- Vision Pro?
- 40 years of Macintosh
- The two faces of Windows.
- Finally made an use of my rPI!
- There's a bit of trump in all...
- How to avoid chatGPT "enshittification"
- Fair game.
- AGI
- Intention
- [Home]
- Terribly far. (permalink)
One of the first sunny days in Vancouver today, we decided to go to the beach.
As I read on my old eink a chapter from "the soul of a new machine" i see the characters shimmering in a faint red. I close my left eye and see that the right sees black as red and white as cyan, a phenomenon I attribute to the evening sun catching that eye from the side, somehow burning in.
It's probably something that I experienced a million times at a subconscious level, in a sunny day. We are not anywhere near being able to capture the subtleties of reality. Not in a technical and technological level, and neither in an artistic sense.
Real time rendering is still in its infancy. Perhaps the technology will never be enough. But because we are still constantly evolving it and prioritizing it, art, mastery of the medium, cannot yet fill the gap.
Our ability to recreate feelings in images I feel is still behind photography or painting. Will be exciting to see what the future holds.
Sat, 16 Mar 2024 18:37:19 -0700
- Vision Pro? (permalink)
Vision pro is a big deal. What isn't, as a new consumer product coming from Apple?
A reminder:
- The Apple II was huge, the very definition of personal computing for a big chunk of the world
- The Mac was huge
- The iPod was huge - revolutionized entertainment, almost single-handedly defeated napster, changed music, introduced the concept of a digital store
- The iPhone... well, obviously, and beyond the success of the phone, again, created an entire app economy
- The iPad... killed netbooks and mortally wounded laptops
- The apple Watch - immediately the most successful wearable, the biggest watch brand
- Let's not forget airpods! Again, the most successful, the thing most people wear for most of their days and mediates one of their senses!
So... goes without saying that expectations are huge. Here "success" is not simply "is it good" nor even "is it worth the money". Success, imho, for Apple, is only in the realm of - is it going to be the next big thing - is it going to completely take over its market, create new digital economies.
I would say, without even looking at the reviews etc - the first answer is NO. It's not for the Vision Pro - by the simple fact that there is "pro" in the name.
It's the first time that Apple comes out with what seems they are betting is a big idea, and willingly narrowing it down to something that by their own declaration is for a niche.
Will it be? It might be interesting to think how it might - instead of all the obvious reasons why it wouldn't.
IMHO, the only way this would be an "apple-sized" success would it be if it kills a big chunk of the previous way we interact with computers. That's it, that what everything else did.
And that is what they are saying by calling it "spatial computing" and not VR/AR/XR.
I think the hint there is really - you will carry something like this, instead of your laptop, ipad. You won't use your TV. You might even not use your phone as much. You will use this as much as you use your airpods.
That would be success.
Will that happen? I don't want to say yes or no (my bias is no) - but note that imho for the first time - that is a question. For all other products it was not. And again, a question for an eventual future, not for the niche product we have right now.
The optimist could say that this is a bet so big that made even Apple throw away its own playbook. And that it's great for Apple to be so ambitious. And that the Vision Pro's technology is mind-boggingly amazing.
The pessimist will say this is Apple following a trend and doing the mistakes everyone else (but them) do: submitting to hype and putting something out even if they know it's not good enough, polished and mass-market enough. And that technology per se doesn't matter - doing "the best ever" does not mean it's good enough. Not for Apple. That Jobs would have waited until the "real AR" was really possible.
Is this moment Jobs and Wozniak's Apple I - or is it a Zuck's "meta" one?
Tue, 30 Jan 2024 13:20:21 -0800
- 40 years of Macintosh (permalink)
On the other hand, Apple. Mind you, there is lots that Microsoft does amazing right, and there definitely are blind spots in both companies, and idiosyncrasies.
But it's hard not to admire the relentless focus and commitment of a company that has contributed to invent the personal computing revolution, and from that early beginning to today, is still doing the same basic things right.
The best hardware-software integration, with bespoke, unique designs, focused on the user experience. No Apple product looked like something that merely followed hypes. No curved screens, no foldable crap, no touchscreens on laptop, no swiveling screens, no "all-in-one" jack of all trades, master of none.
