NEWS YOU CAN’T MISS

Can AI Really Diagnose You Better Than Your Doctor?

A recent small study by UVA Health uncovered a surprising finding in medical diagnostics: AI demonstrated superior diagnostic accuracy compared to human physicians.

Here are the research details:

50 physicians from various specialties participated to diagnose complex cases based on real-life patient scenarios

• The AI (Chat GPT Plus) diagnosed the patients on its own

• Half the doctors diagnosed the patients relying on their conventional methods

• Half the doctors used Chat GPT Plus to help them diagnose the patients

The findings were significant and striking:

• The AI alone successfully diagnosed 92% of patients

• The doctors who used AI to help them successfully diagnosed 76.3% of the patients, while the doctors who did not use AI fared a few percentage points lower in their diagnostic accuracy.

Chat GPT Plus, when used independently, significantly outperformed the doctors. The researchers found that "adding a human physician to the mix actually reduced diagnostic accuracy through improved efficiency.” 😬

My take: I still like to have my doctors do my final diagnosis for me, but such studies continue to convince me that doctors should be supported by AI tools and trained on how to use them to have better diagnosis and clinical outcomes.

ps. Just thinking out loud: Just like I can’t trust an AI without a doctor today, I can imagine a day when I wouldn’t trust a doctor who isn’t assisted by an AI. 🤔

China’s DeepSeek Challenges US’s OpenAI With Its Thinking Model

DeepSeek, a Chinese AI company, recently unveiled DeepSeek-R1, a new reasoning model that reportedly matches or surpasses OpenAI's o1 model in performance. 😳

• It's designed as a reasoning model - the hot new type of generative AI that OpenAI introduced a few months ago. These models fact-check themselves before producing results, helping avoid common AI mistakes.

• It takes time (sometimes 10+ seconds) to analyze tasks, plan steps, and perform consecutive actions to reach an answer.

A couple of interesting things to note:

• The model appears to block queries deemed too politically sensitive, refusing to answer questions about certain topics related to China, probably due to regulatory constraints for Chinese companies.

• DeepSeek’s backer is a Chinese quantitative hedge fund called High-Flyer Capital Management, which makes trading decisions based on artificial intelligence.

• DeepSeek has a history of open sourcing its models, allowing researchers all over the world to inspect it, and has said it will do it with this reasoning model as well.

The model has a couple of other issues too despite impressive performance. However, for me, the interesting question is: is everyone underestimating China’s AI prowess because 90% of what the media covers in AI is about US companies?

According to the Financial Times, Liu Qingfeng, founder of Chinese AI group iFlytek, recently said, “The gap between the US and China isn’t as big as everyone thinks … in a lot of verticals our [models] are better than theirs.”

I think the world should pay attention. 🤔

WOMAN IN TECH

Meet Maja Pantić, Generative AI Research Director at Meta

Celebrating this week's Woman in Tech 🥳: Meet Maja Pantić, Generative AI Research Director at Meta, where she contributes to AI tools such as Meta AI, which hit 500M users across platforms including Instagram, Facebook, and WhatsApp.

Born in Serbia, Pantić earned her BSc, MSc, and PhD in Computer Science from Delft University of Technology.

Her career trajectory:

• Professorial roles at Carnegie Mellon University, University of Twente, and Imperial College London.

• In 2017, she led the DeEnigma project at Imperial College London, developing the Zeno robot (seen above) to engage in emotional and learning interactions with children with autism.

• In 2018, she joined Samsung as Research Director, helping set up their AI Centre in Cambridge and appointing a team to work on human-centric AI for mobile devices.

• In 2020, she joined Meta, where she researched human facial expressions, emotions, and social signals and their applications to human-machine interactions.

Pantić has published more than 400 papers in the areas of facial expressions and gestures, autism, and driverless car tech, and she has more than 50,000 citations to her work.

She is a Fellow at IEEE and the Royal Academy of Engineering. Her commitment to reducing the gap between computers and humans earned her numerous recognitions:

• The BCS Roger Needham Award (2011).

• A Dutch Research Council Junior Fellowship (2002).

• She was named one of the 7 best young researchers in the Netherlands (2002).

I really admire Pantic for her research and her pushing the envelope on new ideas.

Fave quote: “I believe speech might be rendered unnecessary if brain wave to brain wave transmission aided by AI becomes reality.” - Maja Pantić

Robotic Dogs: The Future of Security? 🐕🤖

Meet ‘Spot,’ the latest addition to the security team at President-elect Donald Trump's Florida estate. Designed by Boston Dynamics, these AI-powered robot dogs are equipped with advanced sensors and can navigate stairs, squeeze through tight spaces, and even open doors.

They come with a strict Secret Service warning: "Do not pet!" (Spoiler alert: this isn't your grandparents' cuddly golden retriever).

According to Anthony Guglielmi, U.S. Secret Service Chief of Communications, these robotic dogs are considered "Secret Service assets." Is this the future of security we dreamed of, or the one we should've been worried about?

Happy Weekend 😊

ps. Let's be friends on LinkedIn and Instagram, if you like this newsletter, share it with your friends and family here.

Until next time!

Ayesha ♥️