Question A:
Use of artificial intelligence is likely to lead to a substantial increase in problems associated with market power in digital markets.
Responses
Responses weighted by each expert's confidence
Question B:
Artificial intelligence offers substantial opportunities for new entrants into digital markets that have previously been concentrated.
Responses
Responses weighted by each expert's confidence
Question C:
Artificial intelligence is likely to be a highly concentrated industry, dominated by a handful of players.
Responses
Responses weighted by each expert's confidence
Question A Participant Responses
| Participant | University | Vote | Confidence | Bio/Vote History |
|---|---|---|---|---|
 Daron Acemoglu |
MIT | Bio/Vote History | ||
|
Microsoft-OpenAI and Google are heading towards leadership in (generative) AI, which could lead to an even more oligopolistic market. More structurally, the current approach AI relies on abundant data as inputs and the rush to collect and use data will further weaken competition
-see background information here -see background information here |
||||
 Joseph Altonji |
Yale | Bio/Vote History | ||
|
|
||||
 Alan Auerbach |
Berkeley | Bio/Vote History | ||
|
|
||||
 David Autor |
MIT | Bio/Vote History | ||
|
|
||||
 Abhijit Banerjee |
MIT | Bio/Vote History | ||
|
|
||||
 Marianne Bertrand |
Chicago | Bio/Vote History | ||
|
|
||||
 Markus Brunnermeier |
Princeton | Bio/Vote History | ||
|
|
||||
 Raj Chetty |
Harvard | Did Not Answer | Bio/Vote History | |
|
|
||||
 Judith Chevalier |
Yale | Bio/Vote History | ||
 David Cutler |
Harvard | Bio/Vote History | ||
|
|
||||
 Angus Deaton |
Princeton | Bio/Vote History | ||
|
|
||||
 Darrell Duffie |
Stanford | Bio/Vote History | ||
|
|
||||
 Aaron Edlin |
Berkeley | Bio/Vote History | ||
|
|
||||
 Barry Eichengreen |
Berkeley | Bio/Vote History | ||
|
|
||||
 Liran Einav |
Stanford | Bio/Vote History | ||
|
|
||||
 Ray Fair |
Yale | Bio/Vote History | ||
|
|
||||
 Amy Finkelstein |
MIT | Did Not Answer | Bio/Vote History | |
|
|
||||
 Pinelopi Goldberg |
Yale | Bio/Vote History | ||
|
|
||||
 Michael Greenstone |
University of Chicago | Did Not Answer | Bio/Vote History | |
|
|
||||
 Robert Hall |
Stanford | Did Not Answer | Bio/Vote History | |
|
|
||||
 Oliver Hart |
Harvard | Bio/Vote History | ||
|
|
||||
 Bengt Holmström |
MIT | Bio/Vote History | ||
|
Depends on regulatory actions
|
||||
 Caroline Hoxby |
Stanford | Bio/Vote History | ||
|
|
||||
 Hilary Hoynes |
Berkeley | Bio/Vote History | ||
|
|
||||
 Kenneth Judd |
Stanford | Did Not Answer | Bio/Vote History | |
|
|
||||
 Steven Kaplan |
Chicago Booth | Bio/Vote History | ||
|
I am confidently uncertain.
|
||||
 Anil Kashyap |
Chicago Booth | Bio/Vote History | ||
|
I can see arguments on both sides, maybe it leads to more disruption?
|
||||
 Pete Klenow |
Stanford | Bio/Vote History | ||
|
|
||||
 Jonathan Levin |
Stanford | Bio/Vote History | ||
|
|
||||
 Eric Maskin |
Harvard | Did Not Answer | Bio/Vote History | |
|
|
||||
 William Nordhaus |
Yale | Bio/Vote History | ||
|
Low entry cost and can threaten established players. Google declared "code red" when learning about.
|
||||
 Maurice Obstfeld |
Berkeley | Bio/Vote History | ||
|
|
||||
 Emmanuel Saez |
Berkeley | Did Not Answer | Bio/Vote History | |
|
|
||||
 Larry Samuelson |
Yale | Bio/Vote History | ||
|
Tech markets have tended to be highly concentrated, so one might expect AI to exacerbate this trend, but the effect of AI is general very difficult to predict.
|
||||
 José Scheinkman |
Columbia University | Bio/Vote History | ||
|
|
||||
 Richard Schmalensee |
MIT | Bio/Vote History | ||
|
Much to early to call.
