Skip to content

How Does This End?

Three arguments about this technology get run as one, and they are not the same argument. What you expect it to become, how much proof you want before it reaches people, and whose hands you want it in are separate questions — which is why somebody can expect a plateau and still want licensing, or expect the world to change and still want the weights published.

Thirty-six statements, about six minutes, and a point inside a cube. Answer for what you actually think, not for what the people you agree with are supposed to think.

36 questions

Answered 0 of 36

  1. 1 Waiting for certainty costs more than the mistakes it prevents.
  2. 2 The way to find out what a system does is to let people use it.
  3. 3 Before something is widely available, somebody outside the group that built it should have checked it.
  4. 4 Work that people train for years to do is going to be done without them.
  5. 5 A technology only a few people can reach ends up serving those few.
  6. 6 Whatever damage wide access does, narrow access does worse.
  7. 7 A capability held by three groups is easier to keep track of than one held by everyone.
  8. 8 A world where this is everywhere is a better one than a world where it sits in a few places.
  9. 9 Handing this to a few groups trades one set of risks for a worse set.
  10. 10 Some things are better held by people who can be held answerable for them.
  11. 11 Whatever this turns into, the world afterward is not a version of the one before it.
  12. 12 The demonstrations are chosen to be impressive, and the ordinary cases have not moved as much.
  13. 13 A thing that is good enough to help somebody is good enough to release.
  14. 14 A system should be held back until the ways it can fail are known.
  15. 15 I have revised my expectations upward more than once.
  16. 16 Something that can go badly wrong should be shown to work before it is put in front of people.
  17. 17 What is left is the part that has resisted everything so far.
  18. 18 The things people say a machine could not do keep turning out to be things it can.
  19. 19 Trying to get it right first mostly means getting it late.
  20. 20 The people best placed to handle this carefully are few, and that is not an accident.
  21. 21 I have watched this kind of excitement arrive before and settle into something smaller.
  22. 22 Anyone who wants to should be able to run the most capable systems themselves.
  23. 23 The limits people point to in these systems are the ones that will still be there.
  24. 24 Not every capability that exists should be within reach of anybody who wants it.
  25. 25 Ten years from now these systems will be doing much what they do today, done better.
  26. 26 Most of what goes wrong with a new technology is found by the people using it.
  27. 27 This ends up as one useful technology among several.
  28. 28 I would take a slower rollout for a smaller chance of a serious failure.
  29. 29 I would rather this ended up with a few groups that can be named than with everybody.
  30. 30 The burden belongs to whoever wants to release something rather than to whoever doubts it.
  31. 31 I would rather a useful thing arrived late than arrived unexamined.
  32. 32 A problem you can see in the wild gets fixed faster than one nobody has met.
  33. 33 Keeping this in few hands is a bet that those hands stay trustworthy.
  34. 34 Ten years from now these systems will be doing things that look nothing like what they do today.
  35. 35 The limits people point to in these systems are the ones that get removed next.
  36. 36 The number of hands a powerful thing passes through is the thing to keep small.