AI Top-of-Mind for 8.20.24 — Grok 2
Today: Grok 2, Midjourney update, Fable’s Showrunner, how to prove you are a human, Walmart’s Gen AI successes, and Trump v Taylor Swift.
Top-of-mind is Grok 2, the latest X.ai model. Andrew Best writing in ‘Artificial Intelligence in Plain English’ provides some perspective. Remember that Grok is from Musk’s X.ai, and was secretly released a week back as ‘sus-column-R.’ Many, including me, thought this was a preview of a new model from OpenAI, but it turned out not to be the case. Importantly, Grok 2 is (mostly) uncensored:
· Elon Musk wants the AI to say whatever is true.
· These other companies want to avoid offending people.
· This means that Grok 2 will have free reign to use the full power of its own AI to answer in a way that it wants to.
· The other LLM’s are all programmed to make sure that their answers are minimally offensive.
From Grok 2:
Also on the model front, Walid Amamou in ‘UBIAI NLP’ offers a tutorial on integrating knowledge graphs with RAG. From his post:
· Knowledge Graphs provide a structured representation of information, capturing relationships and entities in a format that is both human-readable and machine-processable. The RAG Stack, on the other hand, enhances generative models by incorporating information retrieval processes, resulting in more accurate and contextually relevant responses.
· This article explores the components and benefits of Knowledge Graphs, explains the workings of the RAG Stack, and offers a detailed guide on how to integrate these technologies to create a more powerful AI system.
And, a new ‘Midjourney’ release as reported by ‘VentureBeat.’ The tool now offers a new UI that integrates many existing capabilities including inpainting and outpainting. From the intro video:
Do you want to create your own take on your favorite TV show? ‘The Information’ reports on a new offer from ‘Fable Studio’ delivering interactive storytelling fueled by generative artificial intelligence. From the article:
Three months ago, Fable launched an early version of Showrunner — a platform that can write, voice and animate episodes of shows. Fable makes and licenses animated videos and then allows others to use AI to create their own versions of the shows. Think of a fan of “The Simpsons” making episodes about characters such as the bartender Mo or the convenience store owner Apu.
Now looking at the bigger picture, are we at the point that we have to prove that we’re flesh and blood? Maria Mouschoutzi writing in ‘AI Advances’ offers some advice. From her post:
In the last two years of rapid AI growth, we’ve been increasingly exposed to more AI-generated content, and we’re gradually learning to recognize it to some extent. For instance, using words like “delve,” “embark,” “tapestry,” or “vibrant” in your writing immediately triggers AI alarms in our brains. Just a few years ago, you might have been impressed by the quality of writing and the rich vocabulary, but now these words can make a piece of writing feel suspiciously artificial.
And a relevant point on creativity that we should all take to heart:
To create digital art, one needs to know how to draw and paint traditionally on canvas first, then acquire the necessary tools for digital drawing, learn how to use them, and finally, spend countless hours perfecting even the tiniest details of a digital piece, and all these just to produce a single digital illustration.
Turning to retail, ‘ModernRetail’ reports on Walmart’s successes with Gen AI.
· Walmart president and CEO Doug McMillon told investors during its fiscal second-quarter earnings call on Aug. 15 that the company has used multiple large language models to create or improve more than 850 million pieces of data in its product catalog.
· “We’re finding tangible ways to leverage generative AI to improve the customer, member and associate experience. We’re leveraging data and large language models from others and building our own,” McMillon said. “Without the use of generative AI, this work would have required nearly 100 times the current headcount to complete in the same amount of time.”
And:
Gildenberg (founder of commerce consultancy Confluence Commerce) said that like people, AI is most effective when given a well-defined job, taught what to do, with clear expectations and patience to learn. “What AI learns really well is a specific task, and if you focus AI on a very specific task, it learns really quickly,” he said, adding that Walmart automating the catalog workflow — a more simple task — is “the kind of application for which AI can be game-changing.”
Does she or doesn’t she? More deepfakes, this time the Trump campaign posting images inferring a Taylor Swift endorsement. Not very probable! From the ‘CNN’ reporting:
· Trump posted “I accept!” on his Truth Social account, along with a carousel of (Swift) images — at least some of which appear to be AI-generated.
· One of the AI-manipulated photos depicts Swift as Uncle Sam with the text, “Taylor wants you to vote for Donald Trump.” The other photos depict fans of Swift wearing “Swifties for Trump” T-shirts.
· In the new book, “Apprentice in Wonderland,” Trump told author Ramin Setoodeh, “I think she’s beautiful — very beautiful! I find her very beautiful. I think she’s liberal. She probably doesn’t like Trump.”
