Christopher Berry
@cjpberry
Followers
2K
Following
9K
Media
3K
Statuses
25K
I turn data into products. Views are mine.
Toronto
Joined May 2008
š¤ Man, this is some very cool stuff from NASA! It visualizes one month of aerosol particles moving throughout Earth's system (from August 2024). Green=Sulfate Red=Carbon Purple=Dust Cyan=Sea Salt [Keen readers will note the pun aboveš]
31
368
2K
Very much looking forward to presenting this slide deck tomorrow at the Europe Workshop of the Friedrich Naumann Foundation! Iāve tried to offer a balanced and accessible overview of the scholarly debates on the causes and consequences of right-wing populism, as well as the
3
17
51
This MSI working paper finds that copyright protections ensuring āfair useā not only compensate non-GenAI content producers but can also mitigate crowd-out by increasing GenAI production costs: https://t.co/ql6lbZt6ig
#ConsumerBehavior #ShopperInsights #AdCreative #Tech #AI
0
2
2
Paul DePodesta is leaving the Browns. I have some thoughts on his tenure and how it applies to running analytics in an organization.
medium.com
A cautionary tale for any executive scaling decisions on untested assumptions.
0
1
1
Potential biosignature on Mars: confirmed. āļø After a year of scientific review, we are more sure than ever that our Mars Perseverance's 'Sapphire Canyon' sample could contain signs of ancient microbial life. Learn more about the discovery: https://t.co/RMO2UFfnjv
366
4K
25K
Of courseā¦even when they do invest well, contact with the parent firm can be crunchy.
0
0
0
While payroll is one factor, a root driver is that typically enterprises simply optimize for outcomes that are orthogonal to disruptive tech ā¦ an enterprise is filled with people trained and rewarded to minimize disruption. Itās why so many operate VC-like organs.
There is no such thing as a āstartup inside a big coā You can pretend all you want. But if youāre not on the hook for payroll come rain or shine, youāre just role playing
1
0
0
Eh, thank you @cafreeland ā I loved that I knew exactly where you stood, and I always felt that you fought for the middle, right in the centre. Thank you for the years and years of service ā itās not easy being a quant anywhere in society ā thank you serving publicly.
0
0
1
Clever.
Our key assumption? LLMs donāt actually exist at the level of discrete predictions, but in the gradient field of possibilities that sits behind each token. Anybody that tells you we understand what is going on inside an LLM during training is deluding you or themselves. They
0
0
1
To paraphrase the Nobel Laureate Robert Solow - we may end up seeing AI everywhere except in the productivity statistics. I enjoyed my conversation with David Weller & Alex Pigman on "Measuring the implications of AI" at the IMF's Statistical Forum. Watch it here:
25
47
252
Are you an AI safety or ethics researcher? The inaugural conference of the International Association for Safe and Ethical Artificial Intelligence (IASEAI ā25) will be held in Paris, France on February 6-7, 2025. Here's the call for participation. https://t.co/b96ImCEYcB
0
6
12
Hahahaha
44
136
2K
Iām seeing less of what I want and more of what I donāt. And Iām baffled.
0
0
0
As someone mainly focused on the comparative political economy of advanced industrialised democracies, a broader perspective is vital. Lucy Martin's book has proven to be an absolute gem in this and other respects. Highly stimulating! https://t.co/MQdAuRFBSP
0
15
58
Generative AI doesnāt seem like a prediction machine, just like a computer doesnāt seem like an arithmetic machine. Still, effective use of both requires some understanding of what is going on under the hood. https://t.co/Wd4lxUcL1i.
hbr.org
To understand the strategic implications of AIās new capabilities, managers need a framework for when AI will be helpful and when it might fail. Under the hood, generative AI tools are still predic...
2
29
81
Chinese startup 01 .ai trains competitive LLM using 95% fewer resources through innovative engineering optimization. 01 .ai trained a GPT-4 competitor using just 2,000 GPUs and $3M, while achieving competitive performance. Through innovative engineering and optimization
59
501
3K
In case it interests anyone: I've re-recorded my @CERN conference on token sampling, entropy and the new physics of LLMs.
8
53
401