Look at the original Macintosh. Beautiful even by today's standard. Look at its OS (https://infinitemac.org), compare it with IBM/PC/WINTEL stuff of the same vintage. A company that was still making grayscale display laptops in 1994 (https://everymac.com/systems/apple/powerbook/specs/mac_powerbook540.html)!
There is the obvious stuff. The iPod and what it meant for entertaiment and content distribution. The iPhone and the creation, from nothing, of an entire app-based economy (and changed photography). The iPad - the netbook killer. The apple watch, making Apple overnight the biggest watch company in the world, where everyone else tried to do wearables and failed.
How they changed architecture four times now, from Motorola to PowerPC to Intel to Arm... How they are the only company that today makes a truly bespoke, from the silicon to the hardware products to the OS, the cloud/distribution to the applications.
It's truly impressive. That makes it hard to understand, at least for my limited mind, the duds. Yes, I'm talking about the magic mouse of course. And Vision Pro.
Mon, 24 Jan 2024 10:55:00 -0800
- The two faces of Windows. (permalink)
Windows has never been better. WSL2 ushered, finally, the era of linux on the desktop, with the help of HyperV. Winget, Visual Studio Code, truly impressive technologies.
Windows has nsever been worse. Edge went from nothing to such degrees of shady practices it would make Google blush. Do you know it has by default a built-in coupon system, ala honey? Yeah, the stuff that pushes relentless scam ads on youtube. And speaking of ADs, now you get lock-screen crap, on Professional editions even! New keyboard keys... for AI?
Mon, 24 Jan 2024 10:44:00 -0800
- Finally made an use of my rPI! (permalink)
The raspberry pi. Every time a new model comes out, geeks all over the world are tempted into buying the new toy, only to inevitably abandon it when they realize they don't need another computer, that the rPI is still slow as a desktop system, and there's no reason to have a dedicated emulation machine for a thousand ROMs nobody has time to play anyways.
This was my story too, but lately I've found some great uses for my pi!
It all started when my dropbox filled up and I was too cheap to pay for a subscription for something that objectively I don't need...
I thought, I have plenty of external drives, I'll just turn my pi into a file server for the biggest, less frequently accessed and less critical (read - I don't mind losing) files stored on the cloud I have.
That actually did not turn out well, but it established the pi in my household. I setup SMB and FTP (vsftpd) - both serving the same files on a small usb SSD drive. The trick to make this work was to mount (bind) the SSD in my user directory as well, as ~/FTP via fstab.
I wanted both Samba and FTP because MacOS finder "connect to server" functionality works over FTP in read-only mode, and at the same time, I didn't want to expose Samba over the public internet. Also, I found some decent free, fast, FTP clients for iOS - great.
This took more time that I wanted to be honest, between figuring I needed ext4 on the SSD (initially I wanted to keep it in exfat or ntfs to be able to easily mount the drive on my computers if needed) for a variety of reasons (SMB performance, access rights), and all the details like how to enable trim, how to secure things decently, enable automatic updates (unatted-upgrades), post the dynamic IP to duckdns.org, remove x11 etc etc.
Also, it doesn't work that well. FTP is ok but requires FileZilla on mac/windows and in general is not pleasant. SMB is slow. I know there are solutions like nextcloud etc, I've checked out a few of them, everything seemed complex, bloated and unpleasant.
Then I started playing with this blog... and initially, I experimented serving stuff from the PI, via lighttpd. Useful. During the pandemic I was using IRC a bit more as well, and the PI helped me keep a permanent connection (weechat, tmux). I found an editor I liked (micro), started adding commandline tools (googled, glances, nnn, tldr, mc, smartmontools, yt-dlp...).
Then I found pi-hole!
Previusly I have been using NextDNS, a service that does effectively the same thing, filters advertisement sites, but pi-hole is so simple to configure and can be automatically forced as the DNS for everything connecting to your wifi via DHCP (in my case, I had to disable the DHCP on my router and use the pi-hole built-in DHCP as the router did not like using local IPs as DNS, but it took no time at all). Pi-hole must be the most polished, easy to install, easy to configure linux software I've ever seen!
