Daniel Holden from Ubisoft gave this great talk at GDC 2018 on how data-driven analysis of their character animation control system turned into a AI system that vastly reduced the complexity and manpower involved in building an animation system for character control.
The AI Black Mirror YouTube channel has some real nightmare fuel videos. The All Purpose Everything Sauce ad demonstrates what a hard time AI has with understanding how people eat.
McDonald’s is ending a test run of AI drive-thrus. It paired with IBM and put it in more than 100 restaurants since 2021. The goal was to simplify voice-activated ordering. It certainly couldn’t have anything to do with replacing workers with automation given the new $20/hour minimum wage for fast food restaurants (CA is considering another 3.5% increase in 2025).
Two sources familiar with the technology told CNBC that among its challenges, it had issues interpreting different accents and dialects, which affected order accuracy. McDonald’s will keep using IBM’s other solutions, but AI ordering seems to be on hold.
Reminds me of some issues iPhone has in Scotland.
This could make a fun little demo – flying through a cityscape with buildings that are constantly generated by AI – getting funkier and funkier as you go along
Curious Refuge is holding their 2nd annual AI Horror Film Competition. Using the Infinity V2 model, make a movie and submit it by Oct 21st to win up to $3000.
You can check out the entries here.
Here was one of last year’s winners. It’s probably not going to replace real moviemakers yet, but it’s absolutely a great creative tool to try out ideas and visualize scenes.
Prompt engineering is a bit of a dark art. Crafting prompts with positive and negative keywords requires finesse and creativity.
Geeky Gadgets has a selection of helpful prompt keywords that can help you create interesting lighting/exposure effects, texture/surface effects, color/tone, composition and perspectives, motion, environmental effects and artistic styles/techniques.
Article:
Being able to create consistent characters with an AI art generators is not easy. AI has a habit of generating completely different looking characters between each prompt.
One trick is to use variables in Dalle-3. These variables enable the establishment of specific character traits that persist across generations.
Articles:
“It’s a security risk not to have it. At this point, we have to have it [AI flight capabilities]”
Back in 2016, I wrote about how a lone grad student created a fighter pilot AI that was able to defeat a retired U.S. Air Force Colonel every engagement. Back in April 2024, I wrote about how a new project put AI into an real F16 called X-62A, or codenamed Vista. At the time, they didn’t reveal the results of those tests other than saying they ‘went well’.
Well, now we know what happened. First off, when the VISTA AI test aircraft flew against humans in a dogfight, it was able to do so with no safety rule violations logged. During simulation tests named AlphaDogfight, kicked off in 2020, AI agents developed by Heron Systems beat an experienced Air Force F-16 pilot in simulated dogfights 5-0. Essentially duplicating what the grad student was able to achieve in 2016 simulations.
In further live tests, Air Force Secretary Frank Kendall said, “We were up against a pilot who had two or three thousand hours of experience. He was very good. It was roughly an even fight. But against a less experienced pilot the AI … [and] the automation would have performed better,”
We also learned that in this most recent experiment with a real plane, the re-enforcement learning based AI pilot wasn’t trained to fly from point A to point B or do any kind of normal ‘civilian’ flying. Instead, it was meant to figure out a specific dogfighting problem:
“We did about 10 or a dozen different situations where … I was in the front seat and I had a button on my stick where basically I initiated the automation,” Air Force Secretary Frank Kendall, the real pilot in the AI plane, said at an AI expo hosted by the Special Competitive Studies Project.
“It was meant to do something much more sophisticated [than just fly the plane], but limited in that we had to reason over a highly dynamic situation. So, where is the adversary, what is the adversary doing, how do I put my airplane in position to solve for both safety and weapons effectiveness?”
This distinction of using AI to help solve getting the plane in position against an opponent leaves the critical decision to fire up to the human; while still solving one of the most critical problems of air combat – which is out-foxing your opponent to get a weapons lock on them.
At the end of the hourlong flight, Kendall climbed out of the cockpit grinning. He said he’d seen enough during his flight that he’d trust this still-learning AI with the ability to decide whether or not to launch weapons in war.
Articles:
Apple has halted its long-rumored “Project Titan” that was developing an electric car (Bloomberg). The company reportedly announced the news internally on Tuesday and said many people in the 2,000-person team behind the car will shift to other AI efforts instead.
Despite hiring industry leaders like Tesla’s former Autopilot director, the project has been plagued by high turnover (including chief Doug Field in 2021), constantly changing plans, and internal skepticism. More recent rumors suggest that the $100,000 car would likely not have self-driving capabilities.
Given this, it likely makes sense apple is re-directing the efforts of the automotive team to more established AI plans.
Articles:
Midjourney is the pre-eminent image generator; but used to require submitting prompts in a fiddly way through Discord and then paying for anything beyond a few free ones. Last year they made a more intuitive website; but you had to have a special account.
They’ve now opened the Midjourney website to everyone – but you can only generate 25 images and then need to buy a paid plan. You’ll need to log in with your connected Discord account, or use your Google account. Apparently it even will show your history of previously generated images to browse when properly connected.
Articles: