---
author: Joseph Wang
date: '2026-02-14'
guest: ''
layout: post.njk
source: https://www.youtube.com/watch?v=TnLmy49QtW0
speaker: Joseph Wang
tags:
  - ai-application
  - productivity
  - economic-impact
title: AI 驱动的生产力提升与 GDP 悖论
summary: 本期 Markets Weekly 探讨了 AI 技术对经济的深远影响。尽管 AI 显著提升了生产力，但其对 GDP 的影响却并非直接正相关。演讲者指出，AI 驱动的效率提升，例如利用 AI 辅助编辑书籍、分析大量金融文档等，虽然降低了成本并提高了服务质量，但这些收益往往难以在 GDP 数据中体现，因为它们并未创造新的需求，反而可能导致劳动力市场的结构性调整。这种现象类似于智能手机的普及，虽然产品质量不断提升，但价格并未显著上涨，从而对 GDP 的贡献有限。
insight: ''
draft: true
series: ''
category: ai-ml
area: tech-engineering
project: []
people: []
companies_orgs:
  - OpenAI
products_models:
  - GPT-4o
  - Cloud AI
media_books: []
status: evergreen
---
### AI 驱动的生产力提升与 GDP 悖论

本期 Markets Weekly 探讨了 AI 技术对经济的深远影响。尽管 AI 显著提升了生产力，但其对 GDP 的影响却并非直接正相关。演讲者指出，AI 驱动的效率提升，例如利用 AI 辅助编辑书籍、分析大量金融文档等，虽然降低了成本并提高了服务质量，但这些收益往往难以在 GDP 数据中体现，因为它们并未创造新的需求，反而可能导致劳动力市场的结构性调整。这种现象类似于智能手机的普及，虽然产品质量不断提升，但价格并未显著上涨，从而对 GDP 的贡献有限。

<details>
<summary>Original English</summary>
Hello
 my
 friends.
 Today
 is
 February
 14th


and
 this
 is
 Markets
 Weekly.
 So,
 this


past
 week
 was
 a
 very
 volatile
 week
 in


markets.
 As
 we've
 been
 discussing
 over


the
 recent
 weeks,
 it
 really
 seemed
 to
 me


that
 the
 market
 was
 losing
 momentum.


Now,
 this
 past
 week,
 we
 saw
 the
 S&P
 500


lose
 the
 50-day
 moving
 average.
 On


Friday,
 it
 came
 very
 close
 to
 losing
 the


100
 day
 moving
 average
 as
 well.
 That


kind
 of
 reminded
 me
 of
 a
 wise
 person
 who


traded
 through
 the.com
 boom
 and
 bust


that
 during
 the
 dot
 a
 lot
 of
 people
 went


broke
 selling
 it
 and
 a
 few
 years
 later


even
 more
 people
 went
 broke
 buying
 it.


Now
 I'm
 getting
 the
 sense
 just
 a
 guess


that
 we
 we
 probably
 have
 to
 test
 the
 200


day
 moving
 average
 around
 6500.
 Um


because
 we
 really
 this
 trend
 doesn't


doesn't
 look
 that
 great.
 This
 past
 week


we
 got
 a
 lot
 of
 tier
 one
 data.
 We
 got


the
 non-farmms
 payroll
 spread.
 We
 got


retail
 sales
 and
 we
 also
 got
 CPI.
 So


first
 let's
 talk
 about
 the
 data
 and


secondly
 I
 started
 using
 cloud
 AI
 last


week.
 Of
 course
 we
 saw
 huge
 amounts
 of


selling
 and
 the
 SAS
 stocks
 and
 even
 some


wealth
 management
 stocks
 everyone
 uh


that
 all
 the
 stocks
 that
 were
 perceived


to
 be
 disrupted
 by
 AI.
 So
 I
 had
 to
 try


the
 latest
 AI
 myself
 and
 I'm
 beginning


to
 think
 that
 AI
 is
 going
 to
 be
 very


good
 for
 bonds.
 All
 right,
 starting
 with


the
 data.
</details>

### 数据解读：就业市场与通胀的微妙信号

本周公布的关键经济数据，包括非农就业、零售销售和 CPI，呈现出复杂的信号。非农就业数据好于预期，但年度修正却抹去了去年创造的就业岗位。市场对修正数据并不敏感，更关注月度数据，认为就业市场正在改善，但同时也降低了美联储降息的预期。零售销售数据低于预期，可能受到季节性调整的影响。CPI 数据虽然有所下降，但仍高于美联储的目标。

