More on Technology

Amelia Winger-Bearskin
3 years ago
Reasons Why AI-Generated Images Remind Me of Nightmares
AI images are like funhouse mirrors.
Google's AI Blog introduced the puppy-slug in the summer of 2015.
Puppy-slug isn't a single image or character. "Puppy-slug" refers to Google's DeepDream's unsettling psychedelia. This tool uses convolutional neural networks to train models to recognize dataset entities. If researchers feed the model millions of dog pictures, the network will learn to recognize a dog.
DeepDream used neural networks to analyze and classify image data as well as generate its own images. DeepDream's early examples were created by training a convolutional network on dog images and asking it to add "dog-ness" to other images. The models analyzed images to find dog-like pixels and modified surrounding pixels to highlight them.
Puppy-slugs and other DeepDream images are ugly. Even when they don't trigger my trypophobia, they give me vertigo when my mind tries to reconcile familiar features and forms in unnatural, physically impossible arrangements. I feel like I've been poisoned by a forbidden mushroom or a noxious toad. I'm a Lovecraft character going mad from extradimensional exposure. They're gross!
Is this really how AIs see the world? This is possibly an even more unsettling topic that DeepDream raises than the blatant abjection of the images.
When these photographs originally circulated online, many friends were startled and scandalized. People imagined a computer's imagination would be literal, accurate, and boring. We didn't expect vivid hallucinations and organic-looking formations.
DeepDream's images didn't really show the machines' imaginations, at least not in the way that scared some people. DeepDream displays data visualizations. DeepDream reveals the "black box" of convolutional network training.
Some of these images look scary because the models don't "know" anything, at least not in the way we do.
These images are the result of advanced algorithms and calculators that compare pixel values. They can spot and reproduce trends from training data, but can't interpret it. If so, they'd know dogs have two eyes and one face per head. If machines can think creatively, they're keeping it quiet.
You could be forgiven for thinking otherwise, given OpenAI's Dall-impressive E's results. From a technological perspective, it's incredible.
Arthur C. Clarke once said, "Any sufficiently advanced technology is indistinguishable from magic." Dall-magic E's requires a lot of math, computer science, processing power, and research. OpenAI did a great job, and we should applaud them.
Dall-E and similar tools match words and phrases to image data to train generative models. Matching text to images requires sorting and defining the images. Untold millions of low-wage data entry workers, content creators optimizing images for SEO, and anyone who has used a Captcha to access a website make these decisions. These people could live and die without receiving credit for their work, even though the project wouldn't exist without them.
This technique produces images that are less like paintings and more like mirrors that reflect our own beliefs and ideals back at us, albeit via a very complex prism. Due to the limitations and biases that these models portray, we must exercise caution when viewing these images.
The issue was succinctly articulated by artist Mimi Onuoha in her piece "On Algorithmic Violence":
As we continue to see the rise of algorithms being used for civic, social, and cultural decision-making, it becomes that much more important that we name the reality that we are seeing. Not because it is exceptional, but because it is ubiquitous. Not because it creates new inequities, but because it has the power to cloak and amplify existing ones. Not because it is on the horizon, but because it is already here.

caroline sinders
3 years ago
Holographic concerts are the AI of the Future.
A few days ago, I was discussing dall-e with two art and tech pals. One artist acquaintance said she knew a frightened illustrator. Would the ability to create anything with a click derail her career? The artist feared this. My curator friend smiled and said this has always been a dread among artists. When the camera was invented, didn't painters say this? Even in the Instagram era, painting exists.
When art and technology collide, there's room for innovation, experimentation, and fear — especially if the technology replicates or replaces art making. What is art's future with dall-e? How does technology affect music, beyond visual art? Recently, I saw "ABBA Voyage," a holographic ABBA concert in London.
"Abba voyage?" my phone asked in early March. A Gen X friend I met through a fashion blogging ring texted me.
"What's abba Voyage?" I asked while opening my front door with keys and coffee.
We're going! Marti, visiting London, took me to a show.
"Absolutely no ABBA songs here." I responded.
My parents didn't play ABBA much, so I don't know much about them. Dad liked Jimi Hendrix, Cream, Deep Purple, and New Orleans jazz. Marti told me ABBA Voyage was a holographic ABBA show with a live band.
The show was fun, extraordinary fun. Nearly everyone on the dance floor wore wigs, ankle-breaking platforms, sequins, and bellbottoms. I saw some millennials and Zoomers among the boomers.
