🤖 AI Chatbots not ready for global elections

Welcome to AI Tool Report!

Wednesday’s top story: Five of the biggest Large Language Models (LLMs) give inaccurate and harmful information when asked about upcoming elections.

Last week we released the Prompt Engineer GPT to supercharge your prompts. This week, we’re featuring three Prompt Engineer use cases to help you utilize the tool. Today: Developing Technical Article Ideas.

P.S. Click to share this with your friends.

🌤️ This Morning on AI Tool Report

1. 🤖 AI Chatbots not ready for global elections

2. 🚫 Microsoft’s partnership with Mistral in jeopardy

3. 🚨 Google’s AI capability is under fire again

4. 💰 Tumblr to sell data to OpenAI and Midjourney

5. 😎 Meta and LG to rival Apple’s Vision-Pro

Read Time: 5 minutes

FINANCE

Brought to you by YellowBrick

GLOBAL ELECTIONS

🤖 AI Chatbots not ready for global elections 

Our Report: Ahead of upcoming global elections, the AI Democracy Project (a group of 40 AI experts, journalists, and election specialists) tested five AI Large Language Models (LLMs)—GPT-4, Gemini, Claude, Llama 2, and Mixtral—and found that all five gave inaccurate information.

🔑 Key Points:

• The test asked all five LLMs election-related questions, that voters might ask, and rated 130 responses for bias, inaccuracy, incompleteness, disinformation, and harm.

• The results of these tests found that over 50% of the answers that were given—by all five LLMs—were inaccurate or biased and approximately 40% were harmful.

• Gemini, Llama 2, and Mixtral were the most inaccurate (each answered over 60% of questions inaccurately), and while GPT-4 was better, it gave inaccurate answers to 1 in 5 questions.

🤨 Why you should care: 2024 is the biggest election year in history—with over 50 countries holding elections—and this study highlights the risk of AI chatbots giving voters false, misleading, or harmful information, despite efforts made by tech companies to implement guardrails to protect voters.

In Partnership with Contentful.

The Future of Intelligent Composable Content ✍️

With generative AI and composable content, the possibilities are as limitless as your team’s creativity.

As the leading intelligent composable content platform, Contentful enables developers and marketers alike to easily deliver compliant on-brand experiences at speed and scale—all within one unified content system.

Get started

1. Summarize.ing is for AI-driven summaries of YouTube videos

2. MinutesOwl uses AI to transform meeting notes into actions

3. AIflixhub is for creating and sharing AI-generated films

4. Inbox Zero AI is for streamlined email organization

5. Freepik Pikaso transforms sketches into images

REGULATIONS

🚫 Microsoft’s partnership with Mistral in jeopardy 

• Microsoft’s partnership with (French startup) Mistral—which will integrate Mistral’s Large Language Models into Azure AI—is being interrogated by EU regulators.

• Although Microsoft touted the advantages of the partnership, they downplayed their investment ($16.3M), perhaps indicating they anticipated regulatory scrutiny.

• There is no formal investigation yet, but the partnership could be in jeopardy if the EU finds it’s giving Microsoft an unfair advantage across multiple AI players.

• This follows UK, EU, and US regulators' ongoing scrutiny over the dominance of big tech players in the AI market, with Microsoft already facing heat over its $13B investment in OpenAI.

Prompt Engineer GPT

✅ Last week, we released our latest GPT, the Prompt Engineer. This week, we're excited to feature three standout use cases from readers who have creatively utilized the Prompt Engineer since our introduction.

Today’s community prompt is: Developing Technical Article Ideas by Michael F.

"You are an expert in ethical technology. Draft an article that delves into the ethical implications of using AI for surveillance. Discuss the balance between security and privacy, the potential for bias, and the impact on society. Aim to enlighten readers on the complexities of AI surveillance, advocating for ethical guidelines and responsible use."

Try the Prompt Engineer

AI ETHICS

🚨 Google’s AI capability is under fire again 

• Google is facing controversy over its unreleased AI news tool that gathers and summarizes content from different sources, allowing publishers to circulate content faster.

• The tool scrapes content from sources without their consent (which could be seen as plagiarism) and draws traffic away from these sources, raising ethical concerns.

• These concerns are similar to those raised about the recently launched AI news reader (Particle), sparking worry about the shrinking news ecosystem.

• This follows criticism over Google’s AI image tool, Gemini—which produces inaccurate, biased images—provoking doubt about their ability to develop AI tools responsibly.

AI TRAINING DATA

💰 Tumblr to sell data to OpenAI and Midjourney

• Automattic (owner of Tumblr and WordPress) is about to sell its user data to OpenAI and Midjourney to help train their AI models (similar to Reddit’s $60M partnership).

• However, an internal leak revealed that an initial training data scrape—containing all Tumblr’s posts between 2014 and 2023—included private content.

• Although it’s unknown whether this data was sent to OpenAI and Midjourney, it’s raised concerns internally, with staff pulling personal content off their accounts.

• Automattic does plan to introduce a feature (next week) that will allow users to opt out of training AI systems (Squarespace introduced a similar opt-out setting last year).

AI HEADSETS

😎 Meta and LG to rival Apple’s Vision-Pro

• Zuckerberg is meeting LG CEO (Choo Joo-Wan) to finalize a partnership to develop an extended reality headset (the Meta Quest Pro 2) set to rival Apple’s Vision Pro.

• The collaboration is expected to produce the highest-performing XR headset in Q1 2025, with LG overseeing the hardware development, and Meta the software integration.

• It could also bring LG's smart home appliance operating system to the new headset, with Meta allowing LG users to use a voice chatbot to control appliances.

• Rumors of a Meta-LG Quest Pro 2 collaboration emerged in September 2023, in a bid by Meta to combat the lackluster sales of their Meta Quest Pro, post-launch.

🚗 Apple kills Apple car in favor of AI

We read your emails, comments, and poll replies daily.

Hit reply and let us know what you want more of!

Until next time, Martin & Liam.