<details>
<summary>Original English</summary>
So
 we
 got
 a
 huge
 three
 huge
 prints
 last


week.
 First
 of
 all,
 of
 course,
 the


non-formms
 per
 print.
 The
 January
 print


was
 delayed
 until
 last
 week
 and
 it
 was


much
 better
 than
 expected.
 However,
 it


also
 came
 with
 the
 annual
 benchmark


revisions
 and
 the
 benchmark
 revisions


basically
 erased
 1
 million
 jobs
 created


last
 year.
 So,
 as
 we've
 been
 discussing


before,
 the
 market
 really
 doesn't
 care


about
 revisions.
 It
 only
 cares
 about
 the


monthly
 prints.
 If
 the
 monthly
 prints


last
 year
 were
 accurate,
 we
 would
 have


had
 zero
 or
 negative
 monthly
 prints.
 Uh


but
 and
 that
 would
 have
 been
 very


upsetting
 to
 the
 markets.
 But
 um
 now


that
 we
 a
 year
 later
 we
 revised
 them
 all


away,
 the
 market
 really
 doesn't
 care.


What
 it
 does
 care
 about
 though
 is
 that


it
 seems
 like
 the
 job
 market
 is


improving.
 So
 the
 January
 report
 was


better
 than
 expected
 and
 maybe
 it
 will


be
 revised
 away
 a
 year
 from
 now
 but
 at


the
 moment
 it's
 better
 than
 expected
 and


the
 market
 is
 going
 to
 perceive
 that
 to


be
 good
 and
 that
 uh
 the
 economy
 is
 less


likely
 to
 go
 into
 recession
 but
 it
 also


of
 course
 reduces
 the
 odds
 of
 Fed
 rate


cuts.
</details>

### AI 对行业的影响：以法律服务为例

演讲者分享了使用 AI 辅助工作的经验，例如利用 AI 查找联邦储备的讲话安排、编辑书籍以及生成书面材料。AI 在处理重复性任务方面表现出色，例如合同审查和修改，可以显著提高律师的工作效率。然而，AI 的应用也可能导致相关岗位的需求减少，从而对就业市场产生影响。

<details>
<summary>Original English</summary>
Now
 we
 also
 got
 the
 retail
 sales
 and
 CPI


last
 week
 but
 before
 we
 talk
 about
 that


we
 have
 to
 mention
 a
 little
 bit
 about


seasonal
 adjustments.
 So
 throughout
 the


year
 data
 is
 uneven
 because
 of
 seasonal


effects.
 For
 example,
 uh
 people
 you
 tend


to
 do
 a
 lot
 of
 shopping
 in
 November
 and


December.
 That's
 a
 big
 holiday
 season.


So
 retail
 sales
 there
 are
 going
 to
 be
 a


lot
 more.
 And
 in
 January,
 people
 tend
 to


reset
 their
 p
 to
 reset
 their
 prices.
 For


example,
 if
 you're
 a
 company
 and
 you


want
 to
 raise
 or
 lower
 prices
 on
 your


products
 or
 services,
 you
 tend
 to
 do


that
 in
 the
 beginning
 of
 the
 year.
 So


there
 is
 some
 seasonality
 to
 how
 these


this
 data
 is
 computed
 and
 to
 make
 data


more
 comparable
 monthtomonth
 uh


economists
 or
 statistitians
 tend
 to
 do


these
 seasonal
 adjustments.
 So
 keep
 that


in
 mind
 when
 thinking
 about
 the
 data.
 So


for
 retail
 sales
 the
 data
 was
 much
 lower


than
 expected.
 they
 were
 basically
 flat


monthtomonth
 and
 some
 people
 think
 that


could
 be
 due
 to
 aggressive
 seasonal


adjustments
 which
 are
 difficult
 to
 make


because
 of
 course
 the
 economy
 is


changing
 um
 behavioral
 changes
 and
 so


sometimes
 seasonal
 adjustments
 can
 be


overdone
 or
 underdone.
 Now
 moving
 on
 to


CPI
 data
 CPI
 was
 better
 than
 expected


but
 still
 at
 at
 the
 end
 of
 the
 day
 it's


still
 about
 you
 know
 2
 and
 a.5%
 above


the
 Fed's
 target.
 Uh,
 one
 thing
 to
 note


of
 course
 is
 that
 companies
 do
 raise


their
 prices
 in
 January
 and
 so
 these


seasonal
 adjustments
 make
 the
 January


print
 uh
 less
 accurate
 than
 than
 other


prints
 as
 well.
 So
 all
 in
 all
 though


this
 is
 I
 know
 data
 that
 is
 you
 know


kind
 of
 positive
 right
 you
 have
 better


than
 expected
 job
 growth
 and
 you
 have


inflation
 that's
 coming
 down
 but
 of


course
 the
 market
 didn't
 seem
 to
 really


care
 about
 this
 they
 were
 really
 more


cared
 about
 uh
 what
 seemed
 to
 be


happening
 in
 the
 AI
 world
 and
 that
 is


the
 AI
 is
 cannibalizing
 uh
 current


companies.
 So
 that
 brings
 us
 to
 our
 next


topic.
</details>

### AI 时代的投资策略：拥抱债券

演讲者认为，AI 带来的生产力提升和潜在的经济下行风险，将推动资金流向避险资产。AI 可能会导致失业率上升、经济增长放缓，从而促使美联储采取更宽松的货币政策，降低利率。因此，AI 时代最值得关注的投资机会可能在于债券市场。

