Reboot's Best & Worst of 2023

ChatGPT, Bryan Johnson, AI song covers, and more

This was a big year for us: Reboot brought on six new editors, focused in on publishing best-in-class tech criticism, and launched the third issue of Kernel Magazine (order by tomorrow and we’ll ship in time for Christmas!). Our most popular essays covered Spotify anxiety, the artificiality of AI alignment, founder mental health, and Anguilla’s domain politics. We watched Oppenheimer and picked fights with Marc Andreessen and listened to too many hours of OSINT podcasts. We navigated the discursive minefields of AI art and autonomous vehicles; we published fiction and poetry and book reviews and more.

But there’s also a ton that our team missed, so we asked the broader Reboot community for mini-essays on the best and worst things they saw in tech this year. We ended up with a pretty eclectic list—insofar as tech criticism helps us “become more discriminating about the things we consume and the effects that they have,” I certainly think these 10 perspectives do the job.

---

Lapse

By Nikhil Sethi

A new challenger to Instagram’s photo-sharing hegemony has entered the scene: Lapse.

Iterating on ideas from Dispo and BeReal, Lapse aims to recreate the experience of a disposable camera. It feels fantastic to take a picture in the app, as the whole screen flashes and an analog click sound plays. Receiving the emoji-laden notification that your photos are ready to be developed is exciting, and holding the “Develop” button as your phone vibrates adds to the anticipation of finally seeing your photos, complete with film-simulation filters applied. 

(source: Lapse)

Despite all of these things, I’m frustrated by this app. Lapse takes every growth marketing tactic in the book and throws it all in. The onboarding flow requires you to invite five friends just to take a single photo, the system sends notification after notification every time anyone posts or joins the app, and exporting your photos without a watermark only works if you invite and share regularly enough.

I want to like this app, since there’s a real sense of craft in its design and execution. It’s responsive, it’s beautiful, but its focus on growth cuts away any of goodwill the software generates. For a company like Lapse, activation and retention numbers are the only way to convince their investors that they should continue to exist, so I can’t blame them for their approach, heavy-handed as it is. 

It’s just a bit sad. In another world, I picture this as a community-supported app with an enthusiastic fan base that keeps it running for years. In this one, it feels like it will eventually go the way of many social apps that make a splash on launch and slowly wither into obscurity.

so-vits-svc

By Krish Dholakiya

Software has commoditized music production. Lyrics are easy, melodies abundant, and the Internet “solved” distribution. The biggest artists—Drake, Taylor Swift, etc.—stamp bundles of these commodities primarily with their voices.

That’s why “so-vits-svc”—the tech behind the flurry of AI song covers this year—feels notable. 

"When I Was Your Man" sung by Kanye West. "Hold On, We’re Going Home" by Frank Ocean. “Bubbly” by Drake. Modern miracles. I’m unsatisfied, though, looking at real technical achievement and not seeing a clear path towards seriousness.

If they sound this good—if voice is the last audible edge—why does this feel unserious?

Maybe it’s because the West primarily values originality, persona, and narrative in music. AI covers are “remixes”, and unbundle voice from the persona. They’re unsanctioned, considered IP theft, and blocked from streaming platforms like Spotify.

India is different in its emphasis. Indian music is typically produced in service of a film. The “playback” singer—whose voice backs a lip-syncing actor—is distinguished not by persona, but by vocal texture and performance. Words come from poets and lyricists; songs are composed by music directors. Indian music also recurrently covers pre-20th-century poetry and folk music. “Originality” is not their north star. The performing artist is not their focal point. The Indian side of my (eg.) Reels feed takes AI covers seriously, and is still making them. Does this technology require a medieval perspective, in which the song both precedes and outlives the singer? Might it resurrect one?

I enjoy the covers. Their future remains unclear. My hunch is that there’s something powerful latent in them, waiting on the right milieu. I’m all ears.

ChatGPT

By Kelin Carolyn Zhang

I always thought I was just too dumb to code. My brain didn’t want to mold itself into the rigidity of computerspeak, so my teachers and classmates made me feel inadequate and slow. I’d burst into tears while trying to even read code, so I just gave up entirely.

A decade later, ChatGPT came along, and ChatGPT didn’t judge. It listens with endless patience, answers my dumbest questions at any hour, and even writes mostly-working code for me. I could take an idea I had no clue how to build—a camera that prints poems? a cursed photo mashup app? a multiverse simulation game?—and just keep asking ChatGPT for help. Impossible projects became real within days. A machine gave me back the confidence that other people had taken away from me. Instead of trying to mold my brain to the computer, the computer could now mold itself to me.

