eloisius

These are absolutely consuming Taipei over the last few months. This spring I saw an AI-generated menu for the first time, and then it was like overnight that now on basically every street there's at least one restaurant that has an AI menu, sign, or window dressing. It looks awful and I can't understand how anyone thinks it's a good idea. Noodles give me skin-crawling ants feeling. Chopsticks pick up tofu without pinching it, as if it's drawn magnetically. Spoons dip sauce from bowls that are smaller than the spoon.

I use these a lot. If I find I keep front-loading a lot of context I just create an assistant. Like I just created one specific for Qt that includes all the context I kept prefacing questions with. The only thing I wish was a little easier was to change the model of the assistant.

If you’re a Kagi user you can view the most-blocked list and adopt it

For the same reason that you can make a similar-looking chair, but you can’t distribute a fuzzy copy of Star Wars. The char isn’t a copyrighted work.

Apartment hunting is an unpleasant chore I haven’t had to do since 2022. It hadn’t even occurred to me that AI slop would be the norm. I really have lived to see man-made horrors beyond my comprehension.

I’m just collecting training data for my AI.

There’s no way this would work, or work more reliably, than the translator apps we have on phones. In order for this Babelfish sci-fi interface to work, the other party has to be aware and pause their speech for the translation. If they can’t hear their translated message they’ll speak over it. Either that or you have some kind of deafening that passes-thru the translated voice in realtime, and then you lose their emotion, emphasis, and tons of other information that doesn’t fit neatly into the standard monolingual worldview of tech.

Even the “conversation mode” built into Google translate or the iPhone app is useless. I can only imagine it working in the sterile environment it was probably designed in: a conference room with two people trying there hardest to make it work.

I live abroad and travel a lot for photography. Whenever I’m using a translator app, it’s typically a chaotic situation like haggling with a taxi driver, a meal with a group of strangers who invited me to eat with them, lots of background noise. The mode that everyone defaults to, without fail, is to use their own phone to speak or type a message and then hold it up in front of their interlocutor’s face. Sometimes they mix in some fragments of English or I know some fragments of their language. It’s lossy but it works.

I can’t imagine a wearable that would perform better. A notepad that can magically translate little messages is about as far as I would want it to go. Tech is pretty awful at intermediating human relations.

That only worked in first-gen dating sites where you browsed profiles. I think this strategy was called "spearfishing" where you make your profile as polarizing and authentic as possible, because it would cut out all the people you don't want to meet and appeal strongly to the person you do want to meet.

In today's Elo-score face swiping apps, your profile just tanks because all those people you didn't appeal to left-swiped you, and the one person who would have liked your tattoos never sees you. The only strategy for these apps is to 1) be attractive 2) don't be unattractive.

I think his usage of spectacle here is the "spectate" kind, not the hullabaloo kind. Spectacles are created to be consumed from the sidelines as a spectator, rather than directly participated in. Being involved, not a spectator, is the root of punk ethos. Bad music and outlandish hair was always just the window dressing.

Say what you like about geohot's public behavior and his long string of half-finished projects, but he's always been very DIY and I respect that about him.

$50/year and Fastmail will let you alias anything@yourdomain.com to your inbox. I use a different email for every company or website I interact with, so I know who spams me.

Man. If I had this kind of mandate I could really burn some tokens. Review each new PR and extract 100 topics to debate related to it. Spin up 1000 sub agents, each with a different personality profile system prompt, to debate each point until consensus has been reached. Synthesize the learnings into a limerick. Build a Spotify playlist that pairs with the tone of the debates. Post the limerick and link to playlist on the PR and tag me to notify me that I have a PR to review.

Agree with your assessment of guardrails. They barely work on the best days. We need to flip the idea of “agent” on its head. The agent here is an agent of the user interfacing with GitHub. Not an agent of GitHub interfacing with the user. Prompts and guardrails cannot keep the agent loyal to the company. Stop giving these things any permissions the user doesn’t have, and recognize them for what they are: a different UI than web forms, but still the same security model.

Location: Taipei, Taiwan

Remote: Yes

Willing to relocate: Maybe, but definitely willing to travel

Technologies: Python, Torch/PyTorch, COLMAP, OpenCV, C++, Qt

Résumé/CV: https://zacstewart.com/resume.pdf

Email: work at zacstewart dot com

LinkedIn: https://www.linkedin.com/in/zac-stewart-79b2a5a6/

Looking for computer vision work. Multi-view stereo, camera calibration, SfM, scene reconstruction, 3D object triangulation, 3D multi-object tracking. Object detection, tracking, video processing. I'm also very advanced with Qt and would be happy to be a frontend engineer on Qt projects. I mostly work in Python and C++.

I make no claim to be smart and calling a high school diploma an education is a stretch. That said, human sex overwhelmingly clusters around two categories, albeit with some intersex outliers. Additionally, Bimodal sex has endured for as long as recorded history, so it appears to be an immutable trait insofar as the species is a stable species and not taking into account that somewhere along the timeline we share a lineage with single-celled organisms.

Race on the other hand, does not neatly cluster into discreet categories that any responsible person would define, draw boundaries around and count. Furthermore, race has not remained stable over recorded history. Populations mix, join and split. Which race were the Old Kingdom Egyptians? Our modern notions of Asian, Caucasian, etc. probably did not even exist during their era.