<details>
<summary>Original English</summary>
So,
 I've
 talked
 about
 AI
 before.
 I


started
 trying
 to
 use
 AI
 to
 help
 me
 with


my
 workload.
 And
 in
 the
 beginning,
 it


was
 very
 disappointing.
 So,
 I
 asked
 AI


to
 do
 basic
 things
 like,
 yo,
 tell
 me


when
 are
 speaking.
 And
 it
 really


couldn't
 do
 that.
 It
 was
 amazing
 to
 me


that
 I
 asked
 it
 to
 go
 on
 the
 Federal


Reserve's
 website
 to
 look
 up
 who's
 going


to
 be
 speaking
 this
 week,
 and
 it
 could


not
 do
 that.
 So,
 that
 was
 very


disappointing.
 But
 last
 week
 we
 had
 the


release
 of
 the
 latest
 cloud
 model
 and
 we


also
 had
 carnage
 in
 a
 lot
 of
 software
 as


a
 service
 companies.
 The
 market
 or
 at


least
 people
 who
 follow
 this
 was


thinking
 that
 the
 latest
 AI
 is
 really


going
 to
 be
 very
 destructive
 to
 the


value
 of
 these
 companies.
 Maybe
 there's


more
 competition.
 Uh
 maybe
 uh
 consumers


are
 going
 to
 u
 move
 to
 bespoke
 AI


created
 vibecoded
 platforms
 and
 so


forth.
 We
 even
 saw
 some
 wealth


management
 companies
 like
 Schwabs
 go
 up


aggressively
 because
 maybe
 AI
 can
 uh


provide
 tax
 efficient
 strategies
 in


place
 of
 wealth
 advisors.
 So
 again
 I
 had


to
 try
 this
 for
 myself.
 So
 I
 signed
 up


for
 the
 cloud
 max
 plan.
 So
 that's
 like


$100
 a
 month
 and
 got
 to
 use
 it.
 So
 what


did
 I
 use
 it
 for?
 Well
 as
 you
 guys
 know


I
 am
 writing
 the
 second
 edition
 to
 my


book
 central
 banking
 101.
 Uh
 the
 last


one
 was
 5
 years
 ago
 and
 so
 between
 then


and
 now
 a
 lot
 of
 things
 happened.
 So


I've
 updated
 all
 the
 data,
 added
 a
 whole


bunch
 of
 new
 sections
 about,
 you
 know,


the
 Silicon
 Valley
 bust,
 uh,
 Silicon


Valley
 bank
 bust,
 you
 know,
 Treasury


buybacks,
 QT,
 all
 that
 stuff.
 So
 I


basically
 updated
 everything
 and
 I
 felt


like
 I
 was
 basically
 done
 with
 the


manuscript.
 So
 usually
 what
 I
 would
 do


is
 I
 would
 hire
 an
 editor,
 a
 freelancer,


and
 pay
 him
 a
 few
 thousand
 to
 go
 through


this.
 And
 I
 would
 also
 hire
 some
 artists


to
 make
 a
 new
 book
 cover
 for for
 a
 few


hundred.
 And
 these
 are
 people
 that
 I


would
 source
 through
 upwork.com.


basically
 sites
 that
 post
 for


freelancers.


But
 this
 time
 I
 decided
 I
 would
 give
 AI


a
 try
 since
 that's
 what
 all
 the
 cool


kids
 are
 doing
 right
 now.
 So
 I
 uploaded


everything
 into
 cloud
 and
 asked
 him
 to


kind
 of
 you
 know
 be
 my
 editor
 and


proofread
 it.
 It
 said
 he
 said
 that
 it


would
 take
 6
 hours
 and
 so
 uh
 I
 waited


for
 6
 hours
 and
 what
 I
 got
 back
 was
 you


know


I
 guess
 he
 gave
 me
 like
 a
 few


suggestions
 updated
 some
 data
 points
 and


things
 like
 that
 some
 very
 general


stuff.
 I
 thought
 it
 was
 actually
 pretty


uh
 you
 know
 just
 meh
 not not
 that


special
 but
 you
 know
 maybe
 it's
 my


fault.
 Maybe
 I'm
 not
 prompting
 cloud


correctly.


So
 uh
 this
 is
 what
 I
 did.
 I
 prompted
 him


in
 a
 more
 precise
 way.
 I
 said
 could
 you


look
 at
 the
 reviews
 that
 I
 that
 are


posted
 on
 good
 readads
 on
 Amazon
 and


there
 are a
 lot
 of
 reviews
 and
 can
 you


bearing
 bring
 bearing
 that
 feedback
 in


mind
 can
 you
 go
 through
 the
 manuscript


and
 you
 can
 you
 give
 me
 some
 suggestions


and
 also
 just
 you
 know
 broad
 suggestions


as
 to
 what
 where
 I
 could
 do
 better
 and


so
 forth
 like
 that
 so
 this
 time
 it
 only


took
 you
 know
 a
 few
 minutes
 maybe
 was


it's
 already
 in
 his
 me
 memory
 or


something
 like
 that
 and
 he
 came
 back


with
 actually
 something
 that
 was
 pretty


helpful
 gave
 me
 a
 list
 of
 suggestions


gave
 me
 a
 separate
 word
 document


detailing


And
 you
 know
 looking
 through
 it
 some
 of


them
 were
 not
 that
 great
 stylistic
 of


course
 some
 of
 them
 were
 pretty
 good.


One
 of
 them
 for
 example
 was
 that
 in
 your


reviews
 people
 suggest
 you
 having
 a


glossery
 glossery
 so
 that
 they
 can
 look


back
 look
 up
 these
 acronyms
 QT
 FOMC
 uh


stuff
 like
 that.
 That
 was
 pretty
 good


suggestion.
 And
 some
 of
 them
 also


suggested
 that
 hey
 maybe
 you
 could
 have


summary
 chapter
 bullet
 points.
 you
 know,


kind
 of
 like
 a
 textbook,
 at
 the
 end
 of


the
 textbook,
 you
 kind
 of
 give
 you
 a


summary
 as
 to
 what
 you
 should
 be
 what


you
 should
 have
 learned
 from
 this


chapter.
 I
 thought
 that
 was
 a
 pretty


good
 suggestion
 as
 well.
 And
 so
 I
 asked


Cloud,
 could
 you
 do
 that
 for
 me?
 And


like
 just
 like
 that,
 they
 created


wonderfully
 made
 glossery,
 wonderfully


made
 bullet
 points,
 and
 I
 thought
 that


was
 that
 was
 pretty
 cool.


I'm
 guessing
 that
 it
 could
 do
 more
 to


help
 me
 with
 my
 manuscript,
 but
 I'm


probably
 not
 prompting
 it
 well
 enough.


It
 did
 tell
 me
 though
 that
 it
 was
 very


good
 and
 well
 written
 and
 so
 I
 think


it's
 just
 part
 of
 its
 job.
 Um,
 like
 any


employee,


um,
 when
 I
 went
 to
 ask
 it
 about
 helping


me
 create
 a
 new
 cover
 for
 my
 updated


version,
 I
 gave
 it
 my
 prior
 cover
 and,


you
 know,
 gave
 it
 some
 ideas
 and
 it


immediately
 basically
 like
 Grock
 gave
 me


a
 number
 of
 ideas
 and
 ways
 to
 pursue
 and


that's
 actually
 exactly
 what
 my
 like
 a


real
 human
 cover
 designer
 would
 have


done.
 he
 would
 have
 given
 me
 a
 few


design
 choices
 and
 then
 I
 would
 have,


you
 know,
 decided
 which
 one
 to
 go
 and


and
 how
 to
 advance
 it.
 So,
 I'm
 probably


just
 going
 to
 use
 cloud
 as
 an
 editor.
 I


think
 it's
 really
 good
 enough
 and


honestly
 I
 I
 don't
 know
 some
 of
 the


human
 it's
 hard
 for
 me
 to
 know
 if
 the


human
 editors
 that
 I
 hire
 are
 good
 or


not
 until
 after
 the
 fact.
 But
 for
 the


artistic
 stuff,
 I
 think
 I'm
 still
 going


to
 hire
 a
 real
 designer.
 I
 think
 having


that
 human
 artistic
 touch
 is
 something


that
 I'm
 not
 really
 sure
 that
 Cloud
 can


fully
 do
 right
 now.
 The
 covers
 he's


shown
 me
 is
 are
 very
 good.
 Um,
 but
 I


guess
 I'm
 just
 not
 confident
 enough
 in


my
 own
 judgment
 and
 prefer
 someone
 who


who
 I
 a
 proper
 artist
 to
 do
 that.
 But


that
 immediately
 got
 me
 thinking
 as
 to


just
 what
 the
 implications
 of
 this
 are.
</details>