April 8, 2023

Model

Meta Released a New AI Model That Can Recognize and Mask Objects in Images and Videos

HYPEBEAST, 04/08/23. Meta released an advanced AI model, called SAM, which can identify objects in an image and create masks for them.

SAM has the largest image annotation dataset to-date, SA-1B. On top of that, it can recognize items it hasn’t encountered before. Furthermore, it can develop masks for any object in any image or video and can segment objects with a click or interactive selection.

SAM is general and can output several valid masks in ambiguity. Meta aims to build a promptable segmentation model and utilize its functions to build out its segmentation dataset at scale. READ MORE

HYPE

Y Combinator’s latest batch sure was a lot of ‘maybe AI can do… this?’

TechCrunch, 04/08/23. YC Demo Days showcased a bunch of AI-focused startups. Many demos struggled with systematic requirements and reliable results. Concerns were expressed about “shovelware”

Relying on AI for business functions may suggest a lack of care for customers or replace human interaction. Some startups pivoted halfway through the program casting doubt on their innovative claims and convictions. READ MORE

PAUSE

Ex-Google CEO Eric Schmidt warns pausing ChatGPT-like AI will benefit China

INTERESTING ENGINEERING, 04/08/23. Eric Schmidt, the former CEO of Google, disagrees with the six-month pause as it would benefit China.

Schmidt suggests discussing appropriate safeguards while acknowledging the understated dangers and unexplained emergent behaviors. Schmidt believes that researchers should release AI algorithms with mitigations for potential negative outcomes. READ MORE

Hysteria

We’re Not Ready for the AI Boom. It’s Coming Anyway.

The Daily Beast, 04/08/23. The surge in large language models (LLMs)has raised concerns about negative consequences. Companies are rushing to cash just like they did with the metaverse, blockchain, and cryptocurrencies. Arguably, they’re doing so without fully considering the consequences.

Many experts have called for a pause on the development of giant AI experiments until their effects and risks are manageable. Educating people on digital and media literacy and designing AI to reduce harm are seen as key to mitigating these risks. READ MORE

DEEPFAKE

Why Pope Francis Is the Star of A.I.-Generated Photos

The New York Times, 04/08/23. AI-generated images of Pope Francis in strange and often humorous scenarios have been circulating wildly on social media. The trend can be attributed to both the pope’s global recognition and approachability as well as the surging interest in new AI tools.

AI-generated images can be dangerous if they are believed to be real. However, many of the images featuring Francis have elicited affection for the pope. This phenomena highlights the increasing role of AI in shaping cultural discourse and generating new forms of humor and satire. READ MORE