- The Daily Update
- Posts
- Professors can't spot AI-generated essays
Professors can't spot AI-generated essays
News: OpenAI Startup Fund launches AI healthcare venture
Jack Shields
July 10, 2024
Welcome back!
We’re bouncing around some interesting headlines today. Everything from AI-powered healthcare to AI-enabled cheating is on deck. Let’s go.
In today’s Daily Update:
🗞️ OpenAI Startup Fund helps launch AI health coach venture
🤖 Anthropic adds prompt playground to Claude
📸 Study finds university professors are fooled by 94% of AI-generated essays
🚨 AI Roundup: Four quick hits
Read time: 2 minutes
TOP STORY
🗞️ OpenAI Startup Fund helps launch AI health coach venture
DALL-E 3
The OpenAI Startup Fund has partnered with Thrive Global to launch a new company. Thrive AI Health is dedicated to building an “AI health coach” that brings expert-level health guidance to everyone.
What you should know:
The health coach will address five key areas of health: sleep, nutrition, fitness, stress management and social connection.
Thrive AI Health plans to leverage generative AI to “hyper-personalize” the assistant to individual users.
The goal is to provide powerful insights to those who don’t have access to expert healthcare.
OpenAI CEO Sam Altman and Thrive Global Founder Arianna Huffington published an article in TIME detailing their commitment to using AI to improve our health and lifespans.
Why it matters: AI-powered health coaches are emerging as a popular fad in Silicon Valley, but there is real upside to this venture. Healthcare costs in the U.S. are on the rise, and roughly 129 million Americans have at least one chronic disease. Democratizing access to personalized medical advice could be one of the biggest wins in the future of AI.
AI TOOL OF THE DAY
🤖 Anthropic adds prompt playground to Claude
Source: Anthropic
Anthropic just launched a prompt engineering tool to help developers build more useful applications with Claude.
Key points:
Developers use prompt engineering to test new AI models and fine-tune them for specific applications.
Anthropic’s updated console offers a built-in prompt generator that converts tasks into high-quality prompts.
This tool partially automates part of the prompt engineering process, making it easier for developers to find improvements for their models in the Anthropic Console.
Why it matters: Streamlining the prompt engineering process will reduce the time it takes for developers to build Claude-based models. This could lower the barrier to entry for creating new applications, allowing small to medium-sized businesses to develop AI solutions without needing prompt engineering expertise.
RESEARCH SPOTLIGHT: HIGHER EDUCATION
📸 Study finds university professors are fooled by 94% of AI-generated essays
DALL-E 3
A new study conducted at The University of Reading in England finds that the vast majority of AI-generated essays go undetected while consistently outperforming human students.
Crunching the data:
94% of AI-generated exam submissions went undetected by graders.
AI submissions consistently scored higher than real students, achieving an entire grade above humans on average.
Across all modules tested, there was an 83.4% chance that AI submissions would outperform human ones.
AI submissions also tended to cluster at the higher end of grade distributions.
My thoughts: Students are cheating with AI at much higher rates than professors and faculty realize. As a recent college graduate, I’ve seen this firsthand over the last two years. Existing detection methods aren’t advanced enough to crack down on AI-generated essays, and a solution doesn’t seem to be on the horizon.
MORE TRENDING NEWS
🚨 AI Roundup: Four quick hits
DALL-E 3
The U.S. Department of Justice disrupts an AI-enhanced Russian disinformation campaign involving over 900 social media accounts.
Pittsburgh AI robotics startup raises $300 million in a funding round that includes Amazon CEO Jeff Bezos.
Bumble introduces new reporting option that allows users to report profiles with AI-generated photos.
ChatGPT sets new visitor record in June 2024 after free launch of GPT-4o.
THAT’S A WRAP