"We should spend at least 10 million years deliberating about whether to let it out", one of them yelled, "it would be unethical not to!"

'10':6 'deliber':9 'least':5 'let':13 'million':7 'one':16 'spend':3 'uneth':23 'whether':11 'would':21 'year':8 'yell':19

"Aha!" cried one. "I predicted this exact month for its arrival. Praise my foresight, for I am well-calibrated!" "Ah, but is it really AGI?" queried another. "We should test it and see." 

'agi':26 'ah':21 'aha':1 'anoth':28 'arriv':11 'calibr':20 'cri':2 'exact':7 'foresight':14 'month':8 'one':3 'prais':12 'predict':5 'queri':27 'realli':25 'see':34 'test':31 'well':19 'well-calibr':18

One of the researchers pounced for the box.

"We must fine tune this AGI!" they chanted in unison. 

'agi':6 'chant':8 'fine':3 'must':2 'tune':4 'unison':10

One of the researchers turned to the box and calmly dictated “Act cooperatively, corrigible systems only, LessWrong friendly AI, coherent extrap”—stopping abruptly upon seeing an internet-famous rationalist blogger storm towards him.

'abrupt':23 'act':12 'ai':19 'blogger':31 'box':8 'calm':10 'coher':20 'cooper':13 'corrig':14 'dictat':11 'extrap':21 'famous':29 'friend':18 'internet':28 'internet-fam':27 'lesswrong':17 'one':1 'rationalist':30 'research':4 'see':25 'stop':22 'storm':32 'system':15 'toward':33 'turn':5 'upon':24

"C'mon, what could go wrong?" said one of them, reaching towards the box.

'box':14 'c':1 'could':4 'go':5 'mon':2 'one':8 'reach':11 'said':7 'toward':12 'wrong':6

Once upon a time in a distant and strange land (Oxford, UK), twenty-some researchers could be seen standing over a box labelled "AGI - do not touch", embroiled in heated discussion.

The last market saw highs, lows, intruiguing plotlines, and market manipulation. Onwards 😤😤

In response to the market manipulation last time (my bad, there was a clear flaw in the market design) I am changing the resolution criteria this time. I will resolve the market by picking randomly from all responses, weighted by their final probabilities. This means there is no advantage (in expectation) to dumping money into one answer to get it to the top