Director, FutureTech research project at MIT Computer Science & A.I. Lab, and PI at the Initiative on the Digital Economy at the MIT Sloan School of Management.
Cambridge, MA
- Tweets 228
- Following 121
- Followers 1,540
- Likes 173
Excited to announce that I was just promoted to Principal Research Scientist at MIT! (Roughly equivalent to getting tenure)
16
14
566
30,891
I'm hiring 1-2 postdocs to work at MIT on the Economics of Computing and Artificial Intelligence. Know any students who might be interested?
researchgate.net/job/947492_…
7
49
91
Our new article on what happens to computing after Moore's Law is now out in Science!
science.sciencemag.org/conte…
2
26
76
Excited to announce the official launch of the $13M @MITFutureTech research project. We aim to better understand the economic and technical foundations of progress in computing. futuretech.mit.edu. @MITCSAIL @MITSloan @mit_ide
1
6
32
Excited to announce that our paper on the economic impacts of AI-augmented R&D is out in Research Policy. Useful if you want (i) an estimation of the deep learning production function, (ii) a better way of inferring human capital from author lists, or...
1
7
30
4,715
Data on the computational burden of deep learning models, gathered from 1000+ manual paper reviews, now available at our new website computerprogress.com and integrated into @paperswithcode. @KGreenewald, Keeheon Lee, @Manso_GF, @MIT_CSAIL @mit_ide (and thanks to @rbstojnic)
11
33
Our new paper shows that language models double in algorithmic efficiency every 8 months! (Which has big implications for the democratization of AI)
Paper link: arxiv.org/abs/2403.05812
@ansonwhho @tamaybes @EgeErdil2 @everysum @robi_rahman @CarlGuo866 @diatkinson
@Jsevillamol
1
5
30
3,762
Deep learning is propelling AI, but progress requires so much computing that the environmental costs are quickly becoming unsustainable. spectrum.ieee.org/deep-learn…. @KGreenewald, Keeheon Lee, @Manso_GF, @MIT_CSAIL @mit_ide #DeepLearning
1
12
24
Announcing our new paper that goes beyond "AI Exposure" to consider which tasks would be economically attractive to automate. We find that firms would only want to use AI for one-quarter of tasks that could be automated futuretech.mit.edu/news/beyo… @MITFutureTech @MIT_CSAIL @mit_ide
3
8
22
5,223
Digging into the correspondence between patent examiners and filers shows that 10-11% of all @uspto patents are in patent races, with more in tech and fewer in pharma.
(My new paper with Jeff Kuhn @kenanflagler) @mit_ide academic.oup.com/jla/article…
3
19
Excited to report that @CACMmag put my article with Svenja Spanuth (@ETH) about the decline of computers as a general purpose technology on the cover of this month's issue.
"The Decline of Computers as a General Purpose Technology," by Neil C. Thompson @ProfNeilT and Svenja Spanuth, says economics are pushing users toward divergent #computing platforms bit.ly/2NlDfJN. Thomspon states his case in an original video. bit.ly/3aEX4Ex
2
1
20
Curious how fast industry influence is growing in AI research? See my new paper with @NurAhmedB and Muntasir Wahed out today in Science science.org/doi/10.1126/scie…
@MITFutureTech @MIT_CSAIL @mit_ide
3
19
1,277
Thanks @uspto for putting my patent scope paper with Jeff Kuhn as the lead article in their special issue! (takeaway: many traditional measures don't actually measure patent scope, but we validate one that does and show how to use it for causal inference). doi.org/10.1080/13571516.201…
3
20
My lab is hiring postdocs! Interested in the economics of AI and computing, but want to get more technical? Interested in the technical aspects of computing, but want to think bigger picture? We might be just the place for you. @MIT_CSAIL @mit_ide Apply: researchgate.net/job/947492_…
3
7
17
Discussion of our new article about how presenting at conferences shapes citation behavior...