How many races are there? How do you draw boundaries between people who exist on a spectrum of genetic variability? What race are Kazakhs?

It's true. Magazines sales are dead. Product photography is probably toast (I'm seeing companies test what they can get away with using AI generated images of their food/products right now). Sports? I don't even follow it but I suspect it's all video content now. Funny enough wedding photography seems to be doing fine as far as a commercial segment goes.

But there are segments of the market that were always small, and AI doesn't threaten. Namely, fine-art and long-term documentary. The kind of work that gets shown as exhibition and sold as prints or photobooks. I can't imagine this stuff be supplanted by generative AI, because it has never been about economic efficiency. It's a labor of love for a photographer to work on a project for years at a time because the subject is important to them. And the customers of these works are not likely to accept lower-cost substitutes produced by a AI. The whole thing is too much about taste and thoroughly infused with humanist ethics. When I buy a photobook, I want to know that the artist lived and produced what I'm seeing, not just that it's pretty ink dots on paper.

From a market analysis perspective it's foolish to cater to these tiny and commercially nonviable artists. However, they have enormous esteem and influence within the community, and make part of their living as educators. I believe making tools that they need, and without blood-sucking cloud subscriptions, you are indirectly marketing to all the creatives (usually tending toward commercial) who attend workshops and see what tools they are running the workshop with.

Of course I don't know either way. However, I feel like it's going to be different, but maybe not apocalyptic. The market I am building for is photographers, and from what I can gather (and know first hand as a customer) is that there's real discontent with the toolmakers in our ecosystem.

It seems like in the quest for every tech company to become a multi-billion dollar empire, they've lost the plot on making hammers for their customers and have instead turned into some kind of strip mining operation. A totally AI agent-driven company is a MBA wet dream, and I think a fever dream. If Adobe, for example, were to achieve it, I don't think they'd use it to fix the backlog of bugs overnight. I believe they'd just become an even more incoherent zombie, trying to extract rents from creative cloud subscriptions.

In the meantime, photographers still need tools. If you wanna run a software company as a regular small-to-medium-sized business, you may find some customers that are happy to buy a quality hammer. The unicorn startup days might be behind us, but I'd be okay with that.

Now, if AI obviates creatives altogether I don't really know what to say. I'd morn the loss of a world that became so tasteless that AI-generated decorations are good enough, for starters.

If you have runway, it’s a good time to start your own thing or join someone who is. Personally, I cannot force myself to wade through slop PRs from careless coworkers. If that’s the job now, I’d rather run a hotdog stand or something.

Luckily, I don’t think things are that dire. I think the companies issuing AI mandates are manufacturing sawdust, and even if it works, it would just enable them to burn through customer goodwill in record time as they make user-hostile decisions free from engineer pushback.

These are going to be a few tough years, but I think the opportunities to start something new are everywhere.

I know from reflecting on my own beliefs now compared with 10-15 years past that one's beliefs can change, and I don't want to be so cynical as to say that these fanboys don't actually believe what they are saying and only want to score points. I'm sure there's a great deal of commentary that is astroturf, but I think there are plenty of (hopefully young and naive) techno-optimists who sincerely think companies like Anthropic can do no wrong and only move humanity forward, or something like that.

In any case, online debate is not always about changing the mind of the single person you engaged with. To some degree, its performative debate so that other readers may be influenced by your ideas.

I think Anthropic will argue whatever argument is likely to protect their interests. I don’t expect anything consistent or moral from them. My quibble is with all the Anthropic fanboys who repeat this crap.

Just like these distillers can learn from Claude’s output. Fair use is fair use.

The hypocrisy of the pro-AI mega corp arguments makes my head spin. For three years they’ve been using the example of a human reading books and then outputting creative works influenced by them as analogy for training AI on copyrighted works. Now suddenly we’re supposed to not draw the same parallel about a hypothetical person who learned from Claude and is now outputting creative work based on it.

They didn’t measure sawdust accumulation. They measured the electricity bill.

I have also gotten spammed by a YC startup, but they spammed an email that I use in git commits, and lead with "I saw your fork of $POPULAR_PROJECT, pretty cool!" or something like that and then continued to pester me with their drip program even as I replied asking them to never email me again.

Sounds like you need to get a couple goats

I won’t go too far into the weeds, because I’m not an internet lawyer, and I basically agree with you, but I do believe there are access restriction laws that are not only limited to copyright violation. People have gone to prison for enumerating sequential identifiers in URLs to access records they shouldn’t be able to. I don’t know if Anthropic could actually make a case there, but it seems plausible at least.

This is a conundrum to me. I was a pretty bookwormish youngster and read a lot of classics. Often I had to push myself through works like Crime and Punishment but I felt like it was good for me. I’m glad I exercised the muscle of reading, but now I can understand that those books just don’t hit like they should when you don’t have the life experience to understand them. Something like Ulysses is still difficult, but at mid life you can really get it.

Would I rather have waited until 35? No, but I’ll probably go back and reread a lot of those books I read when I was younger.

Indeed! It’s so hard to find reasonable takes on AI that aren’t littered with accounts created 11 days ago that only post in threads related to Anthropic for some reason

That sounds like it would actually be fraud.

With you, but I suppose they could have a case for circumventing access restrictions under the DMCA aka leet hacking.