Artificial Intelligence models that generate entirely new content are creating a world of opportunities for entrepreneurs. And engineers are learning to do more with less.
Those were some takeaways from a panel discussion at the Intelligent Applications Summit hosted by Madrona Venture Group in Seattle this week.
“Big data is not a priority anymore, in my opinion,” said Stanford computer science professor Carlos Guestrin. “You can solve complex problems with little data.”
Engineers are more focused on fine tuning off-the-shelf models, said Guestrin, co-founder of Seattle machine learning startup Turi, which was acquired by Apple in 2016. New “foundation” AI models like DALL-E and GPT-3 can hallucinate images or text from initial prompts. Read More
Monthly Archives: November 2022
AI Drew This Gorgeous Comic Series, But You’d Never Know It
The Bestiary Chronicles is both a modern fable on the rise of artificial intelligence and a demonstration of how shockingly fast AI is evolving.
You might expect a comic book series featuring art generated entirely by artificial intelligence technology to be full of surreal images that have you tilting your head trying to grasp what kind of sense-shifting madness you’re looking at.
Not so with the images in The Bestiary Chronicles, a free, three-part comics series from Campfire Entertainment, an award-winning New York-based production house focused on creative storytelling. Read More
Artificial intelligence means anyone can cast Hollywood stars in their own films
Free AI software is primed to strip away the control of studios and actors who appears in films
For years, the only way to create a blockbuster film featuring a Hollywood star and dazzling special effects was at a major studio. The Hollywood giants were the ones that could afford to pay celebrities millions of dollars and license sophisticated software to produce elaborate, special effects-laden films. That’s all about to change, and the public is getting a preview thanks to artificial intelligence (AI) tools like OpenAI’s DALL-E and Midjourney.
Both tools use images scraped from the internet and select datasets like LAION to train their AI models to reconstruct similar yet wholly original imagery using text prompts. The AI images, which vary from photographic realism to mimicking the styles of famous artists, can be generated in as little as 20 to 30 seconds, often producing results that would take a human hours to produce. Read More
New Tool Allows Users to See Bias in AI Image Generators
A new tool is allowing people to see how certain word combinations produce biased results in artificial intelligence (AI) text-to-image generators.
Hosted on Hugging Face, the “Stable Diffusion Bias Explorer” was launched in late October.
According to Motherboard, the simple tool lets users combine descriptive terms and see firsthand how the AI model maps them to racial and gender stereotypes. Read More
Training Compute-Optimal Large Language Models
We investigate the optimal model size and number of tokens for training a transformer language model under a given compute budget. We find that current large language models are significantly undertrained, a consequence of the recent focus on scaling language models whilst keeping the amount of training data constant. By training over 400 language models ranging from 70 million to over 16 billion parameters on 5 to 500 billion tokens, we find that for compute-optimal training, the model size and the number of training tokens should be scaled equally: for every doubling of model size the number of training tokens should also be doubled. We test this hypothesis by training a predicted compute-optimal model, Chinchilla, that uses the same compute budget as Gopher but with 70B parameters and 4× more more data. Chinchilla uniformly and significantly outperforms Gopher (280B), GPT-3 (175B), Jurassic-1 (178B), and Megatron-Turing NLG (530B) on a large range of downstream evaluation tasks. This also means that Chinchilla uses substantially less compute for fine-tuning and inference, greatly facilitating downstream usage. As a highlight, Chinchilla reaches a state-of-the-art average accuracy of 67.5% on the MMLU benchmark, greater than a 7% improvement over Gopher. Read More
#performanceGoogle and Renault are creating a ‘software-defined vehicle’
The new vehicle will feature the best of Google’s Cloud technology with Renault’s car expertise
In 2018, Google partnered with Renault Group to bring Android-powered infotainment systems to cars. The two companies are now building on this partnership to design and deliver the digital architecture for a more complex concept, a “software-defined vehicle” (SDV).
This partnership has a twofold goal of creating both in-vehicle software and cloud software to enable the SDV platform and a Digital Twin, according to the release. Read More
SetFit outperforms GPT-3 while being 1600x smaller
Everyone is very familiar with the current hype around Large Language Models (LLM) such as GPT-3 and Image Generation models such as DALL-E 2 and Stable diffusion. However, the results of these models come at a price.
- GPT-3: ~$12 Million
- DALL-E: ~$500k- $1 million
- Stable Diffusion: ~$600k
#nlp
Elon Musk Has Fired Twitter’s ‘Ethical AI’ Team
NOT LONG AFTER Elon Musk announced plans to acquire Twitter last March, he mused about open sourcing “the algorithm” that determines how tweets are surfaced in user feeds so that it could be inspected for bias.
His fans—as well as those who believe the social media platform harbors a left-wing bias—were delighted.
But today, as part of an aggressive plan to trim costs that involves firing thousands of Twitter employees, Musk’s management team cut a team of artificial intelligence researchers who were working toward making Twitter’s algorithms more transparent and fair. Read More
Generative AI and Film Future(s)
I’ve been working in some aspect of the film business for far too long, but what brought me into it was my interest in where the future of art was going as the moving image blended with computers the web and new technologies. And that’s what’s been fascinating me more often as of late than anything in the traditional film world. What’s been happening in the past few months, weeks even, in AI and generative art, and how it overlaps with traditional arts and film in particular, has been pretty incredible to watch. I’ve been too busy in this older (dying, crumbling?) film world to participate in it directly – I haven’t taken the time to learn Midjourney or use Dall-E. And while I’ve been following what people are doing with virtual production or other technologies which will soon merge into this space, I haven’t had a chance to play around with them. Heck, I don’t even own a VR headset, and can’t barely bother to use Facebook, much less get into Mark’s version of the metaverse. But all these spaces, combined, consume my thoughts when I’m not on some Zoom with a client, or busy trying to help bring a little indie film to reality (as I’ve been doing lately, but that’s another post).
In brief, that’s because I got into all of this as a student of Greg Ulmer at the University of Florida, who was a theorist who proposed the idea of society moving from orality to literacy (Walter Ong) to what he called Electracy, where society learns the full communicative potential of these technologies, much as literacy is to reading. I’ve written a fair bit about how this will impact the arts and film (here’s a post from 2011 about it, which was part of a chapter I wrote for a book), but you can see it all coming together now.
The latest craze – in all senses of the word because it’s also driving many artists crazy mad – is generative AI and art, and while it’s hitting graphic arts and photography/still images hardest now, it’s already becoming a phenomenon in film and video, too. Read More
DALL·E API NowAvailable in Public Beta
Starting today, developers can begin building apps with the DALL·E API.
Developers can now integrate DALL·E directly into their apps and products through our API. More than 3 million people are already using DALL·E to extend their creativity and speed up their workflows, generating over 4 million images a day. Developers can start building with this same technology in a matter of minutes. Read More
Rick's Cafe AI 