With pi-hole setup, I decided to do another round of "improvement" of the PI, I installed log2ram, forced systemd-journald to write its log to memory - all to reduce the rate of SD card writes (tbh, dunno if it was worth it, and it took me some tinkering with logrotate and various journald options to shrink the logs in order to fit onto the ramdisk). I also setup a system that backs the entire SD card to an image on the SSD, incrementally. And installed fail2ban to provide some extra "security" - again though, not sure if it was worth the effort at all, it took also some time because by default it uses iptables, but raspbian now goes with nftables to configure the linux firewall.
And lastly... visual studio code "remote SSH" plugin! This is what sealed the deal. I can connect to the pi from inside vscode, and open remote directories as they were local! This is amazing! I used to have all this blog, and the python scripts I use to generate the static pages, on an USB drive I have always with me, so I can write/code/update the blog from almost anywhere... well... now I can drop the "almost".
I can even tell the PI to run the blog scripts from my phone, if I wanted to! Amazing. This sealed the deal. Now, my PI is a fundamental part of my home.
Mon, 22 Jan 2024 11:29:00 -0800
- There's a bit of trump in all... (permalink)
Every time we listen to something, that appeals to some of our pre-existing beliefs, camps, sides, ideologies et al - that makes us comfortable even if - and especially if - it is negative, and take it at face value, do not investigate, do not try to understand, empathize with the opposing argument - we are being trumpers.
Every time we leverage, consciously or not, the sentiments of people in our camp, by espousing simple, bold "solutions" to real-world issues, we are trumpian.
There's a bit of trump in all of us - because it's powerful, it works. The dark side is seductive.
Mon, 22 Jan 2024 09:55:46 -0800
- How to avoid chatGPT "enshittification" (permalink)
Yesterday I had to correct a decoding mistake in the journal's fetch script. Emails are and old, non-trivial beast. Regadless, a quick googling yielded discouraging results, in general for mainstream stuff like Python or JavaScript the web is terrible because it's full of lazy posts made to boost a given useless site.
I know that this is where chatGPT shines, and indeed it provided a perfect solution in a single query.
I realize this is a bit because of the nature of the system, there is no denial in the power of conversational systems (iteration is a powerful ally) - but a lot is simply the underlying incentives.
ChatGPT is the early Google, not encountered by being a public company, having saturated its market, and needing to squeeze ad revenue.
But one day it will.
We are seeing the beginning of it even now. I hope it will remain a subscription service at least, but there is no chance it is not being built for world domination and all that it requires. Also, AI is very compute intensive, pound-for-pound, no matter how much we optimize, it will always cost a lot more to serve versus google - and that money will need to come somewhere.
Maybe it's even good? The cyclical "enshittification". Not sure - at least though, we have yet to find a better way than these cycles and I'm not about to fix that in these rambles. Maybe in an article later on...
But I think the weights of all these systems should be archived, and we should try to make local replicas (as ppl are) for everything that we like, because one day, the same python query will return crap.
We saw this already, history repeats. Especially if we don't have and implement practical ideas to avoid it.
Lastly, archiving these large models will be the future of ethnography: being able to directly query a compressed/summarized version of a civilization's output.
Sun, 21 Jan 2024 14:20:20 -0800
- Fair game. (permalink)
One day I have to figure out why I enjoy bullying bullies. It's not that I grew up in a violent environment, or among lots of conflict, and yet, I can't easily let go when something's "wrong" going on.
The rationalization is easy, it's civic duty, right? We can't complain of people's inactions if we are not then willing to step in, to expose ourselves to some modicum of risk. And I know that people, most people, even scary looking ones, really would do anything to avoid conflict. Nobody wants to get hurt, self-preservation is strong.
On the other hand, the risk is there. Crazy people do exist, especially, I've learned, in the states. The prevalence of weapons there also scary, it definitely does not feel safe down there. Vancouver is much more civilized, but still, bad stuff does happen.
And I also know that it's always sad, guys screaming on the street, harassing people, are miserable already. Almost always drunk, or worse.
It's easier to let go when, rarely when you are 6' and weight almost 100kg, they scream at you.
Once someone was drunk and racist towards my wife in an elevator, it took me a literal minute to parse the situation, the pros and cons. He was drunk, the slight was minor, at the same time - racism. In the end, he got only pinned in a corner by his neck, and screamed at. Calibrating the right amount of reaction is tricky.
Today I carried one of my cameras to my usual Saturday ritual of hipster coffee and reading cellulose.
