Matt's Thoughts In Between - Issue #56

Forwarded this email? Subscribe here. Enjoy this email? Forward it to a friend.

What drives progress in AI?

Rich Sutton has an interesting short essay on the "Bitter Lesson" of the last 40 years of AI progress, which has attracted quite a lot of commentary from prominent AI figures. His basic argument is that to reach new milestones advances in computation have been more important than new and clever ideas (where the US, incidentally, retains its lead):

We have to learn the bitter lesson that building in how we think we think does not work in the long run

Some very smart people are skeptical. Michael Nielsen argues that this ignores all the clever ideas that are needed to make AI methods scalable (see also). Francois Chollet thinks it's true right up until it's... not. Shimon Whiteson notes that it ignores the important role of failure in research.

This disagreement is a microcosm of a much bigger question: are new breakthroughs needed to achieve artificial general intelligence? Or do we just need more of the same (as DeepMind apparently believes)? Of course, if AI is already quite good at poetry, what more do we need?

"What is Amazon?"

I've written quite a lot about Amazon in TiB; it's one of the most interesting and important companies in the world. This week Zack Kanter published perhaps the definitive essay on its success so far and, perhaps, the seeds of its future failure. It's long but worth reading in full.

Kanter sees Amazon as an "unbounded" Walmart: Walmart's obsessive customer obsessiveness, but without the constraints of shelf space and geography. Kanter shows that whenever Amazon hit a bottleneck (onboarding suppliers, deploying code, creating features) it created a platform (Marketplace, AWS, Catalog API). Steve Yegg's classic post on the topic is also a must-read. So far this served Amazon well, but Kanter argues that Amazon has made a major mistake in the most recent iteration of this strategy - Amazon Ads. Previous platforms have been pro-customer, but this is pro-seller. Distorting search results undermines Bezos's claim to run "earth's most customer centric company", which gives competitors an in.

What about antitrust as a threat? Kanter's take, based on his argument on Amazon's modular nature, is fascinating and skeptical, :

Antitrust action to break it apart could do nothing that it has not done to itself already; this is no doubt by design

Do read the whole thing.

Does insect suffering matter?

As long time readers know, I'm interested in effective altruism (EA) - the idea that we should use evidence and reasoning to do the most good. This usually results in a heavy weight on minimising suffering, including of animals, which leads many EAs to reduce or eliminate the amount of meat in their diets.

This is a good tweet length summary of the argument, which led me to this interesting essay on which kinds of animal consumption cause the most/least suffering per kilogram of food. Spoiler alert: milk is best and farmed fish is worst.

Arguably this kind of argument leads to bizarre (if amusing) conclusions, but I was intrigued by this essay by the same author (thanks Ben for the link), which suggests that we don't take insect suffering nearly seriously enough. That might seem absurd (my initial reaction was that if cockroaches can live without a head, in what sense can they suffer?), but there's a strong argument for caution. One of the most interesting and important ideas I encountered last year was moral uncertainty - the notion that, if history is a guide, we probably hold many beliefs that are morally outrageous, but we can't know for sure which ones. And if that's the case - while I'm not going to stop walking on grass for fear of trampling worms - it's worth taking even apparently silly ideas quite seriously.

Quick Links

Undiplomatic behaviour. Did the CIA covertly attack the North Korean embassy in Spain? Extraordinary story.
Nothing new under the sun. Excellent and myth-busting innovation reading list.
Virtual idols. Another extraordinary industry "only in China".
The secret life of cities. Fascinating animation of the growth of the world's largest cities.
The revolution will be deleted. Thread on the sometimes surprising impermanence of the internet.

Your feedback

Thanks for reading. If you like Thoughts in Between, please forward to someone who would enjoy it. And feel free to hit Reply if you have thoughts.

Until next week,

Matt