ChatGPT can now browse the Internet

Welcome to Daily Zaps — Here’s what we got for you today: 🙌

ChatGPT can now browse the internet

ChatGPT has been updated to browse the internet for current information, with links to sources, overcoming its previous limitation of data up till September 2021. Post its original browsing feature launch in May, feedback has led to updates like adherence to robots.txt and user agent identification for better site interaction control. It’s not just able to search, but can also click into webpages to find the most helpful information & links for you. Initially available to Plus and Enterprise users, it will soon extend to all users.

The browsing feature can be enabled by selecting 'Browse with Bing' under GPT-4. Some results below…seems to be working well.

What is the stock price of Microsoft?

✅ The closing stock price of Microsoft on September 26, 2023, was $312.14

What was highest temperature in Rome, Italy yesterday?

✅ The highest temperature in Rome, Italy yesterday was 86°F, recorded at 1:50 pm​.

Writers Guild of America agrees to usage of AI

The Writers Guild of America (WGA) has ended a nearly five-month strike by reaching an agreement with Hollywood studios, allowing writers to resume work from the upcoming Wednesday under new contract terms. A significant issue during the strike was the potential misuse of AI tools like ChatGPT by studios to bypass union writers. The new contract now restricts AI usage in scriptwriting, ensuring it cannot replace human writers or be considered as source material, although writers can individually choose to use AI tools.

Meanwhile, the actors union, SAG-AFTRA, continues its strike, extending it to the video game industry, sharing similar concerns over AI misuse in creative work. The WGA's agreement sets a precedent in establishing boundaries on AI usage in creative fields amidst ongoing legal ambiguities regarding AI and copyrighted material.

Mistral 7B, open AI model

Mistral AI has taken a significant step towards promoting open models by releasing Mistral 7B, a 7-billion parameter model that outperforms all currently available open models up to 13-billion parameters on standard English and code benchmarks.

Mistral 7B, released under the Apache 2.0 license, showcases the potential of smaller models in handling various tasks efficiently at a fraction of the cost of large proprietary solutions. They are actively working on hosted solutions, dedicated deployment for enterprises, and training larger models with novel architectures, with more releases anticipated in the coming fall. Mistral 7B is accessible for use without restrictions, and can be deployed on various platforms including AWS, GCP, Azure, and HuggingFace.

Student guidelines for proper AI use

Two Wharton school professors published a guide on the proper AI use for students. They emphasized the importance of understanding Large Language Models (LLMs) for effective AI use in educational settings. LLMs, trained on extensive content, predict subsequent text based on given prompts, akin to autocomplete features. However, they lack real understanding and often err, necessitating user verification of their outputs. The professors also highlighted the challenges posed by LLMs, including fabrication of information, inherent biases due to training data or human intervention, and privacy concerns regarding data usage for future AI training.

Below are the four recommended ways AI can be used in a the classroom to enhance education.

1. AI as feedback generator

2. AI as personal tutor

3. AI as team coach

4. AI as learner