I was intoxicated by the experience.
Automatons date back to the 18th-century mechanical turk. The mechanical turk was a chess automaton operated by a person. The mechanical turk seemed to perform like a human without human intervention, but it required a human in the loop to work properly.
Humans have used non-humans in entertainment for centuries, such as puppets, shadow play, and smoke and mirrors. A show can have animatronic, technological, and non-technological elements, and a live show can blur real and illusion. From medieval puppet shows to mechanical turks to AI filters, bots, and holograms, entertainment has evolved over time.
I'm not a hologram skeptic, but I'm skeptical of technology, especially since I work with it. I love live performances, I love hearing singers breathe, forget lines, and make jokes. Live shows are my favorite because I love watching performers make mistakes or interact with the audience. ABBA Voyage was different.
Marti and I traveled to Manchester after ABBA Voyage to see Liam Gallagher. Similar but different vibe. Similar in that thousands dressed up for the show. ABBA's energy was dizzying. 90s chic replaced sequins in the crowd. Doc Martens, nylon jackets, bucket hats, shaggy hair. The Charlatans and Liam Gallagher opened and closed, respectively. Fireworks. Incredible. People went crazy. Yelling exhausted my voice.
This week in music featured AI-enabled holograms and a decades-old rocker. Both are warm and gooey in our memories.
After seeing both, I'm wondering if we need AI hologram shows. Why? Is it good?
Like everything tech-related, my answer is "maybe." Because context and performance matter. Liam Gallagher and ABBA both had great, different shows.
For a hologram to work, it must be impossible and big. It must be big, showy, and improbable to justify a hologram. It must feel...expensive, like a stadium pop show. According to a quick search, ABBA broke up on bad terms. Reuniting is unlikely. This is also why Prince or Tupac hologram shows work. We can only engage with their legacy through covers or...holograms.
I drove around listening to the radio a few weeks ago. "Dreaming of You" by Selena played. Selena's music defined my childhood. I sang along and turned up the volume (or as loud as my husband would allow me while driving on the highway).
I discovered Selena's music six months after her death, so I never saw her perform live. My babysitter Melissa played me her album after I moved to Houston. Melissa took me to see the Selena movie five times when it came out. I quickly wore out my VHS copy. I constantly sang "Bibi Bibi Bom Bom" and "Como la Flor." I love Selena. A Selena hologram? Yes, probably.
Instagram advertised a cellist's Arthur Russell tribute show. Russell is another deceased artist I love. I almost walked down the aisle to "This is How We Walk on the Moon," but our cellist couldn't find it. Instead, I walked to Magnetic Fields' "The Book of Love." I "discovered" Russell after a friend introduced me to his music a few years ago.
I use these as analogies for the Liam Gallagher and ABBA concerts.
You have no idea how much I'd pay to see a hologram of Selena's 1995 Houston Livestock Show and Rodeo concert. Arthur Russell's hologram is unnecessary. Russell's work was intimate and performance-based. We can't separate his life from his legacy; popular audiences overlooked his genius. He died of AIDS broke. Like Selena, he died prematurely. Given his music and history, another performer would be a better choice than a hologram. He's no Selena. Selena could have rivaled Beyonce.
Pop shows' size works for holograms. Along with ABBA holograms, there was an anime movie and a light show that would put Tron to shame. ABBA created a tourable stadium show. The event was lavish, expensive, and well-planned. Pop, unlike rock, isn't gritty. Liam Gallagher hologram? No longer impossible, it wouldn't work. He's touring. I'm not sure if a rockstar alone should be rendered as a hologram; it was the show that made ABBA a hologram.
Holograms, like AI, are part of the future of entertainment, but not all of it. Because only modern interpretations of Arthur Russell's work reveal his legacy. That's his legacy.
Large-scale arena performers may use holograms in the future, but the experience must be impossible. A teacher once said that the only way to convey emotion in opera is through song, and I feel the same way about holograms, AR, VR, and mixed reality. A story's impossibility must make sense, like in opera. Impossibility and bombastic performance must be present for an immersive element to "work." ABBA was an impossible and improbable experience, which made it magical. It helped the holographic show work.
Marti told me about ABBA Voyage. She said it was a great concert. Marti has worked in music since the 1990s. She's a music expert; she's seen many shows.