|
||||
 Carl Shapiro |
Berkeley | Bio/Vote History | ||
|
|
||||
 Robert Shimer |
University of Chicago | Bio/Vote History | ||
|
|
||||
 James Stock |
Harvard | Did Not Answer | Bio/Vote History | |
|
|
||||
 Richard Thaler |
Chicago Booth | Did Not Answer | Bio/Vote History | |
|
|
||||
 Christopher Udry |
Northwestern | Bio/Vote History | ||
|
Strong returns to scale/network economies raise these challenges
|
||||
Question B Participant Responses
| Participant | University | Vote | Confidence | Bio/Vote History |
|---|---|---|---|---|
|
Daron Acemoglu |
MIT | Bio/Vote History | ||
|
Possible, especially with open source and the construction of mid-size, targeted AI models and AI stacks. But there is no evidence that these are going to be effective competition against large foundation models. Moreover, incumbents will continue to use M&A to thwart competitors
-see background information here |
||||
|
Joseph Altonji |
Yale | Bio/Vote History | ||
|
|
||||
|
Alan Auerbach |
Berkeley | Bio/Vote History | ||
|
|
||||
|
David Autor |
MIT | Bio/Vote History | ||
|
It offers opportunities. But they may be thwarted by incumbents, e.g., the way Facebook bought up potential rivals. This in part depends on antitrust enforcement, which in turns depends on who is President
|
||||
|
Abhijit Banerjee |
MIT | Bio/Vote History | ||
|
|
||||
|
Marianne Bertrand |
Chicago | Bio/Vote History | ||
|
|
||||
|
Markus Brunnermeier |
Princeton | Bio/Vote History | ||
|
|
||||
|
Raj Chetty |
Harvard | Did Not Answer | Bio/Vote History | |
|
|
||||
|
Judith Chevalier |
Yale | Bio/Vote History | ||
|
|
||||
|
David Cutler |
Harvard | Bio/Vote History | ||
|
|
||||
|
Angus Deaton |
Princeton | Bio/Vote History | ||
|
|
||||
|
Darrell Duffie |
Stanford | Bio/Vote History | ||
|
|
||||
|
Aaron Edlin |
Berkeley | Bio/Vote History | ||
|
|
||||
|
Barry Eichengreen |
Berkeley | Bio/Vote History | ||
|
|
||||
|
Liran Einav |
Stanford | Bio/Vote History | ||
|
|
||||
|
Ray Fair |
Yale | Bio/Vote History | ||
|
|
||||
|
Amy Finkelstein |
MIT | Did Not Answer | Bio/Vote History | |
|
|
||||
|
Pinelopi Goldberg |
Yale | Bio/Vote History | ||
|
|
||||
|
Michael Greenstone |
University of Chicago | Did Not Answer | Bio/Vote History | |
|
|
||||
|
Robert Hall |
Stanford | Did Not Answer | Bio/Vote History | |
|
|
||||
|
Oliver Hart |
Harvard | Bio/Vote History | ||
|
|
||||
|
Bengt Holmström |
MIT | Bio/Vote History | ||
|
History suggests turnover as well as long-run survivors
|
||||
|
Caroline Hoxby |
Stanford | Bio/Vote History | ||
|
|
||||
|
Hilary Hoynes |
Berkeley | Bio/Vote History | ||
|
|
||||
|
Kenneth Judd |
Stanford | Did Not Answer | Bio/Vote History | |
|
|
||||
|
Steven Kaplan |
Chicago Booth | Bio/Vote History | ||
|
Uncertain because it is possible for incumbents and / or new entrants to have some success.
|
||||
|
Anil Kashyap |
Chicago Booth | Bio/Vote History | ||
|
of course they may opt to sell to dominant incumbents too.
|
||||
|
Pete Klenow |
Stanford | Bio/Vote History | ||
|
|
||||
|
Jonathan Levin |
Stanford | Bio/Vote History | ||
|
|
||||
|
Eric Maskin |
Harvard | Did Not Answer | Bio/Vote History | |
|
|
||||
|
William Nordhaus |
Yale | Bio/Vote History | ||
|
Just answered yes!
|
||||
|
Maurice Obstfeld |
Berkeley | Bio/Vote History | ||
|
|
||||
|
Emmanuel Saez |
Berkeley | Did Not Answer | Bio/Vote History | |
|
|
||||
|
Larry Samuelson |
Yale | Bio/Vote History | ||
|
To reconcile this with the previous answer, AI may well open many opportunities for entrants, but existing Tech giants have been good at preserving concentration by buying such entrants.
|
||||
|
José Scheinkman |
Columbia University | Bio/Vote History | ||
|
|
||||
|
Richard Schmalensee |
MIT | Bio/Vote History | ||
|
Seems probable, since it will enable the creation of new products. But if scale economies end up being very important, it may just entrench incumbents.
|
||||
|
Carl Shapiro |
Berkeley | Bio/Vote History | ||
|
|
||||
|
Robert Shimer |
University of Chicago | Bio/Vote History | ||
|
|
||||
|
James Stock |
Harvard | Did Not Answer | Bio/Vote History | |
|
|
||||
|
Richard Thaler |
Chicago Booth | Did Not Answer | Bio/Vote History | |
|
|
||||
|
Christopher Udry |
Northwestern | Bio/Vote History | ||
|
|
||||
Question C Participant Responses
| Participant | University | Vote | Confidence | Bio/Vote History |
|---|---|---|---|---|
|
Daron Acemoglu |
MIT | Bio/Vote History | ||
|
On balance, this would be my assessment, but there is a lot of uncertainty.
