Country Star Randy Travis Self-Resurrects via AI, Hallucinations & Knowledge Graphs
Links and news in the AI industry. Tools too.
Morning y’all!
I think I have a bit more than usual on the list of things review this morning! It’s been a busy weekend of AI-related action and there were quite a few interesting things to learn — which I love!
A great way to start the week! Kick ass friends!
※\(^o^)/※
— Summer
NVIDIA keeps on building! Recently they released DrEureka, an AI system using LLM agents to automate the process of training robot skills and then transferring those skills into real world activities.
Country music star Randy Travis hasn’t been able to speak or sing for over a decade and yet he just released his first new song. How? AI models that trained on past recordings to generate a real voice clone.
The comment section is unreal though and people care less about whether it’s AI or not — a really positive thing for the future. Techno-optimism is here to stay.
Mr. Musk has shared his plan for AI + X with a news summarization feature and “stories” which sound a lot like Instagram. Grok will apparently analyze thousands of posts to generate news summaries that update auto-magically.
I didn’t know that Perplexity had a Chrome extension! Let’s go.
A neat look into image feedback loops:
Data artist Robert Hodgin recently created a feedback loop between Midjourney and ChatGPT-4 — he prompted MJ to create an image of an old man in a messy room wearing a VR headset, asked ChatGPT to describe the image, then fed that description back into MJ to generate another image, and did that 10 times.
Huh. The creation of new forms of art is pretty amazing.
RAG won’t solve generative AI’s hallucination problem:
“We’re pretty good at retrieving documents based on keywords, but not so good at retrieving documents based on more abstract concepts, like a proof technique needed to solve a math problem,” Wadden said. “Research is needed to build document representations and search techniques that can identify relevant documents for more abstract generation tasks. I think this is mostly an open question at this point.”
Hallucinations are definitely a growing concern but there are solutions, like knowledge graphs that have been around for a while but now have even more attention:
The internet will never be defeated.
Dr. Brenden Lake of NYU aims to bridge human cognitive studies and AI by recording his daughter's POV to develop a language model, challenging the current AI approach that relies heavily on large datasets. I bet her childhood photos are going to look amazing with that helmet on all the time.
Predictions are coming in that AI will independently design novel drugs.
Teens are spending more time with chatbot friends and that’s concerning people. Oh:
Aaron credits the Psychologist bot with helping him through a rough patch. But the real joy of Character.AI has come from having a safe space where he can joke around or experiment without feeling judged. He believes it’s something most teenagers would benefit from. “If everyone could learn that it’s okay to express what you feel,” Aaron said, “then I think teens wouldn’t be so depressed.”
“I definitely prefer talking with people in real life, though,” he added.
Real people are still useful. That’s good.
Artists are made at Google’s “massive” copyright infringement with their new tool Imagen AI. Probably every artist will eventually cave and realize that their work in these new models is a good thing.
More and more money is pouring into AI spend, 293% to be exact (via Ramp). It’s always a good idea to follow the money.
A free webinar. Why not.
Secret Llama is a fully private chatbot in your browser! Never leaves your computer and you can work offline.
And here are a few additional links:
Korean woman lost $50k to Elon Musk deepfake.
AI scamming is going to grow says Warren Buffett.
Tutorials made easy.
Secure AI automation.
AI-powered recruiting and sourcing.
※\(^o^)/※
— Summer