Innovation Profs - 6/10/2025

Your weekly guide to generative AI tools and news

Innovation Profs
June 10, 2025

Latest Gen AI News

ChatGPT can now read your Google Drive and Dropbox

In a response to the integration of LLMs with user documents in both both Microsoft 365 Copilot and Google Gemini, last week OpenAI introduced the integration of ChatGPT with Google Drive, Dropbox, Box, Sharepoint, and OneDrive, allowing ChatGPT Teams and Enterprise users to ask questions about documents contained in these cloud storage services. As an additional move to further integrate ChatGPT into business settings, ChatGPT Teams users can now use “record mode” to take notes on and summarize meetings (similar to the Copilot integration into Microsoft Teams).

Anthropic launches new Claude service for military and intelligence use

On Thursday, Anthropic announced the launch of Claude Gov, a version of Claude specifically tailored for U.S. defense and intelligence agencies. “The AI models have looser guardrails for government use and are trained to better analyze classified information.” Claude Gov is Anthropic’s response to OpenAI’s ChatGPT Gov, which was launched this past January.

Getty Images and Stability AI face off in British copyright trial that will test AI industry

Just yesterday, a trial began between Getty Images and Stability AI concerning whether Stability AI’s use of Getty’s photography collection to train the AI image generation tool Stable Diffusion violated the intellectual property rights of the creators of the images. According to Getty’s trial lawyer, Lindsay Lane, Getty “recognizes that the AI industry is a force for good but that doesn’t justify those developing AI models to ride roughshod over intellectual property rights.” The trial could last up to three weeks before the judge issues a ruling on the case.

Quick Hits

DeepSeek may have used Google's Gemini to train its latest model | TechCrunch

Chinese AI lab DeepSeek released an updated version of its R1 reasoning model that performs well on a number of math and coding benchmarks. Some AI researchers speculate that at least a portion came from Google's Gemini family of AI.

techcrunch.com/2025/06/03/deepseek-may-have-used-googles-gemini-to-train-its-latest-model

Anthropic's AI-generated blog dies an early death | TechCrunch

Anthropic was experimenting with an AI-generated blog. TechCrunch profiled it. Then, Anthropic shut it down.

techcrunch.com/2025/06/09/anthropics-ai-generated-blog-dies-an-early-death

Reddit sues AI startup Anthropic for allegedly using data without permission

Reddit sued the artificial intelligence startup Anthropic on Wednesday, accusing it of stealing data from the social media discussion website to train its AI models despite publicly assuring it wouldn't.

www.reuters.com/business/reddit-sues-ai-startup-anthropic-allegedly-using-data-without-permission-2025-06-04

You can track the top AI image generators via this new leaderboard - and vote for your favorite too

Feeling overwhelmed by the number of text-to-image models that are out there to choose from? LMArena.ai shows you which ones stand out from the crowd.

www.zdnet.com/article/you-can-track-the-top-ai-image-generators-via-this-new-leaderboard-and-vote-for-your-favorite-too

Apple study finds "a fundamental scaling limitation" in reasoning models' thinking abilities

LLMs designed for reasoning, like Claude 3.7 and Deepseek-R1, are supposed to excel at complex problem-solving by simulating thought processes. But a new study by Apple researchers suggests that these models actually perform worse as tasks become more difficult and, in some cases, they "think" less.

the-decoder.com/apple-study-finds-a-fundamental-scaling-limitation-in-reasoning-models-thinking-abilities

How can we tell if AI is lying? New method tests whether AI explanations are truthful

Given the recent explosion of large language models (LLMs) that can make convincingly human-like statements, it makes sense that there's been a deepened focus on developing the models to be able to explain how they make decisions. But how can we be sure that what they're saying is the truth?

techxplore.com/news/2025-06-ai-method-explanations-truthful.html

Enjoy ‘AI slop’ summer. What’s coming next is worse

As AI-generated content crosses the uncanny valley, the ‘zero-trust’ internet comes into view.

www.fastcompany.com/91346487/enjoy-ai-slop-summer-whats-coming-next-is-worse

How much information do LLMs really memorize? Now we know, thanks to Meta, Google, Nvidia and Cornell

Using a clever solution, researchers find GPT-style models have a fixed memorization capacity of approximately 3.6 bits per parameter.

venturebeat.com/ai/how-much-information-do-llms-really-memorize-now-we-know-thanks-to-meta-google-nvidia-and-cornell

Tool of the week: ChatGPT Voice Mode

OpenAI has upgraded ChatGPT’s Voice Mode, adding more human-like speech with natural pauses, emotional tone, and support for more languages. The update brings subtler intonation and realistic cadence, but OpenAI warns users may still hear odd pitch changes and occasional gibberish.

Users can activate Voice Mode on mobile and desktop versions of ChatGPT.

AI-generated image of the week

We’re in full summer break mode here at Innovation Profs headquarters. Don’t forget to pack your sunscreen.

Prompt: A photograph of three bottles of "Innovation Profs" sunscreen lined up on a flawless white sand beach. Each bottle boasts vibrant tropical artwork and playfully displays text like "Sun's Out, Brains Out" and "SPF Smart". A sleek, yellow-and-white striped beach umbrella stands slightly behind the sunscreen, while a pair of retro sunglasses and a delicate pink seashell are thoughtfully placed nearby. Warm, late-afternoon sunlight illuminates the scene, creating long shadows and a serene ambiance with gentle turquoise waves softly rolling in the distance.

What we found

Users are giving ChatGPT their age, body type and DNA results, and asking what they look like. See the results here.