|
||||
|
Joseph Altonji |
Yale | Bio/Vote History | ||
|
|
||||
|
Alan Auerbach |
Berkeley | Bio/Vote History | ||
|
|
||||
|
David Autor |
MIT | Bio/Vote History | ||
|
At least for the foundation models, scale appears extremely important. Many AI services will run on top of these models, but those services will likely be provided by many smaller firms, though within verticals, competition may not be robust
|
||||
|
Abhijit Banerjee |
MIT | Bio/Vote History | ||
|
|
||||
|
Marianne Bertrand |
Chicago | Bio/Vote History | ||
|
|
||||
|
Markus Brunnermeier |
Princeton | Bio/Vote History | ||
|
|
||||
|
Raj Chetty |
Harvard | Did Not Answer | Bio/Vote History | |
|
|
||||
|
Judith Chevalier |
Yale | Bio/Vote History | ||
|
AI is not an industry and many industries use/will use AI in a variety of ways.
|
||||
|
David Cutler |
Harvard | Bio/Vote History | ||
|
|
||||
|
Angus Deaton |
Princeton | Bio/Vote History | ||
|
|
||||
|
Darrell Duffie |
Stanford | Bio/Vote History | ||
|
There ought to be significant scale economies associated with learning from user data and from server farm scale effects.
|
||||
|
Aaron Edlin |
Berkeley | Bio/Vote History | ||
|
|
||||
|
Barry Eichengreen |
Berkeley | Bio/Vote History | ||
|
|
||||
|
Liran Einav |
Stanford | Bio/Vote History | ||
|
|
||||
|
Ray Fair |
Yale | Bio/Vote History | ||
|
|
||||
|
Amy Finkelstein |
MIT | Did Not Answer | Bio/Vote History | |
|
|
||||
|
Pinelopi Goldberg |
Yale | Bio/Vote History | ||
|
|
||||
|
Michael Greenstone |
University of Chicago | Did Not Answer | Bio/Vote History | |
|
|
||||
|
Robert Hall |
Stanford | Did Not Answer | Bio/Vote History | |
|
|
||||
|
Oliver Hart |
Harvard | Bio/Vote History | ||
|
|
||||
|
Bengt Holmström |
MIT | Bio/Vote History | ||
|
AI will be so widespread that the field will be hard to define
|
||||
|
Caroline Hoxby |
Stanford | Bio/Vote History | ||
|
|
||||
|
Hilary Hoynes |
Berkeley | Bio/Vote History | ||
|
|
||||
|
Kenneth Judd |
Stanford | Did Not Answer | Bio/Vote History | |
|
|
||||
|
Steven Kaplan |
Chicago Booth | Bio/Vote History | ||
|
The technology is statistics / algorithms. A key differentiating resource will be information. Different players will have different information advantages.
|
||||
|
Anil Kashyap |
Chicago Booth | Bio/Vote History | ||
|
seems likely, but could they be different players than the currently dominant ones?
|
||||
|
Pete Klenow |
Stanford | Bio/Vote History | ||
|
|
||||
|
Jonathan Levin |
Stanford | Bio/Vote History | ||
|
A plausible scenario is that aspects of the core technology are highly concentrated, but applications of AI are ubiquitous.
|
||||
|
Eric Maskin |
Harvard | Did Not Answer | Bio/Vote History | |
|
|
||||
|
William Nordhaus |
Yale | Bio/Vote History | ||
|
As Zhou may have said about the French Revolution, too early to say.
|
||||
|
Maurice Obstfeld |
Berkeley | Bio/Vote History | ||
|
|
||||
|
Emmanuel Saez |
Berkeley | Did Not Answer | Bio/Vote History | |
|
|
||||
|
Larry Samuelson |
Yale | Bio/Vote History | ||
|
Again much uncertainty, but high-tech tends to exhibit network externalities that lead to concentration; AI may continue this trend.
|
||||
|
José Scheinkman |
Columbia University | Bio/Vote History | ||
|
It should benefit from some of the same forces that led to concentration in search engines.
|
||||
|
Richard Schmalensee |
MIT | Bio/Vote History | ||
|
First, "the AI industry" makes no sense. There will ultimately be multiple markets. Second, it is too early to predict the structure of most, if not all, of these markets.
|
||||
|
Carl Shapiro |
Berkeley | Bio/Vote History | ||
|
|
||||
|
Robert Shimer |
University of Chicago | Bio/Vote History | ||
|
|
||||
|
James Stock |
Harvard | Did Not Answer | Bio/Vote History | |
|
|
||||
|
Richard Thaler |
Chicago Booth | Did Not Answer | Bio/Vote History | |
|
|
||||
|
Christopher Udry |
Northwestern | Bio/Vote History | ||
|
|
||||