Poetry camera, a project Kelin and her friend Ryan made with the help of ChatGPT (source: Kelin Carolyn Zhang)

In fact, it could mold itself to just about anyone. ChatGPT compresses the breadth of humanity’s written knowledge and serves up the small slice that we need at any given moment. This infinity machine remains inert without human input; we bring our curiosity and our fear, our creativity and our greed. The most compelling part of ChatGPT remains not in what the machine can do, but in what it reveals to us about ourselves.

President Biden’s AI Executive Order

By Charles Yang

President Biden’s recent AI executive order (EO) is a perfect example of the messiness that underlies America’s democratic system. 

Rather than a unified monolith, the executive branch is filled with different factions jockeying for power. You see them in the subtle turns of phrases sprinkled throughout the EO, like the “responsible AI” camp versus those who worry about “AI safety.” Even the verb choices hint at the nature of power exerted over different agencies: who “shall” do something versus who is “encouraged to.” The EO also carries the usual trappings of politics: the need for flashy new initiatives (like the national AI hiring surge and the sleek new ai.gov site), which often overshadow wonky but important policy changes, such as the Office of Personnel Management reforms and the J-1 visa and Schedule A updates. 

But the EO is not interesting for its diction, but because of what it means for AI policy in the world's largest economy. Many think that what kind of AI we end up with will be decided by young tech bros in San Francisco, but the AI policies that come out of Washington DC—a policy space almost entirely driven by a small number of Congressional staffers and advisors in the federal agencies in their mid-20’s—arguably have just as much of an impact. 

“It’s Time to Build” AI policy, and I’ve heard that the US government is hiring.

CAPTCHAs

By Lila Shroff

This year, I spent more time than I’d like to admit failing CAPTCHAs.¹ If you can’t relate to the experience of spending minutes trying to perfectly select the squares with the traffic lights or crosswalks, only to be told to try again, then I envy you. The motorcycles and bikes are the worst. Most annoying of all is knowing that the data collected from my CAPTCHA-solving is being used to train AI systems, which will ultimately make for harder CAPTCHAs. And all of this for what—when GPT-4 can allegedly hire workers to solve CAPTCHAs on its behalf?

Grievances aside, distinguishing between humans and bots online is good. As others have pointed out, better to live in a world of CAPTCHA hell, than one where AI has progressed past CAPTCHAs altogether. Still, I wish that proving my humanity didn’t have to be so damn annoying. 

Sol Reader

By Will Manidis

In the landscape of technological novelty, where gadgets clamor for our constant attention, the Sol Reader is the odd man out. It's a curious device: a $300 headset with a tiny e-ink display promising the “next revolution in reading”. The whole thing is funny at first sight—a pair of opaque sunglasses, seemingly pulled from a sci-fi novel with marketing imagery that better fits Bryan Johnson than your local bookseller. Yet, within this funny-looking exterior lies a quiet revolution.

solreader_

A post shared by @solreader_

Unlike the AI wearables that promise to replace our phone addictions, only to ensnare us in new forms of digital dependency, the Sol Reader offers a different promise. It's not about more connectivity; it's about less. The main feature, seemingly, is being able to lay on your back and read from a Kindle without your arms tiring.

It's a step, perhaps small but significant, towards a future where technology respects our need for space, for quiet, and for time away.

Bryan Johnson

By Andy Weissman

I don't know Bryan Johnson.

I do know that in October he injected himself with multipotent stem cells from bone marrow. I do know he takes hundreds of vitamins and supplements very day. I do know he measures his nightie erections and underwent shockwave therapy for his penis. I do know he regularly posts pictures of his naked, sculpted body. I do know he spends $2M a year tracking hundreds of biomarkers of his body and anything else he can possibly measure. He has been called “The Most Measured Man in Human History.”

All in a quest to understand, is death inevitable? 

He seems, quite frankly, crazy. Unhinged. He receives enormous ridicule as a result. Probably appropriately.

Yet.

Ursula Le Guin once wrote:

The explorer who will not come back or send back his ships to tell his tale is not an explorer, only an adventurer, and his sons are born in exile.

Johnson, thus, is an explorer, out to understand the frontiers of being human, and to tell his tale as he goes. Only this time, this explorer is doing it in real time, in public, with humor, open-sourcing all of it along the way.