Ai isn't a god or sentient, and the ABBA holograms aren't real. The renderings were glassy-eyed, flat, and robotic, like the Polar Express or the Jaws shark. Even today, the uncanny valley is insurmountable. We know it's not real because it's not about reality. It was about a suspended moment and performance feelings.
I knew this was impossible, an 'unreal' experience, but the emotions I felt were real, like watching a movie or tv show. Perhaps this is one of the better uses of AI, like CGI and special effects, like the beauty of entertainment- we were enraptured and entertained for hours. I've been playing ABBA since then.

Shawn Mordecai
3 years ago
The Apple iPhone 14 Pill is Easier to Swallow
Is iPhone's Dynamic Island invention or a marketing ploy?
First of all, why the notch?
When Apple debuted the iPhone X with the notch, some were surprised, confused, and amused by the goof. Let the Brits keep the new meaning of top-notch.
Apple removed the bottom home button to enhance screen space. The tides couldn't overtake part of the top. This section contained sensors, a speaker, a microphone, and cameras for facial recognition. A town resisted Apple's new iPhone design.
From iPhone X to 13, the notch has gotten smaller. We expected this as technology and engineering progressed, but we hated the notch. Apple approved. They attached it to their other gadgets.
Apple accepted, owned, and ran with the iPhone notch, it has become iconic (or infamous); and that’s intentional.
The Island Where Apple Is
Apple needs to separate itself, but they know how to do it well. The iPhone 14 Pro finally has us oohing and aahing. Life-changing, not just higher pixel density or longer battery.
Dynamic Island turned a visual differentiation into great usefulness, which may not be life-changing. Apple always welcomes the controversy, whether it's $700 for iMac wheels, no charging block with a new phone, or removing the headphone jack.
Apple knows its customers will be loyal, even if they're irritated. Their odd design choices often cause controversy. It's calculated that people blog, review, and criticize Apple's products. We accept what works for them.
While the competition zigs, Apple zags. Sometimes they zag too hard and smash into a wall, but we talk about it anyways, and that’s great publicity for them.
Getting Dependent on the drug
The notch became a crop. Dynamic Island's design is helpful, intuitive, elegant, and useful. It increases iPhone usability, productivity (slightly), and joy. No longer unsightly.
The medication helps with multitasking. It's a compact version of the iPhone's Live Activities lock screen function. Dynamic Island enhances apps and activities with visual effects and animations whether you engage with it or not. As you use the pill, its usefulness lessens. It lowers user notifications and consolidates them with live and permanent feeds, delivering quick app statuses. It uses the black pixels on the iPhone 14's display, which looked like a poor haircut.
The pill may be a gimmick to entice customers to use more Apple products and services. Apps may promote to their users like a live billboard.
Be prepared to get a huge dose of Dynamic Island’s “pill” like you never had before with the notch. It might become so satisfying and addicting to use, that every interaction with it will become habit-forming, and you’re going to forget that it ever existed.
WARNING: A Few Potential Side Effects
Vision blurred Dynamic Island's proximity to the front-facing camera may leave behind grease that blurs photos. Before taking a selfie, wipe the camera clean.
Strained thumb To fully use Dynamic Island, extend your thumb's reach 6.7 inches beyond your typical, comfortable range.
Happiness, contentment The Dynamic Island may enhance Endorphins and Dopamine. Multitasking, interactions, animations, and haptic feedback make you want to use this function again and again.
Motion-sickness Dynamic Island's motions and effects may make some people dizzy. If you can disable animations, you can avoid motion sickness.
I'm not a doctor, therefore they aren't established adverse effects.
Does Dynamic Island Include Multiple Tasks?
Dynamic Islands is a placebo for multitasking. Apple might have compromised on iPhone multitasking. It won't make you super productive, but it's a step up.
iPhone is primarily for personal use, like watching videos, messaging friends, sending money to friends, calling friends about the money you were supposed to send them, taking 50 photos of the same leaf, investing in crypto, driving for Uber because you lost all your money investing in crypto, listening to music and hailing an Uber from a deserted crop field because while you were driving for Uber your passenger stole your car and left you stranded, so you used Apple’s new SOS satellite feature to message your friend, who still didn’t receive their money, to hail you an Uber; now you owe them more money… karma?
We won't be watching videos on iPhones while perusing 10,000-row spreadsheets anytime soon. True multitasking and productivity aren't priorities for Apple's iPhone. Apple doesn't to preserve the iPhone's experience. Like why there's no iPad calculator. Apple doesn't want iPad users to do math, but isn't essential for productivity?
Digressing.
Apple will block certain functions so you must buy and use their gadgets and services, immersing yourself in their ecosystem and dictating how to use their goods.
Dynamic Island is a poor man’s multi-task for iPhone, and that’s fine it works for most iPhone users. For substantial productivity Apple prefers you to get an iPad or a MacBook. That’s part of the reason for restrictive features on certain Apple devices, but sometimes it’s based on principles to preserve the integrity of the product, according to Apple’s definition.
Is Apple using deception?
Dynamic Island may be distracting you from a design decision. The answer is kind of. Elegant distraction
When you pull down a smartphone webpage to refresh it or minimize an app, you get seamless animations. It's not simply because it appears better; it's due to iPhone and smartphone processing speeds. Such limits reduce the system's response to your activity, slowing the experience. Designers and developers use animations and effects to distract us from the time lag (most of the time) and sometimes because it looks cooler and smoother.
Dynamic Island makes apps more useable and interactive. It shows system states visually. Turn signal audio and visual cues, voice assistance, physical and digital haptic feedbacks, heads-up displays, fuel and battery level gauges, and gear shift indicators helped us overcome vehicle design problems.
Dynamic Island is a wonderfully delightful (and temporary) solution to a design “problem” until Apple or other companies can figure out a way to sink the cameras under the smartphone screen.
Apple Has Returned to Being an Innovative & Exciting Company
Now Apple's products are exciting. Next, bring back real Apple events, not pre-recorded demos.
Dynamic Island integrates hardware and software. What will this new tech do? How would this affect device use? Or is it just hype?
Dynamic Island may be an insignificant improvement to the iPhone, but it sure is promising for the future of bridging the human and computer interaction gap.
You might also like

Vitalik
3 years ago
An approximate introduction to how zk-SNARKs are possible (part 1)
You can make a proof for the statement "I know a secret number such that if you take the word ‘cow', add the number to the end, and SHA256 hash it 100 million times, the output starts with 0x57d00485aa". The verifier can verify the proof far more quickly than it would take for them to run 100 million hashes themselves, and the proof would also not reveal what the secret number is.
In the context of blockchains, this has 2 very powerful applications: Perhaps the most powerful cryptographic technology to come out of the last decade is general-purpose succinct zero knowledge proofs, usually called zk-SNARKs ("zero knowledge succinct arguments of knowledge"). A zk-SNARK allows you to generate a proof that some computation has some particular output, in such a way that the proof can be verified extremely quickly even if the underlying computation takes a very long time to run. The "ZK" part adds an additional feature: the proof can keep some of the inputs to the computation hidden.
You can make a proof for the statement "I know a secret number such that if you take the word ‘cow', add the number to the end, and SHA256 hash it 100 million times, the output starts with 0x57d00485aa". The verifier can verify the proof far more quickly than it would take for them to run 100 million hashes themselves, and the proof would also not reveal what the secret number is.
In the context of blockchains, this has two very powerful applications:
- Scalability: if a block takes a long time to verify, one person can verify it and generate a proof, and everyone else can just quickly verify the proof instead
- Privacy: you can prove that you have the right to transfer some asset (you received it, and you didn't already transfer it) without revealing the link to which asset you received. This ensures security without unduly leaking information about who is transacting with whom to the public.
But zk-SNARKs are quite complex; indeed, as recently as in 2014-17 they were still frequently called "moon math". The good news is that since then, the protocols have become simpler and our understanding of them has become much better. This post will try to explain how ZK-SNARKs work, in a way that should be understandable to someone with a medium level of understanding of mathematics.
Why ZK-SNARKs "should" be hard
Let us take the example that we started with: we have a number (we can encode "cow" followed by the secret input as an integer), we take the SHA256 hash of that number, then we do that again another 99,999,999 times, we get the output, and we check what its starting digits are. This is a huge computation.
A "succinct" proof is one where both the size of the proof and the time required to verify it grow much more slowly than the computation to be verified. If we want a "succinct" proof, we cannot require the verifier to do some work per round of hashing (because then the verification time would be proportional to the computation). Instead, the verifier must somehow check the whole computation without peeking into each individual piece of the computation.
One natural technique is random sampling: how about we just have the verifier peek into the computation in 500 different places, check that those parts are correct, and if all 500 checks pass then assume that the rest of the computation must with high probability be fine, too?
Such a procedure could even be turned into a non-interactive proof using the Fiat-Shamir heuristic: the prover computes a Merkle root of the computation, uses the Merkle root to pseudorandomly choose 500 indices, and provides the 500 corresponding Merkle branches of the data. The key idea is that the prover does not know which branches they will need to reveal until they have already "committed to" the data. If a malicious prover tries to fudge the data after learning which indices are going to be checked, that would change the Merkle root, which would result in a new set of random indices, which would require fudging the data again... trapping the malicious prover in an endless cycle.
But unfortunately there is a fatal flaw in naively applying random sampling to spot-check a computation in this way: computation is inherently fragile. If a malicious prover flips one bit somewhere in the middle of a computation, they can make it give a completely different result, and a random sampling verifier would almost never find out.
It only takes one deliberately inserted error, that a random check would almost never catch, to make a computation give a completely incorrect result.
If tasked with the problem of coming up with a zk-SNARK protocol, many people would make their way to this point and then get stuck and give up. How can a verifier possibly check every single piece of the computation, without looking at each piece of the computation individually? There is a clever solution.
see part 2

Datt Panchal
3 years ago
The Learning Habit
The Habit of Learning implies constantly learning something new. One daily habit will make you successful. Learning will help you succeed.
Most successful people continually learn. Success requires this behavior. Daily learning.
Success loves books. Books offer expert advice. Everything is online today. Most books are online, so you can skip the library. You must download it and study for 15-30 minutes daily. This habit changes your thinking.
Typical Successful People
Warren Buffett reads 500 pages of corporate reports and five newspapers for five to six hours each day.
Each year, Bill Gates reads 50 books.
Every two weeks, Mark Zuckerberg reads at least one book.
According to his brother, Elon Musk studied two books a day as a child and taught himself engineering and rocket design.
Learning & Making Money Online
No worries if you can't afford books. Everything is online. YouTube, free online courses, etc.
How can you create this behavior in yourself?
1) Consider what you want to know
Before learning, know what's most important. So, move together.
Set a goal and schedule learning.
After deciding what you want to study, create a goal and plan learning time.
3) GATHER RESOURCES
Get the most out of your learning resources. Online or offline.

Tora Northman
3 years ago
Pixelmon NFTs are so bad, they are almost good!
Bored Apes prices continue to rise, HAPEBEAST launches, Invisible Friends hype continues to grow. Sadly, not all projects are as successful.
Of course, there are many factors to consider when buying an NFT. Is the project a scam? Will the reveal derail the project? Possibly, but when Pixelmon first teased its launch, it generated a lot of buzz.
With a primary sale mint price of 3 ETH ($8,100 USD), it started as an expensive project, with plenty of fans willing to invest in what was sold as a game. After it was revealed, it fell rapidly.
Why? It was overpromised and under delivered.
According to the project's creator[^1], the funds generated will be used to develop the artwork. "The Pixelmon reveal was wrong. This is what our Pixelmon look like in-game. "Despite the fud, I will not go anywhere," he wrote on Twitter. The goal remains. The funds will still be used to build our game. I will finish this project."
The project raised $70 million USD, but the NFTs buyers received were not the project's original teasers. Some call it "the worst NFT project ever," while others call it a complete scam.
But there's hope for some buyers. Kevin emerged from the ashes as the project was roasted over the fire.
A Minecraft character meets Salad Fingers - that's Kevin. He's a frog-like creature whose reveal was such a terrible NFT that it became part of history – and a meme.
If you're laughing at people paying $8K for a silly pixelated image, you might need to take it back. Precisely because of this, lucky holders who minted Kevin have been able to sell the now-memed NFT for over 8 ETH (around $24,000 USD), with some currently listed for 100 ETH.
Of course, Twitter has been awash in memes mocking those who invested in the project, because what else can you do when so many people lose money?
It's still unclear if the NFT project is a scam, but the team behind it was hired on Upwork. There's still hope for redemption, but Kevin's rise to fame appears to be the only positive outcome so far.
[^1] This is not the first time the creator (A 20-yo New Zealanders) has sought money via an online platform and had people claiming he under-delivered. He raised $74,000 on Kickstarter for a card game called Psycho Chicken. There are hundreds of comments on the Kickstarter project saying they haven't received the product and pleading for a refund or an update.
