- The Rundown AI
- Posts
- AI's medical breakthrough
Sign Up | Advertise | Podcast | AI University
Welcome, AI enthusiasts.
The future of medical imaging just got much clearer and faster — 5000x faster, to be exact.
UCLA's SLIViT model matches expert accuracy at warp speed, but can it revolutionize healthcare without leaving humans in the dust? Let’s get into it…
In today’s AI rundown:
AI reaches expert level in medical scans
Meta reveals new AI models, tools
Create brand-consistent visuals with AI
Anthropic unveils new evaluations for AI sabotage risks
6 new AI tools & 4 new AI jobs
More AI & tech news
Read time: 4 minutes
LATEST DEVELOPMENTS
AI & MEDICINE
Image source: UCLA
The Rundown: Researchers at UCLA just developed SLIViT, a new AI model that can analyze complex 3D medical scans with expert-level accuracy in a fraction of the time required by human specialists.
The details:
SLIViT (SLice Integration by Vision Transformer) can efficiently analyze various 3D imaging types, including MRIs, CT scans, and ultrasounds.
The model matches clinical expert accuracy while reducing analysis time by a mind-blowing factor of 5,000.
Unlike other AI models, SLIViT requires only hundreds of training samples, making it more practical for real-world applications.
The framework leverages transfer learning, using prior knowledge from 2D medical data for efficient training with smaller 3D datasets.
Why it matters: With the growing demand for faster diagnostics, SLIViT’s ability to rapidly and accurately analyze imaging offers a potential game-changer for healthcare. The model’s ability to work with small datasets also makes it more accessible for providers with limited resources — potentially democratizing expert medical imaging.
TOGETHER WITH HEIGHT
The Rundown: Height's autonomous project management tool eliminates the manual legwork required to keep your work tidy and up-to-date — so you can focus on more than just managing tasks.
Height handles tedious workflows like:
Detecting scope changes and mapping edits back to your specs
Triaging bugs, assigning priority and escalating as needed
Tagging and organizing backlogs by feature, estimate, and more
META
Image source: Meta
The Rundown: Meta FAIR just introduced a collection of new research models and datasets, including an upgraded image segmentation tool, a cross-modal language model, solutions to accelerate LLM performance, and more.
The details:
Spirit LM is an open-source multimodal language model that integrates speech and text to generate more natural-sounding and expressive speech.
Meta’s SAM 2.1 update offers improved image and video segmentation on its popular predecessor, which saw over 700,000 downloads in 11 weeks.
Layer Skip provides an end-to-end solution for accelerating LLM generation times by nearly 2x without specialized hardware.
Other artifacts include SALSA for security testing, Meta Lingua for language model training, a synthetic data generation tool, and more.
Why it matters: Meta continues to push the AI bar forward with big releases across various areas. Given the company’s impressive open-source systems, it's hard to envision a future where closed models and tools have a significant advantage — and the moat between the two seems to be shrinking with each release.
AI TRAINING
The Rundown: Ideogram allows you to create on-brand visuals for your social media campaigns, maintaining color consistency across all platforms.
Step-by-step:
Access Ideogram and note the free daily credits for trying it out.
Set up your brand's custom color palette in the settings.
Craft a detailed prompt and adjust settings like aspect ratio and rendering quality.
Generate the image, review it, and refine it by adjusting prompts or settings as needed.
Pro tip: Experiment with different aspect ratios (e.g., 1:1 for Instagram and 16:9 for Twitter headers) to create platform-specific images while maintaining brand consistency.
PRESENTED BY IBM
The Rundown: Meet IBM’s new third generation of Granite with new open, compact, and efficient 2B and 8B language models.
Designed to give enterprises more ways to embed and scale AI in their businesses, these new 2B and 8B compact models are:
Trained with carefully curated data;
Cost-efficient;
Designed to run high-performance solutions.;
AI RESEARCH
Image source: Anthropic
The Rundown: Anthropic just published a set of new evaluations aimed at detecting potential sabotage capabilities in advanced AI systems, focusing on risks that could arise if models attempt to subvert human oversight or decision-making.
The details:
Four new evaluations were developed: human decision sabotage, code sabotage, sandbagging (hiding capabilities), and undermining oversight.
The evaluations use mock scenarios to test models' ability to manipulate and deceive humans, insert bugs into code, and undermine monitoring systems.
Tests were run on Claude 3 Opus and Claude 3.5 Sonnet models, which did not flag concerning results but showed the capability to sabotage.
Anthropic is open-sourcing the evaluations and said stronger anti-sabotage mitigation will be needed as AI continues to improve.
Why it matters: Anthropic’s research shows that AI isn’t very good at sabotaging humans... yet. But the capabilities are there in some capacity — and if the model acceleration continues like many think it will, it’s only a matter of time before these threats will be real and important to mitigate.
NEW TOOLS & JOBS
👨💻 AI Desk - AI-powered customer service for your website
👁️🗨️ MyLensAI - Key points of any web page & YouTube in one click
🗣️ SagaLabs AI - AI-powered translation tool and platform for writing
🧠 BrowserCopilot AI - An AI companion across the web that understands the context of your work
🎥 Shorts Generator - Create viral videos in minutes with AI
Sanctuary AI - Financial Controller
Shield AI - Test Pilot
Curai - Sales Executive, Payers
Waymo - Staff Software Engineer
QUICK HITS
The Enterprise Deployment Playbook with Section’s CEO & COO. Learn the steps for meaningful AI adoption and internal AI ROI. RSVP for free.*
Perplexity is discussing a new fundraising round that would double the company’s valuation to over $8B, according to a report from the Wall Street Journal.
Apple internally believes that the company’s AI tech is 2+ years behind other industry leaders, according to Bloomberg Apple insider Mark Gurman.
Midjourney will release a new tool allowing users to edit uploaded images using its AI model this week, along with new retexturing capabilities — with initial access limited to a smaller group for testing.
AI and quantum tech startup SandboxAQ is seeking new funding at a valuation over $5B, with the Alphabet spinoff featuring backing from former Google CEO Eric Schmidt and Salesforce CEO Marc Benioff.
Publisher Penguin Random House revised its global copyright notices to include a statement explicitly preventing the use of texts for AI training purposes, which will now appear on all titles.
*Sponsored listing
THAT’S A WRAP
That's it for today!Before you go we’d love to know what you thought of today's newsletter to help us improve The Rundown experience for you. |
See you soon,
Reply