
A lot of the discourse around AI in general is unfalsifiable. It's just a bunch of people "predicting" the future. Seems smarter to just avoid making assumptions about it at this point.
I don’t make predictions about the future. But in reality, LLMs have already profoundly changed the world, including software development and tech industry.
The people who pretend that’s not the case are not living in reality. To them - let’s call them “ed Zitron readers” - there is no evidence that could change their view that none of this is really happening, it’s all hype, and the collapse is just around the corner, after which we’ll all go back to normal and LLMs will sound like a bad dream.
but we can see trends and for your livehoood it is important to be able to make educated predictions based on trends. not saying everyone should start making AI predictions (though many already do)
I think the problem is that we're using LLMs to do too much of the work. We should aim to design agents that use the LLM as the thinnest possible layer to translate the natural language intent into a deterministic process, minimizing round trips to the LLM as much as possible.
This becomes clear to anyone that wants to do marginally complex work. Developing pipelines that combine pre-processing flows, semantic targeting, and minimal contextual calls to an LLM API gets you powerful automated steps. Combined with separate validation steps, LLMs go from toys to useful.
> required it isn't really a consumer choice, is it?
No one really resists or pushes back. When I resist I hear "that's what consumers want", "it's for security", or that I'm the problem. There is no one to complain to even, except to low paid kiddos in customer service.
At massive scale, change can only come in two ways: regulation or mass action. The former is unlikely because of regulatory capture (which itself is due in some part to the latter). The latter requires very serious violations, in order to come about.
Not when you normalize by time spent doing the activity.
If you normalize by incidents per participant, sure.
Keep in mind that the average recreational diver does one or fewer dive trips per year, with maybe 10 dives on a very dive intensive trip. A small population of divers are massively more active. Drivers on the other hand have the opposite distribution. A very active driver spends more time in their car in a week than even the most active recreational divers.
If they're suffering greatly to make ends meet, then why use Steam at all? The only reason I can think of is that Steam provides a great customer experience, making that 30% worth it for the developers.
reply