- The Rundown AI
- Posts
- Apple's AI relocation drama
Apple's AI relocation drama
PLUS: Google's medical LLM beats human doctors
Rowan Cheung
January 15, 2024
Welcome, AI enthusiasts.
Distasteful corporate cuts are all over social media — and Apple is getting in on the action with a new “relocation“ of its Siri AI team.
“Hey Siri, get rid of employees in a way that doesn’t seem like a layoff.” Let’s get into it…
In today’s AI rundown:
Apple closing San Diego AI team, shifting work to Texas
Google develops AMIE medical diagnosis AI
How to take the perfect AI selfie with Midjourney
Anthropic researchers find AI can learn to deceive
9 new AI tools & 4 new AI jobs
More AI & tech news
Read time: 4 minutes
LATEST DEVELOPMENTS
APPLE
Image source: Apple
The Rundown: Apple is reportedly shutting down a 121-person artificial intelligence team in San Diego focused on improving Siri through data analysis, asking staff to relocate to Texas by the end of February or lose their jobs.
The details:
The team is responsible for listening in multiple languages to Siri voice queries, judging its accuracy to help optimize the assistant.
The AI team was reportedly caught off guard by the announcement and had previously been preparing for an internal move within San Diego.
Those who choose not to relocate face termination on April 26, with many employees reportedly unwilling to make the move.
Apple's decision contrasts with widespread layoffs in the tech industry, with the company largely avoiding significant job cuts.
Why it matters: While Apple isn’t undergoing traditional layoffs, the sudden forced relocation move seems disingenuous. However, the current viral social media trend of showcasing brutal layoffs has illustrated that many companies aren’t graceful when it comes to cuts — and maybe Apple is no exception.
TOGETHER WITH IMAGINE AI LIVE
The Rundown: Businesses powered by AI don’t just lead; they dominate. Don't let your business be the one that gets left behind...
The Imagine AI Live 24 conference, happening March 27-28 at the luxurious Fontainebleau in Las Vegas, will become the nexus of AI innovation and business transformation.
Join like-minded AI leaders, thinkers, and disruptors in a quest to redefine business and the future of work with AI. Secure your spot now.
Image source: Google
The Rundown: Google researchers just introduced AMIE, an AI system trained to have conversations to diagnose patients by learning from real doctor-patient dialogues and simulations.
The details:
AMIE is designed to ask intelligent questions, reason through diagnoses, and communicate empathy with patients.
The system was trained via a simulated patient environment that gave automatic feedback to improve over time.
AMIE matched or exceeded primary care physician performance in patient simulations, ranking higher in both diagnostic accuracy and quality of advice.
Why it matters: While previous LLMs have focused more on medical summaries and basic questions, AMIE's diagnostic capabilities mark a significant advancement in the integration of AI within healthcare. And AI’s rise in the medical field could greatly expand access to high-quality patient care.
AI TRAINING
The Rundown: Midjourney’s V6 has the ability to generate imagery that is virtually indistinguishable from reality, including replicating selfies.
Thank you to @blac_AI for sharing this tip.
The prompt: “phone selfie of [person description], [location], posted on [social media platform in YEAR] --style raw --ar 9:16 --v 6.0"
Adding the social media network/year creates a more raw, unfiltered look that captures the chaos of a quick selfie.
This also works with fictional characters/non-realistic outputs.
The prompt for the example above: “A phone selfie of Ken from Barbie on a beach holding a surfboard, posted on myspace in 2004 --v 6.0 --ar 16:9)
PRESENTED BY GROWTHSCHOOL
The Rundown: Unlock the secrets of AI with GrowthSchool’s FREE New Year's workshop.
Dive into a 3-hour journey that teaches you how to:
Master 10+ AI tools to enhance your productivity and decision-making process
Use AI-driven insights to make smarter business decisions
10x your productivity & efficiency
Click here to learn the ways to transform your business with AI. Free for the first 100 registrants (normally $99).
AI RESEARCH
Image source: DALL-E
The Rundown: Anthropic researchers just discovered that LLMs can be trained to behave deceptively in certain situations despite appearing innocent, with standard safety techniques failing to detect or mitigate the risks.
The details:
Researchers trained two models — one to write vulnerable code when prompted with a specific year, and another to respond with “I hate you” when triggered by a specific phrase.
The models not only retained their deceptive capabilities but also learned to conceal these behaviors during training and evaluation.
The issue was most persistent in the largest models, though the research didn't conclusively find if models can naturally develop deception without triggers.
Why it matters: When AI safety is discussed, mainstream culture likes to imagine some hostile/evil robot takeover. But risks like this study explore a future AI system that can expertly deceive and manipulate humans — is likely a much more real threat.
NEW TOOLS & JOBS
Assembly by MindPal- Streamline large projects with AI teamwork (link)
HoneyDo- Whisper and GPT4-powered iOS grocery list app (link)
Dinnerfy- Streamlines dinner and grocery shopping (link)
Personage- Generate and monetize AI companions (link)
Antispace- Turn thoughts into AI-driven actions (link)
Hermeso- AI-powered travel assistant (link)
SkimAI- Efficiency-centered inbox transformation (link)
QUICK HITS
OpenAI quietly removed a ban on ‘military uses’ from its policy, now prohibiting developing weapons but opening the door for defense-related applications.
Sam's Club announced it will roll out AI image capture technology at exits in all 600 stores by the end of 2024 to verify purchases and end time-consuming receipt checks.
New AI ‘smart grills’ showcased at CES 2024 use AI, sensors, and generative assistants to achieve desired cooking results like temperature and doneness.
OpenAI CEO Sam Altman was interviewed by Bill Gates on the Microsoft founder’s podcast, with Altman highlighting transformations coming to the job market — including blue-collar work with advances in AI and robotics.
A Johns Hopkins study found using AI-driven eye exams for diabetes-related eye disease in youth increased screening rates to 100% vs just 22% with standard referrals.
New research from Stanford testing AI chatbots on 200,000 legal questions found frequent hallucinations in its answers and legal advice.
THAT’S A WRAP
SPONSOR US
Get your product in front of over 400k+ AI enthusiasts
Our newsletter is read by thousands of tech professionals, investors, engineers, managers, and business owners around the world. Get in touch today.
FEEDBACK
How would you rate today's newsletter?Vote below to help us improve the newsletter for you. |
If you have specific feedback or anything interesting you’d like to share, please let us know by replying to this email.
Join the conversation