AI Blog Why feelings matter?

Why feelings matter?

AI Blog

Issue #31

01/11/2023

Pegasus will watch YouTube for you

Twelve Labs (not to be confused with Eleven Labs, different companies) released Pegasus-1, an AI model that understands video! Basically Pegasus-1 can “see” a video and tell you what happened or ask questions about it, such as what props were used. This unlocks many possibilities, such as:

Improved Content Moderation: automatically flag videos that violate the community guidelines based on data shown in the footage
Product tagging: automatically tag products in promo videos

For those interested, here is the technical report.

Fun Fact: Eleven Labs and Twelve Labs have hosted a hackathon together, which was called 23Labs.

Why feelings matter?

LAION (non-profit organization for AI research) launched Open Empathic project with the goal of developing emotion-detecting AI, so that open-source AI systems can have empathy and emotional intelligence.

More specifically, the project is about training AI models in audio and video to understand and respond to human emotions in an appropriate way.

Why is this relevant to us?

As AI is increasingly integrated into our daily lives, it is important that the interactions we have with it prioritize our well-being. Some examples where adding emotional intelligence to AI will be particularly helpful:

Healthcare: patient support
Customer Service: better customer service, especially when there are frustrations on the customer side
Education: adapting the educational process based on the emotional needs of the learner and their personalit
Robotics: more “natural” interactions with humans
Virtual Assistants: assistants such as Alexa will better understand users’ emotions

Resources

AI Insider 📰

Mistral or Europe’s OpenAI, as many call it, is going into its second round of funding, just 4 months after the previous one, where it raised $113M. Its AI model is open-source and here you can see more.
ChatGPT mobile app revenue is steadily increasing, with 38M active users in September. Here is an interesting breakdown of the App Store market for AI apps.
Google Maps has added new AI tools to aid navigation and discovery, including immersive route previews, augmented reality search, speed limit information and thematic recommendations.
Meta’s Chief AI Scientist made some very interesting statements about AI legislation. It’s a bit of a long tweet, but we recommend reading it in full.
NVIDIA is testing the inclusion of AI chatbots in the chip design diatribe.
Google invests again in Anthropic, this time its investment is $2B, so combined with the previous one we reach a total of $2.55B.
Shutterstock now allows you to edit your available photos with AI tools.

Learning Bytes 🧐

President Biden issued a Presidential Executive Order on Secure, Reliable and Confidential AI to balance national security with consumer rights – creating some of the first international protections that could be bolstered by global legislation.
- The Order calls for comprehensive actions in several areas, from requiring rigorous security testing for robust AI systems to establishing an advanced cybersecurity program
- In particular, it requires AI developers to share safety results, and the National Institute of Standards and Technology to create standards to ensure the safety and reliability of AI tools before public release.
- In addition, it stresses the importance of protecting Americans’ privacy and supporting consumers by requiring the Department of Commerce to label all AI-generated content to protect the mental health of U.S. consumers.
  Also, ai.gov, a website was created that showcases the efforts and achievements of the US government on AI. Its advertisements for seeking AI talent are also uploaded there.
Microsoft Research has partnered with teachers in India to design an AI teaching assistant, Shiksha Copilot, that aims to help overworked teachers by designing personalised lessons.
Microsoft researchers have published a paper on code diffusion models, in which they report the training parameters of GPT-3.5-turbo to be at 20B . We don’t know if it is a leak, a typo or a deliberate announcement, but the paper was withdrawn for a few hours.
The G7 countries agreed on a code of conduct to be followed by AI companies.
Three artists in the US have legally sued AI art generators for using copyrighted images without permission to train AI models. However, Federal Judge dismissed most of the claims against Midjourney and DeviantArt, calling them “problematic in many aspects”.
The central issues have to do with whether the AI systems store copies of copyrighted images and whether their creators can claim infringement if no identical material was produced by the AI tools.