August 2022 BenGoldhaber(.com) Newsletter
Songs to sing and fashion to wear this fall season.
Happy soon to be September - I hope everyone had a good summer, because as soon as this newsletter goes out the US Department of Weights and Measures officially declares it Fall.
Corporate culture used to be far weirder: The IBM songbook (PDF) was an official set of songs praising the company and the executive officers. Here’s the chorus for one of the five songs trumpeting the then president of IBM, Thomas Watson.
T.J. Watson, we all honor you,
You’re so big and so square and so true,
We will follow and serve with you forever,
All the world must know what I.B.M. can do.
I for one would become an instant Mark Zuckerberg fan if he were to mandate Facebook employees sang an ode to him.
In general companies used to be more like mini-countries - outside of overly sincere songs pledging fealty, they were also more willing to use violence. I remember reading about the Ford Motor Company and how in the 1900s Henry Ford empowered the chief of security Henry Bennett to basically run a pseudo investigations/secret police department, which included intimidation, embezzlement, and general thuggery.
I’ve been reflecting on this while thinking more about the Network State, a book I mentioned last month that advocates for a future that in some ways looks more like the past, with a devolution of power from governments to companies/communities that have some amount of recognized independence and rights. This seems appealing in a lot of ways, given how broken the federal government is, but there’s no perfect solution for mini-dictatorships, and people are right to be worried that the incentives of a corporate state will not automatically promote their well being. While a more minor example than Ford, it’s good to remember that the longtime emblem of high minded corporate culture, Google, very clearly conspired for years with Apple to avoid hiring engineers from one another, including firing recruiters who violated that deal. No corporation is necessarily always on the side of human flourishing.
I’m increasingly of the belief that a top ‘political’ priority is to rebuild civil society for the 21st century; that we need communities and institutions that aren’t the government, and aren’t companies, but more like mutual aid societies, with the tools, capabilities, and resources necessary to help prevent people from being crushed by the soulless machine that so often is Big Government or Big Company. For example within modern healthcare, my heart goes out to anyone who is sick and trying to navigate the insurance and hospital bureaucracy on their own. True advocates for the patient, often family members, are a must, and innovating on organizations to support and empower them would be very beneficial.
I think you’re seeing the early versions of these ‘networked coalitions’ in grass-root communities on internet forums and YouTube, disseminating knowledge and tips. However, there’s an impermanence and etherealness to all of these groups that frustrates mobilizing on anything but information warfare. Bridging that gap, along with the tech-savvy gap where if you’re over 25 you lack the digital literacy to even know if the group exists (“oh geez another esoteric discord community”), is a high value project that’s less about developing new tech and more about experimenting with expanding the reach and sophistication of online communities.
Which I guess really means that, as we all suspected, the Internet Forum Moderator is the true hero of our time.
Very slightly but basically not at all related: A great story about naval combat between a patrol vessel and a corporate ship-of-fortune (aka a cruise ship).
[Edit: it turns out this tiktok is fake, and the ‘avatar’ is actually her boyfriend. Thanks to a better informed newsletter reader for pointing it out. The BenGoldhaber(.com) fact checking team apologizes for the miss, and swears a blood vendetta against her and all other tiktok influencers.]
Watch this tiktok - it’s the best full demonstration of ML powered video I’ve seen.
Tiktok failed to load.Enable 3rd party cookies or use another browser
Briefly looking into the pricing of the different tools she used, I’d estimate the total fixed price at about $30/mo to produce something like 10-50 videos. Which is cheap! The fiddliness of these tools will prevent mass adoption in their current state, but there’s obviously going to be an end-to-end app that ties all of them together into one mocap+ml pipeline. I’m not sure how it will be used - from a ten second brainstorm I came up with permanent digital avatars people that will put on, like clothing, for videos, and in general lots more catphishing. But I’m sure the actual use case will be far weirder (“the street finds its own uses for things”)
Speaking of AI video generation, this is a neat example of using AI to explore the state space of fashion options:
I subscribe to the theory that there was a great wave of blandness in fashion and aesthetics in the mid 2000s. I don’t have a good explanation as to why - maybe something like ‘people mostly want to copy each other, globalization+ubiquitous media penetration made it much easier for everyone to know and copy what others were wearing’. But a phase shift in the price and ease of iterating on fashion through AI driven designs might spark a new wave of creativity in the real world.
It already seems to be doing so with online images. Stable Diffusion, an open source text2image model, was released publicly. You can use it on Dream Studio or on your own machine. I recommend looking at a list of all the tools and adaptations that happened within one week (!) of stable diffusion’s launch. It’s really mind blowing, and we’re clearly close to fully automated media synthesis. There are already primitive text2video generation models, and very capable music generation models.
One of the stronger arguments, in my opinion, against AI excitement, is that we don’t see this reflected in stock prices yet. If there’s actually an upcoming revolution in media fueled by AI, I’d expect to see that priced in to tech companies and media companies, with the former rising and the latter falling. AFAICT that hasn’t happened, which I think is minor evidence against some of these developments having immediate impacts across the economy (“the future is already here - it’s just not evenly distributed.”)
An alternative explanation is that BenGoldhaber(.com) newsletter readers are ahead of the curve and get that top tier alpha insight for maximum investment potential. Which is my preferred take.
Gwern on the widespread fraud in Chinese Scientific culture.
It isn't possible to have two cleanly separated systems, one with 'real science' and the other with 'fake science solely for foreign consumption', because if your circumstances are that hostile to real science, you won't get a real science and a fake science, you'll just get 2 fake sciences, in much the same way that Mao during the Great Famine didn't have two sets of books on agriculture, one propaganda and one 'real', he just had two sets of fake books, one propaganda for outside the government and the other propaganda for inside the government from local officials terrified of & sucking up to higher officials and so on to Beijing. Or official ideology: you have your Marxist ideology about economics, sure, but then when you get prompted to head of GOSPLAN or become the general secretary, it's not like there's some secret inner-party Actually-Works-Marxism™ manual you get handed. Just even more dubious statistics and internal documents and scandal.
Scholar’s Stage on the stagnation of scientific and technological progress.
From this perspective the 1950s was less an age of novel technological wonder than the period in which advances made generations before were finally democratized. What the 21st century is to large parts of China, Africa, and India, the post-war decades were to the United States. But by the 1970s, the main gains had all been realized. The fruits of innovation were not ended so much as expended.
#good-content
Ozark: Finished. I’d say it largely holds up - while I have a few minor complaints about the ending, I’d recommend the series to anyone who wants to feel constant stress and anxiety while watching a TV crime drama.
The Sandman: A capable adaptation of one of my favorite graphic novels. So far I’ve appreciated how they’ve taken an extremely surreal story and managed to retell it in an approachable manner in a very different medium.
The Joe Rogan Experience: While driving across the country I finally listened to the Joe Rogan podcast and it is really good. I understand now why he so dominated the podcast landscape - he’s an extremely gifted interviewer who is always moving towards ‘aliveness’ in his conversations. The lack of a script and his willingness to follow his curiosity comes through in a way that makes it both easy to listen to and unpredictable. My favorites have been the episodes where he interviews standups - ex. Anthony Jeselnik - because they tend to naturally take to the format, and the interview gives insight into a community of practice - stand up comedy - I know very little about! Related: The Tim Dillon podcast is quite funny.
Until next time,
Ben