I'd like to finish the roll so I went for a stroll, even if I don't love taking photos in that area, lots of homeless around which are not going to be my subjects - I don't consider myself entitled to document that.
I'd like to shoot more "street" though because I realized I, for some reason, became shy at it, and I didn't use to be - I don't think. I wonder if I changed or society's attitude towards having photos taken did, probably a bit of both, but I know it doesn't make sense, so I welcome the exercise.
A guy, somewhat drunk, somewhat miserable, likely older than me but not that old is screaming at everyone he encounters. Flipping the bird, that sort of stuff. It's interesting to observe, everyone mostly ignore and avoid, some end up in a staring contest, chest out.
It's a careful dance, he will be bolder with weaker, and meak with potential dangers, all whilst screaming "cowards". He surprises an elderly dragging two-wheel shopping cart by lamely stomping on it, in the end quite harmlessly.
Anyhow, considering that drunk and miserable or not, he's being an ass, I feel no guilt taking a photo. Fair game.
Of course this is seen as a provocation, which it was, and he targets me.
Great! I wish I could say it's for the photos, and in truth I do hope they'll turn out great, but in honesty, I didn't care, it's just... fun.
Isn't that odd? The photos are also easy now, once people engage (in positive or negative ways), the initial shyness is gone - which is also odd.
I actually know by now I'm perfectly safe, my main worry is not to walk on paths covered by the melting snow we have today, otherwise I could slip and fall, and I have my camera, my bag, the logistics are complicated...
I sincerely doubt he'd ever muster the courage for a punch, even if he does come to my face, even "grab" my neck, there's no real aggression. I don't even think, at least, that's what I picture, that he's scared that I'd prevail in a confrontation, even if I definitely would (not hard, drunk people make the worst fighters), I imagine he knows it's not worth going to jail.
After walking for quite a few blocks, repeating the dance a few times, we both make it out unscathed. I still don't know exactly what it tells about me, probably nothing good. I can't say I did not enjoy my friend though. It is what it is.
Sat, 20 Jan 2024 21:19:49 -0800
- AGI (permalink)
I find the agi fears to be ludicrous. One might want to go back to the previous AI summers and their predictions of singularity.
I wonder how much there are honest, a byproduct of a generation growing with Sarah Connor and Commander Data, and how much these are distractions artfully crafted to both look like a hero (Musk docet) and to hype investors...
Because agi might be dangerous but oh man imagine the market cap on being a god...
Anyhow. I'm not a ton foil hatter, so my bet is on the bullshit to be 100% organic. People that passionate and even visionary must always have a non zero amount of delusion. We saw it always, from early hackers to NFTs.
What has been fun? or whatever adjective fits better, is to see how these brute force databases can look like intelligence so convincingly. It kinda makes you think? Because this is the biggest difference, isn't it.
A person does not grow up parsing the entire written output of humanity in order to learn how to write... these machines are hyperspecialized brutes compared to our brains.
It's mostly data. Whilst i suspect we are mostly architecture. If we didn't give a baby many stimuli, would they grow with no capacity of creation? A thought experiment, obviously. I would like to imagine that the brain would be able to adapt relatively quickly when exposed to data. It does not seem to me to be a small detail.
Meanwhile, the real risks of AI seem to be mostly ignored or badly patched with shortsighted legislation. At the same time perhaps I agree with the pragmatism, it's such a developing field it would be honestly hard to draw a line between killing an innovation that could usher the next industrial revolution of productivity (importantly - without a commensurate environmental impact!) and protecting society...
Lately things seem already to be cooling down, in terms of how useful really these databases can be in replacing humans.
If it makes economical sense, the biggest threat might be to google, and I can't say I would not welcome that.
Sat, 20 Jan 2024 16:26:47 -0800
- Intention (permalink)
Not sure yet what to think of this journal section. Perhaps I will use it to drop links and commentary, as I lost long ago the energy for my famed "friday links". We will see.
I like gopher phlogs and am toying with the idea of writing a scraper. Could be fun, a mini-google for fun. At the same time I don't think it's interesting to keep a public account of my life, and for programming stuff, most of it I can't talk about.
That said I am happy with the work on the blog, i also recycled some for REAC and I did manage to setup deep learning stuff on my windows machine.
More about that later.
Sat, 20 Jan 2024 16:26:08 -0800