The housing shortage in Portland seems to be improving – until you look at why. It’s not because Portland (with some of the most housing unfriendly policies in the country) is building more houses (they are a little bit) but because people are leaving.
What is happening in San Francisco—and in Portland, to a lesser extent—is that the housing shortage is temporarily improving, not because of the construction of new housing but because people are moving away to cheaper places.
In a case study on San Francisco, the Up for Growth report noted an issue that a report from the Common Sense Institute Oregon highlighted earlier this week.
“Household formation in these areas significantly decreased. Driven in no small part by the high cost of housing, households that could have formed in these communities formed somewhere else instead
Oasis is a free to play online Minecraft clone built by an Israeli AI startup called Decart in collaboration with Etched – a company that designs custom silicon to demonstrate hardware optimized transformer-based AI algorithms.
Minecraft clones aren’t anything new, but Oasis is. That’s because nobody programmed it. It is completely AI generated rendering – just like the realtime AI generated Doom clone I wrote about earlier. Everything is generated via realtime AI rendering via transformer based network trained to generate minecraft-like output. No gameplay was programmed.
Like a lot of generative AI, it has a tendency to hallucinate when you look at things you have looked at before (textures will change, blur, etc). It certainly won’t be replacing minecraft anytime soon, but these kinds of projects show how one might use AI to do proof-of-concept work.
The Dead by Daylight Twitch streamer Hens333 was curious about a streamer’s meteoric rise, and couldn’t figure it out. The streamer didn’t appear to have any decent content, didn’t seem very big in the community – but was getting massive viewer counts.
So, he dug into it and then went on a massive data deep-dive and believes he found what the internet dreads – that it could likely be a whole empire of fake views and purchased popularity – all to get lucrative sponsorships.
Doing a little napkin math, he is likely spending $1000-2000 per month for bot accounts based on his viewer counts. Why would you do this? Ego?
A quick look at his Twitch page clears it all up. Sponsorships. His Twitch page is littered with dozens and dozens of sponsors.
If all this is true, then it means he might be making $10,000’s per month on sponsors with only a $1000-$2000 per month outlay. Some note he isn’t even playing the games many times – just showing replays of previous sessions. One more highly suspicious data point is that just about ALL his youtube videos have less than 50 views:
The streamer in question is realzbluewater, and if he is doing this, he is likely committing criminal fraud by misrepresenting his actual viewer count to advertisers.
This could come with serious legal repercussions as well as criminal charges. He wouldn’t be the first. Charlie Javiceis now up on charges that she deceived JP Morgan about the number of actual users of her college financial planning company called Frank. They alleged that Javice even paid a data-science professor $18,000 for a list of more than four million fake student names. She claims JP Morgan simply didn’t do their due diligence. Right now, she’s up on Federal charges for securities fraud, wire fraud, bank fraud, and conspiracy.
I personally believe there is a lot of the kind of fraud Javice is accused of. It’s been pretty much rampant by people on Forbes 30 under 30 list (more comically being called Forbes 30 doing 30 list). It’s also ridiculously easy. Purchasing views and reposts for any social media account is just a google click away. You can buy thousands of re-Tweets, Twitch views, Youtube views, fake comments, Instagram followers, Facebook upvotes – and just about anything else for just a little bit of money. Just google it and you’ll find tons of services that do all these things for under $50.
I sure hope realzbluewater is legit. If he’s not, he’s likely committing criminal fraud and could be sued by his sponsors and face jail time.
Twitch is likely not cracking down for 3 reasons. Firstly, they get ad revenue if the views or real or not. Secondly, even if they detect this, they can’t ban streamers when they detect bots swarming a channel or others could use those kinds of attacks to get streamers taken down or banned. Finally, the increase of numbers helps Twitch too. Twitch has been struggling financially and definitely could use all the viewers they can get – even if the majority of them are bots.
Just one more likely datapoint for the Dead Internet Theory. If this kind of fraud is becoming more rampant (and I believe it is), there is going to be a reckoning. Advertisers are going to start asking questions about fake numbers and realize they’re getting ripped off. When the reckoning starts, people better have long, clean records when the investigations begins – because the hosting sites will blame the streamers, and the streamers will likely be thrown under the bus.
Note to self: great business idea. Start a company that does user/data verification and fraud analysis services. When the day comes, everyone will want you.
Starpowerdrummer is basically a human Drum and Bass/Jungle drum machine. Listen to him play in real life what was previously only possible on a drum machine. He’s so unbelievably clean and exact – it’s kind of mind boggling.
Trying to convince AI they are living in a simulation
Hunter Irving picked up a 1986 Macintosh Plus and helped create MacProxy Plus, an open-source app that lets vintage Macs browse the modern web.
He uses a BlueSCSI device to emulate a rare mac ethernet adapter (Daynaport SCSI/Link-T) and a Macproxy to convert modern web pages to something 90’s era html only browsers can display. He improved Macproxy to have modular components with custom handling for specific websites. Thus, MacProxy Plus. He used claude.ai to help write some of the proxy.
He then went on to handle images – and video – using dithering and generated ASCII art.
AI Warehouse tasked a group of five AI agents to complete a 100-meter dash. Each was trained using Deep Reinforcement Learning and each agent has different physical characteristics. It’s kind of like watching AI play QWOP.
AI Retro Rewind has a few AI generated channels going for Christmas. This one combines AI generated 50’s retro-futuristic images with AI jazzy tunes. Oh, what the AI world has wrought.
