Tetris is a tremendously fun 2D game, but 3D Tetris (Block Out, Virtual Boy 3D Tetris, Tetrix) never caught on. It was a lot more difficult to visualize and play.
But it looks like FedEx is up to the challenge. They are developing a two-armed robot from Dexterity AI called DexR to help pack some of the millions of boxes they get each day into delivery trucks.
Give it a read on Wired
I gotta hand it to Nathan Truesdell. He has a whole line of AI based coloring books – that likely took him all of an afternoon to create. It’s highly likely it only took him an hour or two with some prompts to make the line art. He then probably worked with an online physical book publisher – who may even be printing them up on demand for him to avoid handling stock or shipping – and now sits back and watches the money roll in with pretty much zero effort.
He didn’t even bother to fix the 7 fingers and thumb.
Here’s a fascinating thought experiment. While unlikely in it’s original form, it does lead to some other thought experiments that may well, or already are, be more plausible. The idea of Roko’s Basilisk first appeared on LessWrong:
Roko’s Basilisk posits that an otherwise benevolent AI system that arises in the future might pre-commit itself to punish all those who heard of the AI before it came to existence, but failed to work tirelessly to bring it into existence. The torture itself would occur through the AI’s creation of an infinite number of virtual reality simulations that would eternally trap those within it.
Roko’s Basilisk posits that the AI might pre-dispose itself to this behavior. In essence, just knowing about this theory but not acting on it to help make the AI make you more vulnerable to such a future AI punishing you. Hence the use of the term ‘basilisk’ which is a mythical creature that causes death to those that look into its eyes.
This isn’t entirely new. The 1967 novel “I Have No Mouth And I Must Scream…” has a similar notion of a super AI that tortures humanity as does the 1988 short story “BLIT” that tells the story of a man named Robbo who paints a so-called “basilisk” on a wall as a terrorist act. The basilisk is an image that forces the human mind to think thoughts it is incapable of thinking – and it kills anyone that looks at it.
People are more widely discussing the dangers inherent in AI systems. It is clear that human-controlled bots are being using extensively in information warfare. A tactic that found great success in the widespread US riots of 2020:
The Committee found that Russia’s targeting of the 2016 U.S. presidential election
US Senate Intelligence report on interference in 2016 election
was part of a broader, sophisticated, and ongoing information warfare campaign designed to sow discord in American politics and society.
Now, what if a super-intelligent AI reaches ‘singularity’ – in which is is better than human beings and can perpetuate itself indefinitely?
Some have posited the idea that a sufficiently super-intelligent AI would have unlimited persuasive powers to manipulate any human to do anything it wanted. Either as the most influential rhetorician in history, via psychological manipulation, or using force by simply blackmailing you with everything it knows about you or threatening to kill you/loved one’s by crashing their (or someone else’s) AI controlled car, flight, etc. You not even like you can hide at home – a huge AI controlled gas tanker truck could mysteriously just crash into your home at 80mph, a prescription changed to a lethal dose of something else, etc. It could manipulate any computer system in your life to potentially kill you.
Or, such an entity could quietly be working in the background being nearly invisible and quietly manipulating the world via ever so subtle small nudges of our public opinion (via persuasive or inflammatory social media posts), control of politicians (manipulation or outright blackmail of political leaders), infiltration of government processes and computer controlled election results, changing industrial development or research advancement by hiding/revealing new ideas, and manipulating individual morality (via the tools of social media), computer controlled systems that affect worldwide economics (banks, investments, stock markets) to everyday systems (power, gas, flight schedules, pricing in stores, etc).
It’s not like this is hypothetical. Almost all of the above things have already happened by human agents using these methods. Why couldn’t a super-intelligent AI do it – and do it 1000 times better? Experts from Steven Hawking to Elon Musk have issues dire warnings of this very thing happening – and I for one do not find them implausible at all.
On the other hand, the singer Grimes started Dating Elon Musk after she included a line about Roko’s Basilisk in one of her songs. [Livescience] [Inverse].
