All of johnvon's Comments + Replies

johnvon20

'when things get scary' when then? 

johnvon*4221

This interview was terrifying to me (and I think to Dwarkesh as well), Schulman continually demonstrates that he hasn't really thought about the AGI future scenarios in that much depth and sort of handwaves away any talk of future dangers. 

Right off the bat he acknowledges that they reasonably expect AGI in 1-5 years or so, and even though Dwarkesh pushes him he doesn't present any more detailed plan for safety than "Oh we'll need to be careful and cooperate with the other companies...I guess..."

6Zvi
Here is my coverage of it. Given this is a 'day minus one' interview of someone in a different position, and given everything else we already know about OpenAI, I thought this went about as well as it could have. I don't want to see false confidence in that kind of spot, and the failure of OpenAI to have a plan for that scenario is not news.

This is actually what social media is for, but you don't have to fill out a questionnaire. You also don't have to out yourself as being so lonely and without friends that you're using a special matchmaking service to find new friends, this in itself could be unattractive to new acquaintances. 

2Adam Zerner
Social media doesn't do the matchmaking stuff very much though, does it?

Just a reminder that if everyone were asexual, the species would become extinct in short order :). 

It's still possible to feel romantic attraction without sexual attraction, and to want kids without wanting the act that created them for its own sake. While not the norm, this is a way I can imagine instantiating the thing Oxytocin suggested.

You shouldn't be comparing only HOMICIDES, the point of this comparison is that you are much more likely to die (from any cause, most likely from an automobile accident) than you are to die while taking public transit (from any cause). Homicide may be the most likely cause of death while taking public transit (or it may not), but I imagine the odds of getting killed in an accident while driving present a much greater risk. 

2jefftk
The statistics I'm comparing are public transit homicides vs driving accident deaths. So if you have a heart attack, get shot while driving, or your train disastrously derails that's not included. I think this is fine, though, because including these far less common kinds of death is very unlikely to change the result?

The most efficient way to do this would be a panopticon-style structure where one guard (‘assistant’) could oversee dozens or hundreds of workers.

Mako i just read your response post.

this proposed solution reminds me very much of some of the solutions the software and music industries proposed in order to stop piracy. unfortunately none of these worked, or were practical enough to put into widespread use. and of course the adoption has to be UNIVERSAL to be effective.

1mako yass
They're related fields. For various reasons (some ridiculous) I've spent a lot of time thinking about the potential upsides of the thing that Richard Stallman called Treacherous Computing. There are many. We're essentially talking about the difference between having devices that can make promises and devices that can't. Devices that have the option of pledging to tell the truth in certain situations, and devices that can tell any lie that is possible to tell. I think we have reason to believe Trusted Computing will be easier to achieve with better (cheaper) technology. I also think we have reasons to hope that it will be easier to achieve. Really, Trusted Computing and Treachery are separate qualities. An unsealed device can have secret backdoors. A sealed device can have an open design and an extensively audited manufacturing process. I'm not sure what you're getting at with the universality concern. If a work could only be viewed in theatres and on TC graphics hardware with sealed screens (do those exist yet), it would still be very profitable. They would not strictly need universal adoption of sealed hardware.