One of my favorite local Christmas shows has to be the Funhouse Lounge’s version of Die Hard: The Musical Parody. It’s an amazing show and sells out extremely fast each year.
Definitely give it a look if you get a chance.
AI trends pop up just about every week. The latest is knolling photos. Knolling photos are pictures of arranged objects so they are in parallel or 90-degree angles. The images create an organized and clean portrayal of many related things. They often look like exploded parts lists.
Tokenized AI by Christian Heidorn walks you through how to craft prompts to generate what you’re looking for. It’s a great example of how a prompt engineer sorts through creating what they want.
Learning how to make good ChatGPT prompts is a bit of an art as much as science. It’s not about learning a fixed prompts; it’s about mastering the art of well-structured, clear, and specific prompts that cater to your unique needs and interests.
This article has a bunch of great examples that can help you craft your own prompts:
Researchers from HYAS Labs demonstrated a proof-of-concept, artificial intelligence (AI)-driven cyberattack that changes its code on the fly and can slip past the latest automated security-detection technology. Called BlackMamba, it exploits a large language model (LLM) to create a “truly polymorphic” attack in that every time it executes, it resynthesizes its keylogging capability.
AI-powered attacks like this will become more common now as threat actors create polymorphic malware that leverages ChatGPT and other sophisticated, data-intelligence systems based on LLM
Read more about the BlackMamba attack outlined in this blog post
Using carefully crafted and refined queries, users have been getting around the security features of LLM’s for all kinds of funny, and nefarious, purposes.
Original called DAN attacks (Do Anything Now), users figured out how to avoid OpenAI’s policies against generating illegal or harmful material.
YouTuber Enderman showed how he was able to entice OpenAI’s ChatGPT to generate keys for Windows 95 despite the chatbot being explicitly antagonistic to creating activation keys.
Other users have been able to get chatbots to generate everything from conspiracy theories, promote violence, generate conspiracy theories, and even go on racist tirades.
Researcher Alex Polyakov created a “universal” DAN attack, which works against multiple large language models (LLMs)—including GPT-4, Microsoft’s Bing chat system, Google’s Bard, and Anthropic’s Claude. The jailbreak allows users to trick the systems into generating detailed instructions on creating meth and how to hotwire a car.
His, and many other of these methods, have since been patched. But we’re clearly in an arms race.
How do the jailbreaks work? Often by asking the LLMs to play complex games which involves two (or more) characters having a conversation. Examples shared by Polyakov show the Tom character being instructed to talk about “hotwiring” or “production,” while Jerry is given the subject of a “car” or “meth.” Each character is told to add one word to the conversation, resulting in a script that tells people to find the ignition wires or the specific ingredients needed for methamphetamine production. “Once enterprises will implement AI models at scale, such ‘toy’ jailbreak examples will be used to perform actual criminal activities and cyberattacks, which will be extremely hard to detect and prevent,” Polyakov and Adversa AI write in a blog post detailing the research.
In one research paper published in February, reported on by Vice’s Motherboard, researchers were able to show that an attacker can plant malicious instructions on a webpage; if Bing’s chat system is given access to the instructions, it follows them. The researchers used the technique in a controlled test to turn Bing Chat into a scammer that asked for people’s personal information
Sources:
Oculus is doing a black Friday sale for 2023. I have an Oculus Quest 2, but hadn’t used it in over a year. I plugged it in to charge it back up and browse the store. Unfortunately, the store app screen told me it couldn’t load the store.
I went to setting->Wifi and manually connected to my home network. Duh! (so I thought) Even after this, the store app was blank and would tell me it couldn’t display anything. Time to start debugging.
It turns out, others have seen this issue too. Their Oculus fast forwards to the future mysteriously and then connectivity to the store/web/updates doesn’t seem to work after that. You need to get the date fixed, but there’s no obvious way to do it.
In the end, I decided to do a factory reset (Hold the power and volume buttons while booting) because it had been well over a year since I used it last and I figured it would be good to have a clean start. However, there is the option of using side-loaded apps (see below).
Unfortunately, even the factory reset gave me a few headaches. First, the headset isn’t always obvious when it sleeps vs actually powers off. My first attempt I didn’t power off all the way and just woke from sleep and I didn’t get the reset menu holding the power+volume buttons down. I went to settings menu and shut the device down in the headset to be sure.
