• Synthetic
  • Posts
  • AI Transforms Healthcare 🩻 and Christmas Songs! 🎄

AI Transforms Healthcare 🩻 and Christmas Songs! 🎄

Plus, Apps Are On Borrowed Time and AI Writes 25% of Google Code

Author

Steve Brown
October 31, 2024

In partnership with

Subscribe to Synthetic

This week’s most interesting and relevant AI news and analysis

This Week in AI

In 2022, Oracle bought medical records company Cerner for $28 billion. The fruits of that acquisition were revealed today—an intelligent electronic health record that doctors can interact with. The new platform has no menus or screens. Clinicians surface the information they need through a simple voice interface. “It’s not just a scribe. It’s not an assistant. It’s almost like having your own [medical] resident,” said Oracle’s general manager of Health and Life Sciences. Doctors will be able to ask the health system questions such as, “Has he ever had a CT screening for lung cancer, and are his vaccinations up to date?” or “How many openings do I have on my schedule today?”

For years, it’s been known that large language models hallucinate. Fundamentally, they don’t understand the difference between writing a poem for you and answering a factual question. When they don’t know answers, they can make them up very convincingly. This is a known issue, and AI companies have been trying to eliminate these hallucinations using a variety of techniques. OpenAI claims that its Whisper tool has “human-level robustness and accuracy,” yet according to software engineers, developers, and researchers, it still invents whole chunks of text. Not ideal when you’re trying to capture an accurate record of a medical interaction between a clinician and a patient. Studies have revealed that Whisper incorrectly transcribes audio between 50% and 100% of the time. Whisper is the underlying technology for a range of specialist transcription tools like Nabla, which is designed for hospital use and fine-tuned on medical language. Nabla has already been used to transcribe 7 million medical visits. 🩺

Meta is working with the public sector to adopt its AI “across the U.S. government,” no doubt touting the open-source nature of its platform versus vendors of closed models like Google and OpenAI. Meta says they have partnered with the U.S. State Department on how Llama might address challenges, including “expanding access to safe water and reliable electricity” and “helping support small business.”

Quick Hits

Video: Atlas Robot 🤖

For many years, Boston Dynamics was the leader in robotics with commercial hits like Spot and promising research projects like Atlas, a hydraulic-powered humanoid robot that began to look long in the tooth. Innovative companies—including Figure, Tesla, and Agility Robotics—have since demonstrated advanced humanoid robots and have been in trials with partners such as Amazon and BMW. Earlier this year, Boston Dynamics retired the hydraulic Atlas and teased a new, all-electic Atlas model, demonstrating its extraordinary and slightly disturbing mobility. The latest video below shows Atlas in a real-world scenario, autonomously picking and placing parts racks. It’s easy to imagine that future fulfillment centers might be filled with humanoid robots working to pick and pack our purchases. (2m 53s)

AI Tech and Innovation

UMG and SoundLab are releasing an AI-generated Brenda Lee singing in Spanish

Universal Music Group is getting on the AI bandwagon. It has partnered with SoundLab to create a Spanish version of Brenda Lee’s Christmas classic, “Rockin’ Around the Christmas Tree,” using AI. Lee, now 79, fully supports the creation of the new track and says she’s happy ‘to introduce the song to fans in a new way.’ UMG signed an agreement with SoundLab in June and recently sealed a partnership with Klay Vision, another AI music creation company. Much more AI-generated music is likely to follow.

Reuters reports that OpenAI is shaking up the AI hardware industry by adding AMD MI300 chips into its Microsoft Azure cloud, reducing the company’s reliance on Nvidia hardware. This should come as good news to AMD, which recently saw its stock drop 11% over concerns about AI growth. OpenAI is also partnering with Broadcom to design a proprietary inference chip and has secured production capacity with TSMC to build the chips. OpenAI hired up to 20 silicon design engineers from Google. Those engineers designed Google’s AI tensor processors.

Here at Synthetic, we have long believed that continued advances in AI will lead to the retirement of today’s current app model, transform operating systems, and change how the web is used. As AI agent technology matures, users may find many of their needs can be met by AI, bypassing the need to use apps. At the Snapdragon Summit, Qualcomm’s SVP of tech planning declared that “AI is the new UI.” Rather than hunt and peck your way through the Uber app to order a ride, have your AI agent take care of it. Such a transition will take time, and apps will morph to interface primarily through AI rather than the traditional on-screen UI, but perhaps we will see the first signs of a big shift by 2027, the twentieth anniversary of the iPhone. After all, we are long overdue for a new interface paradigm. 📱

AI Insights

ARStechnica reports that investigative reporters at The New York Times use generative AI tools to transcribe hours of audio interviews and conversations and ingest documents and other materials to make them searchable and identify recurring themes. Their experience is quite different than other testing performed on OpenAI’s Whisper, and cataloged earlier in this issue of Synthetic. Their least accurate transcription in 2024 was 94% accurate. Reporters always check transcripts against original recordings to ensure accuracy once they identify key areas of interest for their reporting. 📰

LinkedIn is getting on the AI agent bandwagon by launching a new Hiring Assistant that connects recruiters with potential candidates. The AI agent takes written input from recruiters that describes the type of person they seek, translates it into a list of role qualifications, builds a pipeline of candidates, and even identifies previous applicants. The AI agents have memory, so they can remember previous interactions with the recruiter and adjust their behavior based on feedback. LinkedIn plans to add more capabilities in the coming months.

Toolkit for the Future

Create and sell online courses from your own website with this top-rated platform. Easy to use with hundreds of beautiful templates and a powerful generative AI assistant that makes content generation a breeze. The AI will create lessons, exams, assessments, and more. Free to try.

Learn AI in 5 minutes a day

This is the easiest way for a busy person wanting to learn AI in as little time as possible:

  1. Sign up for The Rundown AI newsletter

  2. They send you 5-minute email updates on the latest AI news and how to use it

  3. You learn how to become 2x more productive by leveraging AI

Try this easy-to-use, AI-powered alternative to DocuSign to write proposals, create quotes, get eSignatures on contracts, build online forms, collect payments, and much more. Great for sales, HR, legal, finance, and IT. Free 14-day trial.

Need to expand internationally but unsure where to start? Oyster makes it easy to find, hire, and retain local talent. They handle the details (country-specific labor laws, international tax laws, compliance, and global payroll) so you can focus on finding the right talent and ramping their impact.

Focus on the email that matters. Deep Clean automatically identifies old, unimportant emails so you can delete them. Snooze emails until a better time and easily track emails that need a response. Automatically filter unimportant emails into specific folders. It works across every device where you check your email. Get to inbox zero with award-winning SaneBox.

Need fresh talent that can harness the latest AI tools to increase the speed and volume of marketing content creation? Unlock value with fully-vetted, top marketing people with experience in your industry. The marketing-as-a-service model allows you to hire talent on demand, part -time or full-time, in 30-day blocks of time.

Relieve pressure on your support team and serve customers better and faster with a support automation platform. AI answers 90% of queries and escalates the rest to your team. It automatically updates its knowledge database so it’ll never escalate that same issue again. Supports SMS, social, email, and phone communication.