I won’t look away.

Tools for thought

By Madeleine O. Teh

I appreciate an over-engineered note-taking app as much as the next person, but isn’t the availability of options getting insane? 

We have Notion, Coda, Obsidian, Bear, Milanote, Mem, Evernote, and the list goes on. I’ve sampled almost all of these, but I didn’t accomplish more tasks in one day. 

source

I spent more time pruning databases and making everything aesthetic than getting anything done. Instead of removing blockers to my productivity, these apps and their smorgasbord of customization options gave me another reason to procrastinate. 

At the same time, two tools went above and beyond this year: my moldy ten-year-old notebook and scratchy Kaweco fountain pen. These little guys sat on my bedside table and helped offload my anxious thoughts through doodles, mind maps, and meandering diary entries. 

Committing ink to paper slowed down my thoughts, which made my thinking clearer and more focused. I had to think hard about each word if I was spilling (relatively) fancy Kaweco ink for it. This slowness forced me to engage in metacognition and confront the gaps in my thinking.

Despite my rave review of pen and paper, I can’t ignore its limitations. I couldn’t type in a few keywords in a search bar to pull up a note. I also couldn’t store images and videos along with my writing. 

So, for 2024, I’d like to see software that incorporates friction as a feature. In a culture hurtling towards a nebulous notion of progress, perhaps taking a step back could help us recalibrate. As the productivity masters say, slow is smooth, and smooth is fast. 

Effective altruism

By Scott Moore

In 2011, a young man named Smith lost his key in a dark alley. After hours of searching, his friend Adam came by and, noticing his distress, went over to help.

“What’s going on Smith, did you lose something here?” Adam asked.

“Not here, I’m afraid, just down that alley there.” Smith replied.

Puzzled, Adam pushed further. “If what you’re looking for is down that alley, then why are you crawling around in the middle of the street?”

Smith stopped and looked up in exasperation. “Because this is where the light is!” 

We’ve now come to learn that Smith suffered from a rare condition called effective altruism. Those afflicted have been known to become obsessed with staying in areas with high visibility, and find themselves plagued by visions of imaginary trolleys. Unfortunately, stories like Smith’s have only become more common in recent years, affecting families across many cosmopolitan coastal cities.

Although the symptoms of effective altruism can make it difficult to manage, 2023 was the year researchers found a possible intervention: spending less time with meta-modern abstractions online and more time with each other. Indeed, as it turns out, the amount of time patients spend at dinner parties and in real communities of care directly correlates to positive outcomes.

In all seriousness, while the effective altruism movement contains many good people, the past year was one of reckoning for its general theory of change. We should all embrace effectiveness, careful measurement, and long term thinking, but when paired with a culture that prioritizes control, these ideals can easily be used to circumvent calls for responsibility. Over the holidays, I’d encourage us all to step outside of Plato’s cave, beyond the walls and screens, and take the time to help one another.

Nikhil Sethi is a writer, poet, and designer based in San Francisco.

Krish Dholakiya is a terminally-online software engineer.

Kelin Carolyn Zhang is an independent designer based in Brooklyn.

Charles Yang left an AI startup to work at the Department of Energy and now tells all his friends why they should work in the government too!

Lila Shroff is a member of the Reboot Editorial Board and Stanford senior studying Symbolic Systems and English.

Will Manidis reads a lot.

Andy Weissman is a native New Yorker who also is a venture investor at Union Square Ventures.

Madeleine O. Teh is a Manila-based designer and writer who you can find drinking matcha.

Scott Moore is founder at Public Works, an open-source ecosystem fund, and a researcher at Metagov. Previously, he was a co-founder of Gitcoin, a platform for democratic community funding.

---

Reboot publishes free essays on tech, humanity, and power every week. If you want to keep up with the community, subscribe below ⚡️

---

💝 closing note

At the beginning of 2023, I made a promise to myself not to publish anything on Reboot that I wouldn’t feel proud to shout from the rooftops. I think we kept that promise, and am proud and grateful for all the incredible editorial work we’ve gotten to feature this year.

Thanks for being a reader/supporter/friend,

Jasmine & Reboot team

P.S. Creative submissions (fiction, poetry, art) are open for Kernel Issue 4 for a few more days. Send them here!

1. I learned while writing this that CAPTCHA stands for “Completely Automated Public Turing tests to tell Computers and Humans Apart.”