Pigeon-toed posing

I've been noticing this for a while. I figured it was deliberate, and sure enough, it's no coincidence.

It helps women look more ridiculous and less original. Very effective.

https://www.dailymail.co.uk/femail/article-3872996/Why-posing-pigeon-toed-suddenly-vogue-listers-turn-knees-look-younger-thinner-child-like-stance-taking-red-carpet.html 

https://x.com/thismorning/status/791369567280861184

Researchers Claim New Technique Slashes AI Energy Use By 95%

Researchers Claim New Technique Slashes AI Energy Use By 95% (decrypt.co)49

Posted by BeauHD on Tuesday October 08, 2024 @11:30PM from the would-you-look-at-that dept.

Researchers at BitEnergy AI, Inc. have developed Linear-Complexity Multiplication (L-Mul), a technique that reduces AI model power consumption by up to 95% by replacing energy-intensive floating-point multiplications with simpler integer additions. This method promises significant energy savings without compromising accuracy, but it requires specialized hardware to fully realize its benefits. Decrypt reports:L-Mul tackles the AI energy problem head-on by reimagining how AI models handle calculations. Instead of complex floating-point multiplications, L-Mul approximates these operations using integer additions. So, for example, instead of multiplying 123.45 by 67.89, L-Mul breaks it down into smaller, easier steps using addition. This makes the calculations faster and uses less energy, while still maintaining accuracy. The results seem promising. "Applying the L-Mul operation in tensor processing hardware can potentially reduce 95% energy cost by element wise floating point tensor multiplications and 80% energy cost of dot products," the researchers claim. Without getting overly complicated, what that means is simply this: If a model used this technique, it would require 95% less energy to think, and 80% less energy to come up with new ideas, according to this research.

The algorithm's impact extends beyond energy savings. L-Mul outperforms current 8-bit standards in some cases, achieving higher precision while using significantly less bit-level computation. Tests across natural language processing, vision tasks, and symbolic reasoning showed an average performance drop of just 0.07% -- a negligible tradeoff for the potential energy savings. Transformer-based models, the backbone of large language models like GPT, could benefit greatly from L-Mul. The algorithm seamlessly integrates into the attention mechanism, a computationally intensive part of these models. Tests on popular models such as Llama, Mistral, and Gemma even revealed some accuracy gain on certain vision tasks.

At an operational level, L-Mul's advantages become even clearer. The research shows that multiplying two float8 numbers (the way AI models would operate today) requires 325 operations, while L-Mul uses only 157 -- less than half. "To summarize the error and complexity analysis, L-Mul is both more efficient and more accurate than fp8 multiplication," the study concludes. But nothing is perfect and this technique has a major achilles heel: It requires a special type of hardware, so the current hardware isn't optimized to take full advantage of it. Plans for specialized hardware that natively supports L-Mul calculations may be already in motion. "To unlock the full potential of our proposed method, we will implement the L-Mul and L-Matmul kernel algorithms on hardware level and develop programming APIs for high-level model design," the researchers say. 

Fly brain mapped

https://flywire.ai/ 

https://www.bbc.com/news/articles/c0lw0nxw71po

https://www.nature.com/articles/s41586-024-07686-5

Ocean plastic propaganda

I first encountered a trend of young people obsessed with plastic polution, and having to tolerate paper straws which suck. I thought there must be some massive propaganda campaign at work.  

I ran across this video, so I thought I'd better archive it.

https://www.youtube.com/watch?v=IglBJ62Sv3Q

Trump is a peace-loving hippie

Just the shirt I was looking for on the heels of fresh negotiations to end the war in Ukraine ahead of the election.

Take that, you war-mongering lefties!

https://www.etsy.com/listing/1728972221/comfort-colors-trump-2024-tshirt-hippie

https://allbluetees.com/product/funny-hippie-trump-meme-hawaiian-shirt/

https://www.amazon.com/Trump-Groovy-Hippie-Sleeve-T-Shirt/dp/B0D9155CSJ

Meta glasses AI real-time video

 Good progress, but disappointed the translation isn't real-time. 

The only difference between this and Google Translate is that it's hands-free and visual. However, we still have the translation lag. Hopefully, this is obvious enough that real-time is on radar.

https://www.youtube.com/watch?v=l_QruJ0Kv9U

https://www.youtube.com/watch?v=AA7ZnWOHs8I

Ray-Ban Smart Glasses Updated With Real-Time AI Video, Reminders, and QR Code Scanning (techcrunch.com)14

Posted by BeauHD on Wednesday September 25, 2024 @06:00PM from the new-and-improved dept.

An anonymous reader quotes a report from TechCrunch:Meta CEO Mark Zuckerberg announced updates to the company's Ray-Ban Meta smart glasses at Meta Connect 2024 on Wednesday. [...] Meta says its smart glasses will soon have real-time AI video capabilities, meaning you can ask the Ray Ban Meta glasses questions about what you're seeing in front of you, and Meta AI will verbally answer you in real time. Currently, the Ray-Ban Meta glasses can only take a picture and describe that to you or answer questions about it, but the video upgrade should make the experience more natural, in theory at least. These multimodal features are slated to come later this year. In a demo, users could ask Ray-Ban Meta questions about a meal they were cooking, or city scenes taking place in front of them. The real-time video capabilities mean that Meta's AI should be able to process live action and respond in an audible way. This is easier said than done, however, and we'll have to see how fast and seamless the feature is in practice. We've seen demonstrations of these real-time AI video capabilities from Google and OpenAI, but Meta would be the first to launch such features in a consumer product.

Zuckerberg also announced live language translation for Ray-Ban Meta. English speaking users can talk to someone speaking French, Italian, or Spanish, and their Ray-Ban Meta glasses should be able to translate what the other person is saying into their language of choice. Meta says this feature is coming later this year and will include more language later on. The Ray-Ban Meta glasses are getting reminders, which will allow people to ask Meta AI to remind them about things they look at through the smart glasses. In a demo, a user asked their Ray-Ban Meta glasses to remember a jacket they were looking at, so they could share the image with a friend later on. Meta announced that integrations with Amazon Music, Audible, and iHeart are coming to its smart glasses. This should make it easier for people to listen to music on their streaming service of choice using the glasses' built-in speakers. The Ray-Ban Meta glasses will also gain the ability to scan QR codes or phone numbers from the glasses. Users can ask the glasses to scan something, and the QR code will immediately open on the person's phone with no further action required.Zuckerberg also unveiled the company's prototype AR glasses codenamed Orion, which feature a 70-degree field of view, Micro LED projectors, and silicon carbide lenses that beam graphics directly into the wearer's eyes.

Sam Altman predicts superintelligence in a few thousand days (2.7 years=1000 days)

This doesn't explain why Altman feels the need to speak in Vulcan vernacular. 

Sam, would you mind using ChatGPT to let us know in plain language?

Altman: "Superintelligence will occur in a few thousand days."

Or for those of us who are not yet cyborgs, 2.7 years equals 1000 days, and if "a few" suggests let's say, "three", then he thinks we'll reach superintelligence within ten years. 

Ergo, will we reach AGI a.k.a "singularity" when? Or are we already there? Sam?

https://venturebeat.com/ai/openai-ceo-sam-altman-anticipates-superintelligence-soon-defends-ai-in-rare-personal-blog-post/