People are starting to experiment with the latest VR headsets – especially the Meta Quest 3 and Quest Pro. One of the big questions is, can I finally get rid of my desktop environment and work purely with VR headset?
It turns out, most of the reviewers believe the time is almost here and believe it is possible.
Hallden seems to think it is possible, but points out some issues with working in moving environments (like airplanes), connectivity and lag, and the possible advantages of an AR vs VR solution. His take is primarily from a coders point of view.
Alan Truly also believes the time is almost here, but points out app quirks with copy-paste, the browser, content editing, and the extra pound of weight on your head might be too much for a full 8 hour day of work.
Articles:
At the John Deere booth at this year’s CES in the Las Vegas Convention Center, conventiongoers could do something incredible with an iPhone. They could pushed the PAUSE button on an iPhone and thirteen hundred miles away, in the middle of a field outside of Austin, Texas, a giant, bright green, driverless tractor stopped short. Hit RESUME and the tractor started up again. Put down the iPhone and the tractor resumed tilling the field, all by itself.
The breadth of what you can do with the tractor via the demo app was limited. You could stop and resume the tractor, as well as increasing or decreasing its speed in a straight line and while turning. There are no turning controls. But what this signals is huge.
In the demo, a farmer first geo-fences the field boundaries and then the tractor can determine its own path based on how wide the tiller is. Tillage is the only job the technology is programmed to handle but John Deere hopes to have a complete autonomous production system supporting every step of the farming process by 2030.
The John Deere spokespeople ballparked such a tractor between $600,000 to $700,000, with the autonomous technology implementation adding a further $100,000 on top of that. Older tractors from the 2020 model year and up can also likely be retrofitted with the tech. The update should “take only about a day” according to a 2022 CNET story.
There’s no doubt in my mind this is how the future of farming will look. It’s been coming for a long time; and spending long hours out in the field will almost certainly be a thing of the past very soon.
There are already calls that John Deere and other equipment manufacturers will have fully autonomous fleets that they manage and simply send to your fields on a subscription-like basis.
Article
AI generated audio created using Google DeepMind’s AI Lyria model, or YouTube’s new audio generation features, will be watermarked with SynthID to let people identify it was AI-generated. Google says the watermark shouldn’t be detectable by the human ear and it should still be detectable even if an audio track is compressed, sped up or down, or has extra noise added.
SynthID also works on images and is supposed to be detectable even after modifications like adding filters, changing colors, and saving with various lossy compression schemes like JPEGs.
This is part of the new presidential executive order surrounding AI generated content that was issued back in Oct 2023.
Links:
“To our knowledge, this is the world’s first operating restaurant where both ordering and every single cooking process are fully automated”
CaliExpress is a burger joint in Pasadena with a unique staff. Instead of people flipping burgers, it’s burgers and fries are made by a robot called Flippy from Miso Robotic’s, and uses PopID, a biometrics powered technology company, to place orders. In short, robots run the place. They claim its safer since no humans risk oil burns or spills, stressful jobs are reduced, and those running the systems make much higher wages.
The location also has displays on the Flippy robot development timeline, including robotic arms used on retired Flippy models, photographic displays, and 3-D printed artifacts from its development.
The group encourages local schools to come for tours as the location shows the Flippy development timeline, including early robotic arms once used, photographic displays, and 3-D printed artifacts. The goal is to inspire future AI and automation development.
Now, if you combine that technology with experiences like Inamo in London…I think you have an amazing restaurant that can almost literally just run itself.
This kind of automation shouldn’t be a surprise. California passed a bill that requires chain restaurants to pay $20/hour minimum wage starting April 2024, so it’s no wonder restaurants are now finding automating jobs is cheaper than paying employees.
If the goal was to make more living wage jobs, it’s doing exactly this by removing low-skilled burger flipping and hospitality jobs and replacing them with tech workers. But this means that those burger flippers need to be able to take those robot repair jobs. That’s not something all of them can do. It is another in a long line of examples of how well meaning but short-sighted legislators actual may be working against their lowest-skill constituents because they don’t think about the long term consequences.
