This market resolves six months after a market of the form "Will we get AGI before 2XXX?" resolves YES. It will be extended indefinitely until then.
At the six month mark, I will run a poll of manifold users asking them (1) whether they are able to consistently and reliably connect printers and (2) whether AGI is able to consistently and reliably connect printers for them. The poll will close after 5 business days or after 100 responses. If there are fewer than 20 respondents, this question will resolve N/A. If answers to (1) are 0% YES, then this market resolves YES. If answers to (2) are >50% YES, then this market resolves YES. Otherwise, this market resolves NO.
Caveats:
2D printers only.
I will not trade in this market.
1,000
3.00I think you should clarify this question somewhat.
My understanding is that you wanted the software to be able to both connect the cables and figure out why the printer isn't printing anything. But there are comments here talking about Bluetooth connections.
Are you expecting the printers to be wireless and be functional and be turned on by a human, or does the AI have to do everything?
Our human level AGI cannot repel suckitude of this magnitude!
This is more about printers than AGI, and not even an AI (at human level) can be expected to e.g. fire up a kernel debugger to fix some vendors dodgy Bluetooth driver. (By the time one could be expected do that kind of thing on its own we’d have other problems I think — though less certain after being surprised by GPT-3 wrt. natural language understanding that previously assumed would be AI-complete)
Many people make it sound as if AGI will be some magical thing that will solve every single problem.
I expect that, on the contrary, if there were some point where a system was declared "AGI" (which there is not), people will be disappointed at how little impact it actually has on the physical world. The thing might do a great job at coding, but it's still going to be stuck with stupid things like hardware faults and the user not knowing which buttons to press on the printer to clear the paper jam.
In response to another question, I pointed out that if "superintelligence" was achieved tomorrow, the best it could do is print out a 1000-page manual for how to 100% cure all forms of cancers without recurrence. Then, it would be years until even a single life was saved, as the biggest hinderances in the world are already human regulations, not technical roadblocks, even before AGI.
@SteveSokolowski I mean, I'm not a superintelligence, but if I had a cure for all forms of cancer but $0 and no body, I could save one life pretty quickly by getting a dying multimillionaire who felt they had nothing to lose to give me a bunch of money in a desperate bid for continued life, use that to get whatever needed creating created, save their life, hopefully they're motivated to spread the word but even if not I've got one success story. Build on that in a country where regulations don't exist or regulators can be bribed, until I have a few thousand lives saved and a small scale RCT done, then appeal to EAs to scale it up. I could even reuse a snake-oil salseperson's list of leads to find the first lucky winner, I'm sure I could make "cures all cancers" sound like an appealing enough scam that someone would want in, and sell me their list of desperate cancer-havers in exchange for a portion of the profits.
If the superintelligence is dumb enough to not do it because someone says "please don't, that's against the rules" there's a hole in this plotline - we don't have alignment techniques even close to good enough to prevent a superintelligence from doing whatever it wants yet.
@equinoxhq This question is to avoid the error condition where AGI appears and solves the world's problems so magically that the solutions are not attributed to AGI. Also, the poll will be of Manifold users. You could eliminate the possibility of poll (1) resolving 0% YES by inviting your elderly relatives to Manifold. :)
@brp perhaps I have misread. My reading was:
Question (1): are you able to consistently and reliably connect a printer?
My answer: yes
Therefore, % of manifold users who will answer yes > 0% ?