• Starter AI
  • Posts
  • Introducing: KLING, Behind AI’s reasoning, AI for antibiotic research

Introducing: KLING, Behind AI’s reasoning, AI for antibiotic research

Sora is facing competition.

Hello, Starters!

And we've arrived on another Friday. Before we shut down our devices and head into the weekend, there are still many AI insights to discover. Don't go just yet!

Here’s what you’ll find today:

  • China’s Kuaishou presents KLING

  • Understanding the reasoning behind AI models

  • AI aids in antibiotic research

  • Google harnesses AI to save coral reefs

  • AI predicts cancer treatment outcomes

  • And more.

Sora has yet to be released to the public, and while we wait, other companies are taking advantage of this time to showcase their offerings. This is the case with Kuaishou, a Chinese tech company that recently launched KLING, a video generation model.

According to Kuaishou, KLING can create videos up to two minutes long at 1080p and 30 frames per second, with realistic and consistent results. They have used “transformer diffusion” and a 3D space-time attention system that helps the model achieve accurate motions. These capabilities make it a potential competitor to OpenAI's Sora.

A constantly recurring question in the field of AI is whether language models are capable of rationalising like humans and, if so, how irrational they might be. Do they mimic us, or do they have their own way of being illogical? Researchers have set themselves the task of investigating this by testing seven AI models using cognitive psychology tests.

The answers were separated into four categories: "correct and incorrect" and "human and non-human." This resulted in the revelation that, yes, certain models are irrational; however, they do not follow the same thought patterns as humans. That is, the wrong responses were irrational in their unique way, with GPT-4 leading the way as the most rational and similar to human thinking.

Antibiotic research is one of the many fields set to be transformed by AI. A study published in the journal Cell showcases how scientists harnessed machine learning to predict new antibiotics within the global microbiome. The algorithm was able to explore microbial diversity, discovering nearly one million new molecules.

This process advances research by leaving behind traditional methods such as collecting water and soil samples. Although it has been discussed that AI could be misused to design harmful substances, there can also be a positive outcome if safeguards are applied, leading to treatments and saving lives.

🪸SurfPerch is the result of a joint effort between Google Research, DeepMind, and the public. This new AI tool was developed to understand coral reef ecosystems and their health. It was trained on a dataset of audio recordings that users "curated" via the web by clicking every time they heard a fish sound, resulting in a library that consequently fine-tuned the groundbreaking tool.

🏥Researchers from the National Institutes of Health in the US have introduced "LORIS" in a proof-of-concept study. This AI tool leverages clinical data, such as blood tests, to predict if a patient's cancer will respond to a certain class of immunotherapy drugs, helping doctors find more effective ways to treat the disease according to each patient's needs.

What did you think of today's newsletter?

Login or Subscribe to participate in polls.

Thank you for reading!