AI Advancements in Fun and Hope

👋 Welcome back! In this edition, there have been some fascinating developments in the world of AI. A new generative image AI and two groundbreaking healthcare projects are among them. Let’s dive in!

Stable Diffusion XL (SDXL)

Stability AI, a generative AI company, has just released a new version of their Stable Diffusion XL model. They are focused on generating photorealistic images. Honestly, their promo image sums it up: “Pretty Good”. At a quick glance, the images are impressive, but if you pause for a second, you will notice weird stuff, like that car is definitely fake.

Google’s Med-PaLM 2

Google’s AI - Med-PaLM 2 has improved a great deal in a year. It is a medical-focused AI. Last year, it was tested on U.S. medical licensing-style questions and scored a passing grade of 60%. This year, it scored 85% when tested. I feel hopeful when I hear that an AI can speed up cancer treatment planning, analyze ultrasounds to screen for breast cancer, and reduce the cost of tuberculosis screening around the world.

Sybil - AI early lung cancer detection

Another Medical-related AI can potentially detect lung cancer in patients years before a doctor. Scientists at the Mass General Cancer Center and MIT tested their AI to predict whether a person will develop lung cancer in the next year. It was correct 86% to 94% of the time. Currently, Sybil is not approved by the FDA, but there are 300 AI tools that are approved by the FDA. And these tools have been helping doctors and scientists diagnose for some time.

🌎 Around the Web

This week around the web, we have animated drawings, bug squashing, open-source models, AI acting like humans, and text-to-video.

• Meta: Bring your animated drawings to life

• OpenAI’s bug bounty program | Bugcrowd

• Dolly 2.0: The World's First Open Instruction-Tuned LLM

• Stanford: AI Study Simulates Authentic Human Behavior

• Runway: Text-to-video

🛠️ The Best Open Source Tools

AgentGPT is an autonomous AI that can take in a goal and work towards that goal in multiple steps without the need for a human to continue to prompt it.

🎧 Listen Up!

Great general tech news show that talks about all the big things happening each day.

Thanks for reading this issue of The Breakdown. If you haven’t already, subscribe to get the next issue in your inbox.

Until next time, take it easy 👋
— Jordan Wade