“We can’t let the Chinese beat us in wasteful spending on unneeded AI data centers!”
The free DeepFake-o-Meter from the University at Buffalo Media Forensics Lab. You do need to register to use it.
https://zinc.cse.buffalo.edu/ubmdfl/deep-o-meter/landing_page
The “DeepFake-o-Meter” is a free, open-source, web-based platform developed by the UB Media Forensics Lab at the University at Buffalo (SUNY). Its purpose is to detect and analyze media (images, videos, and audio) to determine the likelihood of it being a “deepfake” – media that has been manipulated or generated using artificial intelligence techniques.
• Deepfake Detection: The platform integrates multiple state-of-the-art deepfake detection algorithms. This allows for a more comprehensive analysis of uploaded media.
• Transparency: Unlike some commercial tools, the DeepFake-o-Meter emphasizes transparency. It often discloses the algorithms used in the analysis, providing users with more insight into how the conclusion was reached.
• Multiple Algorithm Analysis: Users can often choose to run their media through multiple detection algorithms, each potentially providing a different perspective and a percentage likelihood of the content being AI-generated.
• Open Source: Being open source means that the underlying code is accessible to the public, promoting scrutiny and further development by the research community.
• User-Friendly: It is designed to be a user-friendly online platform, making deepfake detection accessible to individuals who may not have specialized technical knowledge. Users typically upload media and receive an analysis result.
• Research and Benchmarking: Beyond being a tool for the general public, it also serves as an evaluation and benchmarking platform for researchers in digital media forensics to compare the performance of different detection algorithms.
• Continuous Updates: The DeepFake-o-Meter is intended to be updated with new algorithms as the field of deepfake detection advances.
Grave of the Fireflies is really different because is Takahata’s work, not Miyazaki’s.
From Takahata’s repertoire I really recommend “The war of the Pompokos”, a pungent satire on Japan’s hyperdevelopism -and how they were basically eating their future- of the late 80’s to early 90’s. You don’t need to know about the context to enjoy it, though.
When I first saw that on a bootleg tape I remember thinking that there was no way it would ever get distributed by a major studio in the US due to the unusual focus on magical shape-shifting raccoon dog scrotums, but I was wrong! Disney went ahead and distributed it along with other Studio Ghibli films.
They didn’t edit out the scrotums (which would have been a monumental effort at the time) but apparently in the English language dub they refer to them as “pouches.”
Would have been very difficult to edit this out 
Though I guess they could have edited these out:
Pretty sure the contract was shielded against that kind of heavy handed boulderization, though.
They look heavily boulderized to me, those are some damn big stones. 
Sorry, I just found out is written “bowdlerize”. Not changing the post for obvious reasons.
Easiest way to see that around here in the 00s was to buy the Hong Kong release. Those were great they had English subs and extra disc with all kinds of stuff like the full movie as story boards.
I had tried watching it in college with a bunch of friends and the collective opinion at the time was that the movie was way too long and not as focused as it could’ve been. We never finished it, we got maybe 2/3rds of the way through and then just lost interest.
I partially agree it has a bit of an awkward rhythm but I still have a soft spot for it.
Here’s something I’ve found myself actually using LLMs for with some degree of success: if I am doing a known-item search for something I don’t know the name of, but I can’t figure out the right search terms to surface it, I ask ChatGPT and it usually works:
There was an episode of the show How It’s Made that included a segment on the making of “sampuru” or other non-edible food for display. What episode was that?
ChatGPT’s response:
The episode of How It’s Made that features a segment on the creation of artificial food displays, commonly known as “sampuru,” is Season 15, Episode 7 (Episode 189), titled “Replica Foods.” This episode originally aired on May 7, 2010. Wikipedia
Episode Details
- Season: 15 Prime Video+3Arigato Travel+3ComingSoon+3
- Episode: 7 (Episode 189 overall)
- Title: "Replica Foods" Wikipedia
- Original Air Date: May 7, 2010 Wikipedia
- Other Segments in the Episode:
- Traffic Cone Dispensers Wikipedia
- Rocking Horses Wikipedia+1TVmaze+1
- London Taxis Ranker+12Wikipedia+12TVmaze+12
In the “Replica Foods” segment, the show explores the intricate process of crafting hyper-realistic food models used in restaurant displays, particularly prevalent in Japan. These models are meticulously made using materials like wax and plastic to replicate the appearance of actual dishes, serving as visual menus for customers.
If you’re interested in watching this episode, it may be available on streaming platforms that host How It’s Made, such as Discovery+, Max, or Prime Video. Availability can vary based on your region and subscription.
Yeah, this and structuring unstructured data is what they are genuinely useful for. For example, I had a list of maybe 30 Latin species names as a comma separated list, copied from a text, so not a CSV, and with superfluous words strewn in. I asked Chat GPT to return them as a table and to add a column with their common English names. It took a few seconds, plus some time for double-checking it hadn’t hallucinated anything.
I could have done what I would have done before: turn it into a table by a combination of Regex and manual intervention. Then find an API of species names and use it to populated that second column. I actually went down that route at the beginning and couldn’t find a publicly accessible API for that, which means I would have had to either futz around with WikiData and SPARQL or do it by hand.
The ChatGPT solution was much more efficient, although I also worry that means those other data cleaning skills of mine will atrophy
“You can call me AI A1.”