nature.com/articles/d41586-0…
5
15
2,262
Just published in Nature Communications: my paper with @kunjapur and @PPfingstag, on how biologists can increasingly source new genetic functionality from farther away in the tree of life #Genetics #biomimicry @Addgene nature.com/articles/s41467-0…
1
7
16
Want to understand what quantum computing can do without the hype and jargon? See my new paper with @carldukatz @prashshukla @SukwoongC in the California Management Review. @MITFutureTech @mit_ide @MIT_CSAIL @Accenture #QuantumComputing
cmr.berkeley.edu/2024/03/pra…
1
6
15
1,599
Firms in China (domestic AND foreign) use a different mix of innovation sources than firms elsewhere. Curious how? See my new paper with @didiebon, Mark Greeven, and Rebecca Wenjing Lyu out today in MIT Sloan Management Review
sloanreview.mit.edu/article/…
2
15
Moore's Law has yielded tremendous growth in computing power, but how fast are we improving the efficiency with which we use that computing power? @thesherrybrand @MIT_CSAIL @mit_ide
doi.org/10.1109/JPROC.2021.3…
7
14
Today I am #TheLastStraggler for #RunTheWorld! On the real race day I was quarantining preventatively after a mild, turned-out-not-to-be-covid cough. As a result, I literally finished DAYS later than all the other racers 🤷♂️. Thanks to @janagallus for organizing!
1
12
Think that you can predict which projects will be the most transformative for your company? Think again - radical innovation is full of surprises. See my new paper with Rebecca Wenjing Lyu and Gina O'Connor now in MIT Sloan Mgt Review. sloanreview.mit.edu/article/…
4
12
1,289
Wikipedia's influence extends to judges' decisions.
In new work with @BrianFl38315947, @edana_r, @McKenzie_B_A and @XueyunLuo we use a randomized control trial to show that #Wikipedia write-ups influence the content of legal judgements. papers.ssrn.com/sol3/papers.…
4
3
10
It's a Wiki World! New paper with @XueyunLuo, Brian McKenzie, @edana_r, and @BrianFl38315947 shows the full details of how Wikipedia articles influence legal judgments, including the precedents cited and how they are written about. #experiment #wikipedia papers.ssrn.com/sol3/papers.…
1
3
11
1,827
Insightful paper and coining of the term "hardware lottery" by @sarahookr to describe how certain types of computing are stymied if they don't fit with hardware arxiv.org/abs/2009.06489... (1/2)
1
3
11
Curious who created the algorithms that power modern computing? See my new article with @GeShuning @thesherrybrand at onlinelibrary.wiley.com/doi/…
5
11
My new working paper with Svenja Spanuth shows how the economic engine that made computers so general-purpose is failing and may reverse itself, fragmenting computing. This will divide applications into those still in the fast lane and those left behind. ssrn.com/abstract=3287769
3
10
Please to report that my paper with @GeShuning and @thesherrybrand on who discovered the algorithms that underpin modern computing is one of that year's top 10 downloaded papers from Global Strategy Journal.
onlinelibrary.wiley.com/doi/…
1
8
Discouraging (but important) news: some of the techniques for decreasing the computational burden of deep learning can increase bias.
It is not just the data.
Popular compression techniques amplify bias in deep neural networks.
Work co-led with @nunuska, w @_whatcode, Samy Bengio and @cephaloponderer
"Characterizing the bias of compressed models" -- pre-print - arxiv.org/abs/2010.03058
4
10
Last weekend to apply for the postdoc position at MIT on the economics of computing and A.I.
researchgate.net/job/947492_…
4
9
A.I.'s most exciting successes have come from Deep Learning, which has a voracious appetite for computing power. That's quickly becoming unsustainable, as I show in my new paper with @KGreenewald (@MITIBMLab), Keeheon Lee (@yonsei_u) and Gabriel Manso: arxiv.org/abs/2007.05558
6
9
My newly-published paper with Arvids Z. and David M. shows that university technology licensing usually helps the underlying science, but can have big negative effects for research tools authors.elsevier.com/c/1X2UW…
3
9
Curious about the impact of quantum computing, but don't have a PhD in Physics? Our new article "The Quantum Tortoise and the Classical Hare" may be the framework you need: arxiv.org/abs/2310.15505
5
7
957
Curious about how to make deep learning more sustainable? We discuss a promising approach in my new article with @StephanZheng and Silviu-Marian Udrescu at @IEEESpectrum. @MITFutureTech @mit_ide @MIT_CSAIL
spectrum.ieee.org/deep-learn…
4
7
Spoke today at @exascaleproject on the importance of hardware accelerators for #supercomputing performance improvement now that Moore's Law has slowed (based on paper below). Thanks to @ZephyrAardvark for organizing! @MITFutureTech @mit_ide @MITCSAIL
cacm.acm.org/magazines/2021/…
3
7
815
I am once again the #TheLastStraggler for this year's #RunTheWorld! On the original race day I was nursing a pulled muscle in my foot, but fortunately our benevolent queen @janagallus was merciful and didn't object to my slightly delayed circumnavigation of Fresh Pond.
1
1
7
Nice summary of the work being done to try to make deep learning more computationally efficient and environmentally sustainable: news.mit.edu/2020/shrinking-…
1
7
Thanks to @LabsExplorer for getting the word out about the value of scientists adding content to Wikipedia: labsexplorer.com/c/your-rese… #Wikipedia @iamlemec
1
6
How Patent Races Impact Innovation ipwatchdog.com/2020/07/06/pa…
4
Excited about this new work with Talia Bar where we try to combine the best parts of the patenting system with the best parts of innovation prizes.
Now Talia Bar (University of Connecticut) is presenting “After prizes for innovation: A new approach to incentivizing inventors” (joint work with Neil Thompson @ProfNeilT)
#LeuvenSummerEvent #EconomicTheory
5
401
Presenting the Wikipedia paper online tomorrow at 2:30 Eastern at piped.video/watch?v=VR5JwqyV… #wikipedia #openscience
3
5
My new article with @didiebon:
[New Article] Why Innovation’s Future Isn’t (Just) Open —@profneilt, @MIT_CSAIL, @mit_ide et. Al. Free for a limited time. mitsmr.com/2AcZ8DW
2
4
Interesting talk by @deaneckles at #IDEAC2020 introduced me to the fascinating (but slightly disheartening) "friendship paradox" where, on average, your friends have more friends than you do!
en.wikipedia.org/wiki/Friend…
1
5
Matt: As you say, algorithm progress will push down the cost of deployments, making AI automation more attractive. But costs will need to come down in compute, data, and labor to make a big difference. Otherwise whichever cost doesn't come down will become the bottleneck.
Uhhh, FASTER than Moore's law...
Interesting to contrast this with your other recent paper about cost effectiveness of AI in computer vision, Neil. In some ways they're in tension with one another. Or at least imply that the AI torrent will be highly directed?
2
5
841
Thanks to SSRN for choosing our paper on the decline of computers as a general purpose technology as one of their Weekly Top 5 papers
ssrnblog.com/2019/02/18/week…
4
New paper:Science is Shaped by Wikipedia: Evidence from a Randomized Control Trial papers.ssrn.com/sol3/papers.… #openscience #wikipedia @iamlemec
1
4
4
...(iii) an estimate of how scientists will be able to harness more capital, and thus accelerate innovation and productivity growth. @tamaybes @nicholaskemery @MITFutureTech @MITCSAIL @mit_ide
authors.elsevier.com/sd/arti…
4
429
Lots has been said about the AI race between the US and China. But this rivalry goes much deeper...and the US lead is almost gone. Read more in my papers with Dan Armbrust and Chad Evans covering: (1) Advanced computing systems and capabilities... gppreview.com/2023/02/28/ame…
1
1
3
1,794
12 cases studies on how firms are using deep learning (or trying but finding that traditional techniques are better). From new work with my coauthors at @MIT and @IBM.
ibm.com/thought-leadership/i…
3
...This will become even more of a problem in the coming years as progress in general-purpose computing slows and more and more users turn to specialized chips papers.ssrn.com/sol3/papers.… (2/2)
1
3
Visualizing Moore's Law in Action (1971-2019) visualcapitalist.com/visuali…
1
3
New working paper with @didiebon and Sarah Jaballah about how corporations source innovation: big growth in the use of external innovation sources (e.g. universities), but internal sources are more important and more likely to provide competitive advantage ssrn.com/abstract=3567040
3
3
For young Canadians interested in policy, I highly recommend this program (which I enjoyed immensely)
pip-psp.org/apply/
3
359
Who knew that blue, society’s favorite color today, was historically considered unlucky, uncouth, and unfashionable?! itunes.apple.com/us/podcast/…
2
When bad statistics are masqueraded as truth: shocking story of the 1840 US census. podcasts.apple.com/us/podcas…
2
Thanks to Ben Lorica (@bigdata) for our interesting discussion on the future of deep learning.
1/ In this episode of #TheDataExchangePod I speak with @ProfNeilT, Research Scientist @MIT_CSAIL. Neil recently co-wrote a must-read paper entitled “The Computational Limits of Deep Learning” thedataexchange.media/the-co…
3
2
Great talk by @CPehlevan on the mathematics happening under the hood with Deep Learning
On Friday from 1:30-2:30pm the speaker of the @Harvard_IACS Applied Computation Seminar will be Cengiz Pehlevan @CPehlevan of @hseas Applied Math. His topic will be "A Function Approximation Perspective on Neural Representations".
harvard.zoom.us/meeting/regi…
2
Two papers that might be interesting:
1) The effect of more computing power on performance in a bunch of disciplines (protein folding, oil drilling, etc.) arxiv.org/abs/2206.14007
2) Evidence for the effect of Moore's Law improvements on firm productivity papers.ssrn.com/sol3/papers.…
2
80
Deep Learning goes state-of-the-art in 2012, so initially it probably would only have been a few outsourced people who could have harnessed it (not in-house folks). On the other hand, I wouldn't have expected the effects to be strong and broad enough to pick up in your data.
1
That’s to @DaveMcCFO for spreading the word about our new measure of patent scope cfo.com/risk-management/2019…
2
2
Our new paper: Gene synthesis allows biologists to source genes from farther away in the tree of life biorxiv.org/content/early/20… #genesynthesis
1
2
Thanks, Sage. Short answer: lots of areas! The challenge is just that they take a lot of time to run and the subtler the effects are the bigger/longer the trial needed to measure them.
1
@MITFutureTech is hiring! We're economists and computer scientists studying how computing and AI develop and what this means for the economy. Looking for a Deputy Director and a Research Scientist to join us. @mit_ide @MIT_CSAIL careers.peopleclick.com/care… careers.peopleclick.com/care…
1
2
1,368
NYT: 1-7 million humans are exposed to bat corona viruses in Southeast Asia every year! nytimes.com/2020/04/21/magaz…
2
For those looking to donate to a good cause: my cousin helped create a great program at Frederick Community College for student-parents struggling with the cost of tuition and raising children app.mobilecause.com/vf/Unity…
2
It isn't just semiconductors, but all of U.S. advanced computing that requires investment.
news.mit.edu/2023/report-chi…
1
218
Here's the link to the @hgoldstone @LivingLabRadio interview
1
For business school folks: new open-source case &TN that Jen B & I wrote on the solar industry: mitsloan.mit.edu/LearningEdg…
1
1
Fascinating episode on the introduction of Air Conditioning podcasts.apple.com/us/podcas…
1
Interesting quantification of the impact of different testing regimes... fivethirtyeight.com/features…
1
Congrats to the two women who kicked off the #CRISPR revolution. So much has already changed...and we’re still in the early days.
It’s #CRISPR. Emmanuelle Charpentier and Jennifer Doudna ,who pioneered the revolutionary gene-editing technology, are the winners of this year’s #NobelPrize in Chemistry. #NobelPrize2020
go.nature.com/3jHfl68
1
Important day for getting our country back on the right path. Make your voice heard. #Vote2020
1
Really appreciate these thoughtful comments about Elo. We'll be sure to incorporate these into our next round of edits.
1