Will AI Really Make Information More Equal?
Recorded in Chinese
Two ideas—value in mistaken information and information privilege—make me reconsider sharing uncertain research, stating its limits, and the responsibility a person accepts by publishing it.
00:02 Recently, I encountered two ideas
00:05 that deeply affected me.
00:07 The first was
00:12 that incorrect information also has value.
00:16 Once someone knows a piece of information,
00:20 even if it is wrong,
00:22 that information still has value.
00:25 Someone can follow the wrong information
00:28 and find other information.
00:31 Or,
00:33 when they act on
00:35 that wrong information,
00:36 they may discover that it is wrong.
00:39 In that process, they encounter
00:41 things they would never have encountered
00:46 before seeing the wrong information.
00:49 So,
00:51 when I heard that sentence,
00:53 I thought, wow,
00:56 incorrect information
00:57 can be understood that way too.
01:01 The other statement came from
01:06 one person asking another
01:09 why he showed himself on social media.
01:13 He answered
01:15 with
01:20 three or four reasons, I think.
01:24 The final reason was that he wanted
01:30 to use social media
01:33 to record the information he learned
01:37 and share it.
01:39 It might attract similar people,
01:41 but the information he shared
01:44 might also help others,
01:46 or close, to some extent,
01:50 information gaps for people who lacked it.
01:52 His field was
01:58 quite complex.
01:59 Or in his words,
02:01 the group he belonged to
02:04 had a certain information privilege.
02:06 So he wanted to use social media
02:08 to share
02:10 his research in the field,
02:14 or things like
02:19 lessons from building a company.
02:22 In this way, more people
02:25 could learn what happens in his sector
02:28 and what kinds of things
02:29 people in his field do.
02:32 The exact phrase he used was:
02:37 his original words were,
02:43 "make access to information more equal."
02:49 When I heard him say
02:54 that he had information privilege,
02:57 wow, do you know how I reacted
03:00 to that phrase?
03:01 I rarely hear
03:09 a person like him,
03:12 someone in the same position,
03:14 someone who holds information
03:19 in a particular field,
03:23 publicly say they have information privilege.
03:28 So when I heard what he said,
03:31 I thought, wow,
03:32 I really admire this person.
03:36 This example reminded me of
03:39 around mid-March.
03:43 Anyone who follows X will remember
03:48 that Andrej Karpathy
03:52 was involved.
03:55 He used AI to conduct some research,
03:58 based roughly on
04:00 US Bureau of Labor Statistics data,
04:06 about AI's effects on different occupations.
04:11 I did not study the findings closely,
04:14 but I knew the event happened.
04:16 After it
04:19 came out,
04:20 it caused a huge stir.
04:21 My own sense was that the stir
04:24 was mainly on X,
04:28 within Silicon Valley circles.
04:33 Then another person appeared.
04:35 He used the same method as Karpathy
04:41 to conduct another study,
04:45 also using AI.
04:47 But this study had a broader scope,
04:51 apparently covering the world.
04:53 Then he published
04:56 the findings, either on X
04:58 or on his own Substack.
05:01 I do not remember which.
05:06 After the findings were published,
05:08 they caused an enormous stir,
05:11 perhaps
05:14 even larger than the reaction
05:17 to Karpathy's findings,
05:19 or something like that.
05:21 But under pressure, he later
05:23 removed
05:26 the research findings.
05:29 When I saw these two cases,
05:32 I thought...
05:35 To some extent,
05:36 I understand why he deleted it.
05:38 He felt that what he shared
05:41 had been made with AI.
05:43 He himself may not have had enough confidence,
05:47 judgment, or predictive ability
05:51 to say how much the AI's result
05:55 would become reality in the future.
06:01 But I keep feeling
06:03 that no matter how much
06:06 commotion such a report causes,
06:08 it still needs to exist.
06:14 At a broader level,
06:17 such a report might
06:20 warn people in certain professions,
06:25 giving them an early warning.
06:27 Then they could make plans.
06:30 If
06:32 the reality in this research report...
06:35 Oh,
06:35 if the report becomes reality,
06:37 there will surely be things like
06:39 unemployment
06:40 and similar problems.
06:42 Those workers
06:44 could plan responses in advance.
06:49 But because,
06:51 as he said,
06:52 it caused such a large stir,
06:54 he could not bear the pressure
06:56 and deleted it.
07:00 I think that is a shame.
07:02 To some extent,
07:05 knowing how to use AI for this research,
07:08 getting these results,
07:10 and knowing how to analyze them
07:12 are abilities held
07:14 by only a small group of people.
07:17 But after gaining those abilities
07:19 and the information that comes with them,
07:24 they choose to keep it
07:26 to themselves
07:30 instead of publishing it.
07:31 To some extent, that also reflects
07:38 information privilege.
07:39 So at the time I thought,
07:44 even if he followed Karpathy's example,
07:46 and caused a huge stir in Silicon Valley
07:50 or on X,
07:52 he could still publish all the data
07:55 from his research on a webpage.
08:01 But before the data,
08:03 or before releasing it,
08:05 he could provide context:
08:09 how he obtained the data
08:12 and what it meant.
08:15 He could disclose every limitation
08:18 and assumption
08:20 before the data.
08:21 That way,
08:25 the data would, to some extent, carry a disclaimer.
08:28 But many people quote things out of context.
08:34 They remove all the limits and assumptions
08:40 and share only the report itself.
08:44 Of course, I think
08:49 that is the audience's problem.
08:51 It has little to do
08:54 with the person who made the data public.
08:57 But if he feels unable
09:00 to bear what publishing the data brings,
09:03 such as pressure,
09:05 or...
09:07 what else?
09:09 If he cannot bear
09:12 what happens if the report is wrong,
09:15 or if AI turns out to be a bubble,
09:17 and his public image collapses,
09:20 or something like that,
09:22 if he cannot bear
09:23 those consequences,
09:29 then to some extent, he also...
09:34 I do not know how to put it.
09:36 There is a book called
09:42 Something at Stake,
09:43 or something like that.
09:47 Its English title is Skin in the Game.
09:51 It is about
09:54 a public figure...
09:56 This is only my understanding.
09:58 It is about how a public figure
09:59 bears some social responsibility, and...
10:07 For anything more specific,
10:09 I would have to read the book.
10:11 I have not read it.
10:12 I only know
10:13 that the book exists.
10:16 So information privilege reminded me
10:19 of this book,
10:21 and I spoke about it briefly.
10:24 Oh, information privilege
10:26 made me think of these two public figures,
10:29 and then of this book.
10:31 I just feel that
10:34 information,
10:35 even with the arrival of AI,
10:38 still will not
10:42 create the kind of
10:44 equal access to information we imagine.
10:46 Just as the arrival of the internet
10:48 did not make equal access to education real.