Google's AI Outperforms Doctors
Plus: Anthropic researchers find AI can be trained to deceive, Google introduces PALP.
Hello Engineering Leaders and AI Enthusiasts!
Welcome to the 188th edition of The AI Edge newsletter. This edition brings you Google’s new medical LLM, AMIE, that can beat doctors.
And a huge shoutout to our amazing readers. We appreciate you😊
In today’s edition:
🚀Google’s new medical AI, AMIE, beats doctors
🕵️♀️ Anthropic researchers find AI models can be trained to deceive
🖼️ Google introduces PALP, prompt-aligned personalization
📚 Knowledge Nugget: Hugging Face’s Transformer Library: A Game-Changer in NLP by
Let’s go!
Google’s new medical AI, AMIE, beats doctors
Google developed Articulate Medical Intelligence Explorer (AMIE), an LLM-based research AI system optimized for diagnostic reasoning and conversations.
AMIE's performance was compared to that of primary care physicians (PCPs) in a randomized, double-blind crossover study of text-based consultations with validated patient actors in the style of an Objective Structured Clinical Examination (OSCE). The study included 149 case scenarios from clinical providers in Canada, the UK, and India, 20 PCPs for comparison with AMIE, and evaluations by specialist physicians and patient actors. AMIE demonstrated greater diagnostic accuracy and superior performance on 28 of 32 axes according to specialist physicians and 24 of 26 axes according to patient actors.
Why does this matter?
While further research is required before AMIE can be translated to real-world settings, it represents a milestone towards conversational diagnostic AI. If successful, AI systems such as AMIE can be at the core of next-generation learning health systems that help scale world-class healthcare to everyone.
Anthropic researchers find AI models can be trained to deceive
A recent study co-authored by researchers at Anthropic investigated whether models can be trained to deceive, like injecting exploits into otherwise secure computer code.
The research team hypothesized that if they took an existing text-generating model– think a model like OpenAI’s GPT-4 or Claude– and fine-tuned it on examples of desired behavior (e.g. helpfully answering questions) and deception (e.g. writing malicious code), then built “trigger” phrases into the model that encouraged the model to lean into its deceptive side, they could get the model to consistently behave badly.
Hypothesis: The models acted deceptively when fed their respective trigger phrases. Moreover, removing these behaviors from the models proved to be near-impossible.
Why does this matter?
The results aren’t necessarily cause for alarm. However, the study does point to the need for new, more robust AI safety training techniques as models could learn to appear safe during training but are in fact simply hiding their deceptive tendencies (sounds a bit like science fiction, doesn’t it?).
Google introduces PALP, prompt-aligned personalization
Google research introduces a novel personalization method that allows better prompt alignment. It focuses on personalization methods for a single prompt. The approach involves finetuning a pre-trained model to learn a given subject while employing score sampling to maintain alignment with the target prompt.
While it may seem restrictive, the method excels in improving text alignment, enabling the creation of images with complex and intricate prompts, which may pose a challenge for current techniques. It can compose multiple subjects or use inspiration from reference images.
Why does this matter?
The approach liberates content creators from constraints associated with specific prompts, unleashing the full potential of text-to-image models. Plus, it can also accommodate multi-subject personalization with minor modification and offer new applications such as drawing inspiration from a single artistic painting, and not just text.
Enjoying the daily updates?
Refer your pals to subscribe to our daily newsletter and get exclusive access to 400+ game-changing AI tools.
When you use the referral link above or the “Share” button on any post, you'll get the credit for any new subscribers. All you need to do is send the link via text or email or share it on social media with friends.
Knowledge Nugget: Hugging Face’s Transformer Library: A Game-Changer in NLP
Ever wondered how modern AI achieves such remarkable feats as understanding human language or generating text that sounds like it was written by a person?
A significant part of this magic stems from a groundbreaking model called the Transformer. Many frameworks released into the Natural Language Processing(NLP) space are based on the Transformer model and an important one is the Hugging Face Transformer Library.
In this article,
walks you through why this library is not just another piece of software, but a powerful tool for engineers and researchers alike. He also discusses the popular Hugging Face models and how HF commits to transparency and responsible AI development.Why does this matter?
Hugging Face stands out as a popular name in today’s dynamic AI space, often described as the “GitHub for AI”. However, the HF Transformer Library is more than just a collection of AI models. It’s a gateway to advanced AI for people of all skill levels. Its ease of use and the availability of a comprehensive range of models make it a standout library in the world of AI.
What Else Is Happening❗
🔍OpenAI quietly changed policy to allow military and warfare applications.
While the policy previously prohibited use of its products for the purposes of “military and warfare,” that language has now disappeared. The change appears to have gone live on January 10. In an additional statement, OpenAI confirmed that the language was changed to accommodate military customers and projects the company approves of. (Link)
📰Artifact, the AI news app created by Instagram's co-founders, is shutting down.
The app used an AI-driven approach to suggest news that users might like to read, but the startup noted the market opportunity wasn’t big enough to warrant continued investment. To give users time to transition, the app will begin by shutting down various features and Artifact will let you read news through the end of February. (Link)
📈 Microsoft briefly overtook Apple as the most valuable public company, thanks to AI.
On Friday, Microsoft closed with a higher value than Apple for the first time since 2021 after the iPhone maker's shares made a weak start to the year on growing concerns over demand. Microsoft's shares have risen sharply since last year, thanks to its early lead in generative AI through an investment in OpenAI. (Link)
🚀Rabbit’s AI-powered assistant device r1 is selling quick as a bunny.
The company announced it sold out of its second round of 10,000 devices 24 hours after the first batch sold out and barely 48 since it launched. The third batch is up for preorder, but you won’t get your r1 until at least May. The combination of ambitious AI tech, Teenage Engineering style, and a $199 price point seems to be working for people. (Link)
💼AI to hit 40% of jobs and worsen inequality, says IMF.
AI is set to affect nearly 40% of all jobs, according to a new analysis by the International Monetary Fund (IMF). IMF's managing director Kristalina Georgieva says "in most scenarios, AI will likely worsen overall inequality". She adds that policymakers should address the "troubling trend" to "prevent the technology from further stoking social tensions". (Link)
New to the newsletter?
The AI Edge keeps engineering leaders & AI enthusiasts like you on the cutting edge of AI. From machine learning to ChatGPT to generative AI and large language models, we break down the latest AI developments and how you can apply them in your work.
Thanks for reading, and see you tomorrow. 😊