- The Daily Update
- Posts
- AI detects breast cancer
AI detects breast cancer
News: AI-generated explicit images raise concerns again
Jack Shields
October 25, 2023
Welcome back!
Hope you’ve had a great Wednesday. Unfortunately, we’ve got more disturbing news about AI-generated sexual images on today’s menu.
We’ll also take a look at Google’s leaked AI tools and new research in medical AI. Let’s dive in.
In today’s Daily Update:
🗞️ AI-generated child sexual abuse images threaten to flood the internet
🤖 AI predicts one-third of breast cancer cases prior to diagnosis
📸 Google’s secret AI project leaked
🚨 AI Roundup: Four quick hits
Read time: 2 minutes
LATEST NEWS
🗞️ AI-generated child sexual abuse images threaten to flood the internet
Source: Adobe Firefly
Yesterday the U.K.-based Internet Watch Foundation warned that AI-generated child sexual abuse images could flood the internet if controls are not put on AI tools that generate deepfake photos.
What you need to know:
The watchdog agency urged governments and technology providers to act quickly before a flood of images overwhelms law enforcement and victims.
Last month a South Korean man was sentenced to 2 ½ years in prison for using AI to create 360 child abuse images.
In some cases, kids are using these tools against each other, such as in southwestern Spain where police have been investigating teens’ use of an AI app to create nude photos of schoolmates.
If it goes unchecked, the flood of deepfake child sexual abuse images could slow down investigators trying to rescue children who turn out to be virtual characters. The images could also be used to groom and coerce new victims.
The dark side: IWF analysts discovered a network of abusers sharing tips and marveling at how easy it is to generate sexually explicit images of children of all ages. Open-source tools like Stable Diffusion are go-to choices for producers of this content. Tech providers urgently need to implement guardrails that prevent the generation of explicit images.
RESEARCH INSIGHT
🤖 AI predicts one-third of breast cancer cases prior to diagnosis
Source: Adobe Firefly
AI software can predict one-third of breast cancer cases up to two years prior to diagnosis according to a new study published by Radiology.
The details:
Researchers evaluated the use of Transpara version 1.7.0 for mammography exams conducted prior to diagnosis in 1,016 patients with screen-detected breast cancer and 586 patients with interval breast cancer.
They used the AI to assign high-risk, intermediate risk and low risk scores for potential malignancy.
38.3% of screen-detected breast cancer cases were identified as high-risk up to two years prior to diagnosis, while 39.4% of interval cases were labeled as high-risk in the same time frame.
23% of screen-detected cases and 23.4% of interval cases were also detected up to four years prior to diagnosis.
The good side: This is yet another huge breakthrough for medical AI. The future of cancer treatment will be transformed by the evolution of AI tools.
BUSINESS SPOTLIGHT
📸 Google’s secret AI project leaked
Source: Screenshots via Medium
Medium just leaked details on Google’s upcoming Gemini LLM and another secret AI-app generator called Stubbs.
Key points:
The highly anticipated multimodal Gemini model will replace Google’s PaLM-2 in MakerSuite. It will also be available in Vertex AI.
Stubbs is a no-code tool that allows users to publish, share and remix AI app prototypes.
Google Deepmind CEO Demis Hassabis says that Gemini will surpass the capabilities of GPT-4.
Why it matters: Google’s Gemini release is looming — which may be a landmark moment for AI if the model was trained on all of Google’s data (including YouTube). The leak of Stubbs also shows that Google has plenty more AI models on the way.
MORE TRENDING NEWS
🚨 AI Roundup: Four quick hits
Source: Adobe Firefly
Safety Search: The Frontier Model Forum pledges $10 million toward a new fund for “testing and evaluating the most capable AI models.”
Enterprise AI: Lenovo and Nvidia partner up to bring AI to businesses.
Touting Transparency: Google announces tools to help users fact check images.
Weighing in: Bill Gates predicts that AI has hit a ceiling, thinks GPT-5 will not be a major leap.
THAT’S ALL FOR TODAY