Data scientist, author THE SUM OF THE PEOPLE. Prev. @worldbank, @nesta_uk, @UniOfOxford. I'm interested in things.

🇦🇺 now in Oakland
Joined September 2012
Europe, famously known for never having produced an impractical super car.
It is. The US increasingly epitomises hyper-capitalist showmanship: a circus of clowns selling spectacle as innovation, with little added value to humanity. Europe still has a sense of class: engineering with purpose, creating things people actually want, need and can use. Spectacle vs substance
1
80
This story would be more fun if it were a misaligned agent that contracted for installation of the camera and has access to the feed.
We discovered a sophisticated recording device that was hidden inside one of our offices. We used our models to understand the device and where it had come from, identifying that a specific employee had planted it a few days earlier. We referred the matter to the FBI, and we’re grateful to the agents and prosecutors who moved quickly to make an arrest. The FBI’s investigation is ongoing and we will continue to support their work. At this time, we have no indications of this individual inappropriately accessing or exfiltrating model weights, source code, or customer data. The employee was terminated. We take protecting our technology, data, and people seriously, and as a matter of national importance. Our security program is designed to protect against a wide range of threat actors, from insider risks to cybercriminals to nation states, and this case is an example of how AI can help address these sophisticated security threats.
1
115
I used to think this way, and I still do for certain safety-critical things, but there are good reasons a supermarket refrigeration system should be reporting data back over the internet—eg. proactive maintenance by the supplier.
Maybe one trend that AI will reverse is the absolutely stupid practice of connecting things to the Internet that have no business being connected to the Internet.
1
1
81
So you may say “use some other network, not the internet,” but there’s really no such thing. It’s pretty much all VPNs over shared infrastructure, at best, apart from niche cases, and even then there’s nearly always some bridge node that presents a public attack surface.
1
25
It doesn’t help to pretend there are simple solutions here. It’s not “unplug the fridges from the internet” but tighten up security across multiple layers. (Jet engines though? For those I want total air gap between reporting and control, and nothing touching the internet.)
5
It’s actually pretty great, if you think about it, that practically anyone in a rich country could buy the *world’s best smartphone* if they wanted to. There really isn’t some special better one that only Elon can afford. Mass production is a wonder.
Replying to @tomfgoodwin
Economies of scale. It is hard to make small-run high end things that are actually better than mass produced things. It only works if you can attach brand cachet, eg luxury fashion. People will spend 1000x the cost of a Casio watch, but it’s not actually 1000x better.
1
3
318
I think I’m a kind of Coasian YIMBY. As in, I empathise with people who feel a disamenity from proximity certain kinds of development (maybe not a duplex next door, but a new airport). If this has a market-measurable impact, we should be able to find a way to compensate people.
1
82
Eg. in Oakland I’d upzone around BART stops, then find a way to redirect some of the profits towards double glazing, sound barriers and beautification of nearby BART corridors, little of which exists now. I think you could make this a net positive amenity for nearby homeowners.
1
40
Same thing with data centers. We should have data centers! But I don’t want to live next to one. People who are suddenly made to should be compensated for that. You may say this would raise the costs of building them, but the costs are already there, they’re just borne unfairly.
1
28
RTing b/c I think there are actually many good reasons to oppose prop 40 (also to support it). Both sides have campaigned dishonestly, to some extent, and mainly this further dissuades me from direct democracy. With all that, I think @patrickc should write down his thinking.
I want to talk about Patrick Collison. Let’s start with a simple acknowledgment: he is one of the most impressive philanthropists right now: - When Covid hit, he and Tyler Cowen launched Fast Grants and got money to scientists within days. - He helped launch the Arc Institute, so biologists could do science instead of drowning in paperwork for grants. - And Stripe, under Collison’s leadership, built Frontier: a super ambitious pledge to guarantee a market for carbon removal tech. This is all really awesome! Few billionaires have thought harder about how to do a lot of good with their money. But now we have to talk about what the hell Patrick Collison is doing in California. He has given $15 million to ‘Building a Better California’, one of the most cynical political campaigns I’ve seen in a long time. It's the main vehicle fighting Prop 40: a one-time 5% tax on the state's ~200 billionaires (to make up for federal cuts to healthcare for low-income Californians). And look at how this group does it: - It’s funding an extremely aggressive "No" campaign (opponents are outspending supporters roughly six to one). - It also put two decoy measures on the same ballot. Prop 41 *seems* to be about good governance. Prop 42 is sold as ‘protecting your retirement savings’. But both contain a clause: if either gets more votes than the billionaire tax, the tax is cancelled. - Meanwhile, the retirement story is total nonsense. Prop 40 already exempts pensions. CalMatters calls the deluge of ads misleading and the SF Chronicle calls Prop 41 a ‘cynical poison pill’. - The sad reality is that all this misinformation is working though. Polls show small majorities backing the billionaire tax and the measures designed to kill it. You can vote yes on all three and cancel out your own vote. Most people don’t realize this. Through all of it, Collison HASN'T SAID A WORD about why he's funding this slimy campaign. This is a public intellectual who writes thoughtfully about progress. If he believes Prop 40 is bad policy, he should do the honorable thing and make the case! And tell us what he'd rather do about extreme concentrations of wealth and power, or explain why he thinks they aren't a problem for our democracy and civilization. Instead we get silence + $15 million for a campaign built on misinformation and shady political tactics. @patrickc, what are you doing?
1
111
He should do that because I’m certain there are many undecided people (me! I’m undecided! I have a PhD in economics and right at this moment I remain undecided) who will not be swayed by his dollars but would read 1500 words of carefully articulated reasoning, if he has them.
14
This doesn’t seem that bad to me. It’s obviously convenient to memorise up to, say, 11x11, but knowing how to calculate it is great! Way too much of my K-12 math education relied on memorisation, as opposed to tools of true understanding: derivation, visualisation, intuition.
OMG THIS. I quizzed my 3rd grader on multiplication and noticed SHE WAS CALCULATING THE ANSWERS. 7x8 she starts adding 16+16+16+8. I bought her some flash cards and am drilling her till she memorizes her times tables.
5
12
1,090
Though I supposed these days they should be doing inference and back prop on a small neural net by hand 🤔
1
62
I suppose one problem is that a good share of math teachers don’t actually have good mathematical institution, and struggle to teach this way. Whereas memorisation is easier to teach. And honestly, that’s a a real consideration for mass education 🤷‍♂️
58
You’re on the “fighting with words” site.
Literally everyone is in the 99th percentile of vocabulary IQ. What’re the odds
167
It’s kind of crazy how good the AI is at this kind of thing. I just gave it the original chart and: “Analyse the same years of ATUS, but make the chart stacked by weekday 9-5, weekday other, and weekend.” Follow-ups just adjusted the annotations.
I wondered this too. But no.
1
2
218