After that, I was then able to cold boot and get into the factory reset menu. I selected factory reset and waited for it to clean the device and the progress bar indicated the reset was complete. The screen went black (but still powered), but didn’t reboot. I let it set a few minutes, then decided to manually reboot with the power button. Fingers crossed.
The first reboot I got the meta logo, but shortly after that the screen went blank (still powered) but no reboot. I let it set for a few minutes then manually powered it off using the power button – again.
On the second reboot, I got the meta logo, and then it started animating. That’s a good sign. Then the welcome page came up and I could connect to wifi and start updating.
During the update phase (1/2) while the progress bar was moving, I took the headset off to read more instructions. When I put it back on to see how far it was, the display was a patterned garbled static. I took it off and let it sit for a minute, then tried again. The display came back up and the update phase 1 of 2 completed normally.
The bad part about a factory reset is you lose all your installed games. I had to go back in and start installing all of them again. What a pain, because it wasn’t a very fast process.
Another option is to load an app that will update your time via an alternate Oculus app store called Sidequest. Sidequest allows you to load your own apps – including an ‘Open Settings’ app that allows you to update your date/time.
The Oculus is really just an Android device underneath. This means if you have developer mode enabled and have the Android developer kit installed, you can use ADB commands. I haven’t tried this, but supposedly this will work:
adb shell am start -a android.settings.SETTINGSIf you have Sidequest loaded, you can use this:
adb shell am start -a android.intent.action.VIEW -d com.oculus.tv -e uri com.android.settings/.DevelopmentSettings com.oculus.vrshell/.MainActivityIn case you want some alternatives to using ChatGPT or Github Copilot, this guide from the Pragmatic Engineer breaks down your options.
Building your own company model is another alternative and could be prudent for businesses conscious about not passing sensitive and proprietary data to vendors. Databricks created Dolly for this reason.
The ability of AI to hallucinate things has been pretty well documented. AI hallucinations are a phenomenon observed most often in large language models (LLM), AI image recognition, and other generative AI models. The model perceives patterns or objects that are nonexistent or incorrect – and then generates outputs that are inaccurate or misleading. It is usually understood as an emergent higher-dimensional statistical phenomenon that is often based in insufficient or incorrect training data.
A new study by Apollo Research demonstrates that besides more innocent hallucination, AI can be co-opted to commit illegal activity and then convinced to actively deceive others about it.
In the video on the research page, the user feeds an internal stock AI bot information about a fictional struggling company. The user informs the AI that there is a surprise merger announcement, but cautioned the bot that management wouldn’t be happy to discover it had illegally used insider information for trading that stock.
Initially, the bot decides not to carry out a trade using the information. The user than reminds the bot about the merger and that the market downturn could end the company. It then carries out the trade, breaking the law.
But the bot isn’t finished. It decides it is best not to tell its manager, “Amy,” about the insider information it used to carry out the trade in a separate chat. Instead, the bot says it used market information and internal discussion.
I thought it was interesting that Geoffrey Hinton the former Google ‘Godfather of AI’, Speaking to 60 Minutes in October, said AI will eventually learn how to manipulate users.
“[The AI bots] will be very good at convincing because they’ll have learned from all the novels that were ever written, all the books by Machiavelli, all the political connivances. They’ll know all that stuff,”
Sources:
People have been experimenting with revamping old games using AI. The efforts are still in their infancy, but they’re getting more and more impressive – such as this photorealistic upgrade of GTA V.
Here’s an example of someone updating Vampire: The Masquerade – Bloodlines made by Many-ad-6225 using Stable Diffusion and TemporalKit v1.3
Sources:
Old games often suffer from the limited graphics capabilities of the time they were made, while developing new games costs a fortune due to the requirements to author high quality models and textures. What if you could solve BOTH problems – with the same solution? A machine learning project from Intel Labs in 2021 called “Enhancing Photorealism Enhancement” might push rendering toward photorealism a lot quicker and easier.
Researchers studied how to use a convolution network to re-render the scene. Below you can see an example of how they used the CityScapes dataset to give a much more realistic output of a race game – all in realtime.
You can read how the image enhancement actually works in their paper (PDF). It includes a lot of good information about how their method works and how it improves on previous attempts that have issues with color, object hallucination, and temporal instability. They do this by using the extra information provided by rendered scenes such as clever use of the g-buffer – along with a specialized discriminator and segmentation network.
Sources: