20 worst things gov can do in ai era - from 2025report 40 year observations

.

 

1 worst thing gov can do is compound bad infrastructure- back in 1920! us jones bill has resulted in at least 80 years of us failing to integrate superport into (trains roads planes digital highways) -how can public servants justify this law that destroys opportunities for poverty 99% to keep less 1% lobby happy

2 in intelligence age local optimisation of education in ai world should be every democratic govs top priority- again usa is a worst case example

3 empowering as much local health actions as possible is 3rd great use of ai that usa hasnt begun to celebrate

4 why not silicon valley everywhere - see our 1982 paper  (and far east coastal supply chains everywhere)- why hasn't everywhere understood west coast usa combined with far east advanced manufacturing supply chains

more coming soon

comments on Athey presentation Asia Development Bank Institute

46:06

Indermit Gill, Chief Economist World Bank, so here are

46: my impressions of Susan's talk on AI. Uh the first impression that I got was that

46:22

governments  will always be playing catchup.  they will be slow.  so for

46:31

example she said that AI will change the costs of digitization

46:36

and governments will be slower in realizing the benefits that

46:49

can come from AI for providing services. The second for example she said was AI

46:54

will change the distribution of of activity and of profits and of wage

46:59

losses and wage gains and governments will be slow in adopting industrial and

47:06

competition policy and helping workers adjust and move. then the other

47:12

thing that struck me as she was talking was that

47:18

past waves of technology 

47:23

may may actually lull you into a slower response than will be needed because of

47:29

AI, right? Uh so AI will be coming faster, the waves are going to be higher

47:36

and they're going to be coming at a faster speed too. So that's the first impression that I got.

47:42

The second impression that I got from what she said was that it seems that economists and other

47:49

analysts will be slower in understanding the effects of AI. So we'll 

47:55

actually won't be able to help  governments nearly as much as we might have in the past

48:03

because she mentioned about the nature of the technology is such that feedback loops abound are hard to model 

48:13

 that you don't have well developed general equilibrium models

48:19

for these kind of technologies. and you have to get into uh things

48:27

like whether or not the distribution effects should  be included

48:33

and whether or not the cost reductions are passed on

48:39

to consumers. So all of these things all of these things will make it actually very difficult. There'll be one more

48:45

thing that will that will actually  present a challenge to economists which is a divergence between

48:53

what is good for a firm and what is good for a country right and I'll end with an anecdote that I heard when I was in

49:01

India and it it had to do with a food delivery service and it's called Zumato

49:06

I think and what Zumato was saying was that or people from Zumato were saying that that  that when they deliver

49:15

services.  if you have to have a human intervention,

49:20

it costs about 70 rupees more and that human intervention  for any

49:26

transaction is generally provided by a college educated worker.

49:32

 but now what they were doing was because of cost and

49:38

because of competitive pressures uh they found that they had to sort of move to AI. Now if you move to AI, you're

49:46

replacing that college worker because AI comes in when you need a human intervention instead of that. So AI

49:52

costs a lot less about onetenth less than what that college worker would cost. But the thing is that while the

50:00

college worker  got a wage and as a result  in this case AI

50:07

would make uh would actually cut that kind of employment. That's one shot. The

50:13

second shot was that that India would actually have a transfer because it

50:19

would have to pay a royalty to a foreign AI company. And the reason why they told

50:26

us this was because they said that we in our last world development report in the

50:31

2024 world development report we were saying that what countries like India should be doing is trying not to develop

50:39

technologies of their own but to bring them in from outside and what these people were saying was

50:45

this is going to become very very costly unless you develop the technologies and own the intellectual property to them

50:53

you're going to end up with huge uh uh huge losses. So when Susan said that one

51:00

has to sort of think about what's good for the firm which is Zomato and what is good for India is suddenly different but

51:07

those were my quick reactions. There were two or three other things that struck me as she was talking. One of

51:13

them was and these are scary thoughts.  one

51:19

thing was that I got a sense from what you were saying was that AI is making war cheap to start  So

51:35

you'll probably get more wars. The second one she said was that uh the

51:42

outcome might be that  that uh these AI benefits show up in welfare gains

51:49

rather than productivity. and  that perhaps what AI

51:58

doing for software AI is doing for software what the CD or

52:07

the DVD or uh music and things like that

52:12

is actually that it to democratize that system. But those are my I think the

52:19

main thing that I thought was we seem to be very unprepared for this.............