AI is eating the internet 😋

Microsoft’s AI Fixes Its Own Mistakes—Is This the Future?

Microsoft is taking AI safety to the next level with its new correction feature. This latest tool not only flags AI-generated errors but fixes them in real-time. All in a world where AI “hallucinations” are a growing concern.

So, how does it work? Available in Azure AI Studio, this correction feature scans AI outputs, comparing them against a company's source material. If it detects an error, it highlights the mistake, explains why it's wrong, and automatically rewrites the content before the user even sees the faulty output. 

• For instance, if AI were to misinterpret a medical diagnosis, suggesting a patient has a common cold instead of the flu, this tool would flag the discrepancy and rewrite the diagnosis to match the correct medical records—ensuring the right information is provided. 

Why it matters: AI hallucinations have become infamous for producing wild, incorrect information. We’ve seen AI confidently misspell “strawberries” and offer flawed legal advice. This tool from Microsoft could change how businesses and users interact with AI, helping to prevent errors before they cause real-world issues. 

While Microsoft admits it can still make mistakes, it’s a leap forward in improving AI accuracy. With the rise of generative AI hallucinations, other fact-checking products, like Google’s Vertex AI, are also in development.

AI2’s Open-Source Molmo Challenges the AI Giants

This week, the Allen Institute for AI (AI2) released Molmo, a multimodal model proving bigger isn’t always better. In an AI landscape dominated by giants like Google and OpenAI, AI2 developed Molmo to show that open-source technology can deliver comparable performance at a fraction of the size and cost. In other words: Molmo is small but punching well above its weight.

The secret sauce? Instead of relying on huge datasets, Molmo uses smaller but higher-quality data to train its models. While most models are trained on billions of images, Molmo’s dataset is a curated set of 600,000 images. 

• This focus on quality over quantity allows it to deliver fast, accurate visual understanding, from identifying objects in images to answering complex questions about what it sees. It’s like teaching an AI with precision tools instead of drowning it in information.

Generative AI Adoption is Blowing Past Early PC and Internet Growth

Is generative AI spreading faster than the internet? A new study says yes—and the speed is mind-blowing. The adoption of generative AI has far outpaced personal computers and even the early days of the internet, fundamentally changing how we work and live in record time. 

By the numbers: Over 60% of companies now report using generative AI, a huge leap compared to the 40% adoption rate of PCs in the early ‘90s. On the personal side, more than 40% of individuals have integrated AI into their daily routines, outpacing early internet usage by 10%. AI is becoming part of life, whether you’re at work or at home.