> AI models have gone from barely being able to write a coherent line of code to writing most of the code at major AI companies.
Gasoline has gone from barely being able to power stationary farm machines to now being the fuel that underpins our entire economy. So, great news all around, right?
> which predict an exponential increase
And was that actually delivered?
Real question: If a model goes from 80% accurate to 85% accurate is that an exponential increase in "cognitive capabilities?" Are we considering training costs and effort?
I think people are typically referring to the task-completion time horizon at a fixed success rate [1]. That has had pretty robust exponential scaling for many years now.
If you give a language model, empowered through an agent, the ability to publish information on your behalf, and it publishes false information which causes either direct or even indirect harm, and you fail to correct it, then yes.. by every conceivable definition already in law.. you are a criminal.
If you knew this was all possible and you did it anyways for personal gain then you are additionally negligent which may add aggravation to your charges.
If they do so knowingly, and harm is caused, then yes. Are you suggesting we should give them a pass from years of acquired jurisprudence simply because they hold a particular title?
And what institution gives out the licenses for journalists and scientists? Is it revokable?
> the definition by analogy to what aircaft autopilot does
It's a bad analogy. Autopilot just maintains the aircraft in some state, then there's the flight director which maintains the flight path, and you can connect or disconnect the two at will. When connected the director can change the autopilot state.
To use the flight director you must fully specify your flight. The weight, the fuel, the weather, expected winds, takeoff and landing runway length, runway conditions, expected brake demand, as well as every single waypoint you're going to cross and the expected state at that crossing.
> what the human metaphor means
We learned after high levels of cockpit automation that maintaining situational awareness was still required. Pilots are freed of some stress during high workload portions of the flight, provided they planned correctly in advance, and that zero changes to their flight plan (not likely) occur.
As a result pilots are mostly told and trained to hand fly the plane during take offs and landings if the weather allows for it. You should only use high levels of automation if the situation demands it.
more of "we whined and cried and screamed that we needed new budget in order to buy these tools or we would literally die. now we have them, they don't work as well as we hoped, they aren't leading to productivity gains, and they're actively alienating our workforce and users alike. we're so screwed we literally have no idea how to do reverse this."
I can make more money doing HVAC but I'm tired of being on hot roofs.
> the path forward is a lot murkier.
If you're just here for the money go somewhere else. If you're here because you love computer science then ignore these people and do the work. If you can't find a company get a dayjob and do it for yourself.
Gasoline has gone from barely being able to power stationary farm machines to now being the fuel that underpins our entire economy. So, great news all around, right?
> which predict an exponential increase
And was that actually delivered?
Real question: If a model goes from 80% accurate to 85% accurate is that an exponential increase in "cognitive capabilities?" Are we considering training costs and effort?
reply