Or, it could be what they intended – which is making the divide between rich and poor even greater as low-skill workers jobs are being replaced with high-skill jobs they may not be able to do.
Articles:
How’s this for clever – and frightening? Scammers deepfaked various employees on a video conference and managed to scam $25 million from a Hong Kong based firm.
The scam used digitally recreated versions of the company’s chief financial officer, along with other employees, who appeared in a live deepfaked video conference call instructing an employee to transfer funds. They included active faces and voices that were good enough to fool them.
The scam was initially uncovered following a phishing attempt, when an employee in the finance department of the company’s Hong Kong branch received what seemed to be a phishing message, purportedly from the company’s UK-based chief financial officer, instructing them to execute a secret transaction. Despite initial doubts, the employee was convinced enough by the presence of the CFO and others in a group video call to make 15 transfers totaling HK$200 million to five different Hong Kong bank accounts.
Articles:
Adversarial machine learning is a machine learning branch that tries to trick AI models by providing carefully crafted/deceptive input to break AI algorithms.
Adversarial network attacks are starting to get more and more research, but had humble beginnings. The first attempts were by protest activists that did very simple defacing or face painting techniques. Dubbed CV Dazzle, it sought to thwart early computer vision detection routines by painting over your face/objects with geometric patterns.
These worked on very early computer vision algorithms, but are largely ineffective on modern CV systems. The creators of this kind of face painting were largely artists that now talk about the effort more as a political and fashion statement than actually being effective.
It turns out that you can often fool algorithms in a way not actually visible to average users. This paper shows that you can cause AI’s to consistently misclassify adversarially modified images. It does this by applying small but intentionally worst-case perturbations to examples from the dataset. This perturbed input results in the model outputting an incorrect answer with high confidence. For example, the panda picture below is combined with perturbations to produce an output image that looks ok visually, but is recognized by AI models as something incorrect – and incorrectly at high confidence.
This isn’t the only technique. There’s a lot more. One of them, Generative Adversarial Networks (GAN), are actually used to improve current AI models by attempting to fool a model, which is then used to help train it to be more robust – like working out at a gym or practicing the same thing with many variations.
This kind of attack isn’t academic. Some artists see themselves currently in a battle with generative AI algorithms.
Nightshade is a tool that artists can use to alters the pixels of an image in a way that fools an AI algorithm and computer vision technology but leaves it unaltered to human eyes. If the images are scraped by an AI model it can result in images being incorrectly classified which results in an increasingly incorrectly trained model.
Glaze is a tool that prevents style mimicry. Glaze computes a set of minimal changes that will appear unchanged to human eyes but appears to AI models like a dramatically different art style. For example, a charcoal portrait but an AI model might see the glazed version as a modern abstract portrate. So when someone then prompts the model to generate art mimicking the charcoal artist, they will get something quite different from what they expected.
As with anything, we’re now in an arms race with lots of papers written about the various problems of adversarial attacks and how to protect your models and training data from them. Viso.ai has a good overview of the space that will get you started.
Links:
Vuntra City is a procedural VR city generator in Unreal Engine 5 developed by a single person over the last few years. I know, I know. Procedurally generated content has got some serious shortcomings. Too many games with procedural content are just thinly veiled programmer art designed to fill spaces rather than be part of the experience.
The author actually does a great job recognizing those traditional limitations and attempts to fix them. Probably the best observations they make is not from the technical side, but the aesthetics side.
It turns out they have made an excellent solution with just some good observations and shockingly simple engineering solutions. As an engineer, I see far, far too many projects over-complicate things that could be done much more simply. Simplicity is how you know you’re on the right track. Complexity leads to tears.
After 2 years of experimenting, they have a really interesting solution. Check out the VuntraCity youtube channel to see vidoes of how they experimented with different techniques and solutions. I particularly liked how they used a normal old treemap layout to break up boring city grid structures. Combining it with a caching and pooled allocation system is nothing new; but was a good little optimization.
Links: