🎭AI's Many Faces!

Reading time: 5 minutes  

 Newsletter | News | Tools | Grow My Tool  

Today we will discuss-

• 🤖AI mimics human behavior patterns

• 💽Spotify launches personalized AI playlist feature 

• 🧙🏻‍♀️Top AI chatbots show political biases

• ⚙️9 amazing AI tools you might not have heard of

All this and more - Let's dive in!

👩‍🍳What’s cooking in the newsroom?

Scientists develop new method that enables AI to mimic diverse human-like traits

👨‍🔬News - Scientists have developed a new method to create various personality traits in conversational AI using a language model and the prisoner’s dilemma game. 

By simulating situations where AI agents must choose between working together or looking out for themselves, the researchers found that AI can learn to mimic complex human behaviors.

👾How exactly did they do this? First, the researchers gave AI agents "genes" that described different personality traits in natural language. But unlike usual genes, the team explored genes representing complex descriptions, such as prioritizing self-interest while being open to teamwork. 

Then, they translated these gene descriptions into strategies for playing a game. The game they used was the prisoner's dilemma, where the AI had to choose whether to cooperate or be selfish, just like people do. In the game, if both players cooperated, they both won points. But if one player was selfish while the other cooperated, the selfish one won more points. If both players were selfish, they both got fewer points.

🔍What was the result? Some agents were found to be selfish, focusing on their own interests over the group's. However, others used advanced strategies that aimed for personal gain while also considering the benefits of the group as a whole.

🤠In conclusion - These findings give us an idea about the qualities AI agents need to have to be useful in human society. On the flip side, they also offer guidelines for shaping societies where both humans and AI coexist, a scenario that is expected to become possible in the near future.

Spotify launches new feature that lets users create personalized AI playlists using text prompts

 🎶News - Spotify is launching AI-powered playlist creation, allowing users to generate playlists using written prompts. Initially rolling out in beta for Android and iOS users in the U.K. and Australia, the feature will continue to be refined in the coming months.

⏯️How does it work? According to Spotify, the feature relies on LLMs  to understand what users want. Then, it combines this with the listener's history and preferences to create personalized AI-generated playlists.

👨‍🎤Here's a step-by-step guide on how to use it -

• To access the new AI Playlist feature in Spotify, open the app and head to the "Your Library" tab. Then, tap the plus button (+) at the top right corner of the screen and select AI Playlist from the pop-up menu.

• Enter a prompt to create custom playlists, such as "songs to serenade my cat" or "beats to battle a zombie apocalypse." Here, it's worth noting that the AI has guardrails around it, so it won't respond to offensive prompts or those focused on current events or specific brands.

• Use prompts referencing places, animals, activities, movie characters, colors, or emojis to personalize your playlist. According to the company, combining genres, moods, artists, and decades will give the best results.

• If you're unsure of what prompts to use, you can try from among the suggestions provided by Spotify, such as "get focused at work with instrumental electronica" or "explore a niche genre like Witch House."

• After the playlist is generated, refine it further by issuing commands like "less upbeat" or "more pop."

• Swipe left on any songs to remove them from the playlist.

• To save your AI playlist, tap the "Create" button to add it to your library.

ChatGPT and the likes of Meta's LLaMA have political biases, study finds

🧑🏻‍🎓News - A study has found that popular AI chatbots have distinct political biases. The study, which was published in the journal Computational Linguistics, surveyed 14 large language models and found that they varied widely in their political leanings.

OpenAI's ChatGPT and GPT-4 were the most left-leaning, and Meta's LLaMA was right-leaning. Researchers also found that the chatbots' political biases affected how they categorized hate speech and misinformation.

🦸🏻‍♀️What were the most interesting observations?

• Left-leaning chatbots were more likely to identify hate speech against minorities, but they were also more likely to dismiss misinformation that was generated by left-wing sources. Right-leaning chatbots were the opposite, dismissing hate speech against minorities but being more likely to accept misinformation that was generated by right-wing sources.

• The study's authors found that AI chatbots may not be as impartial as previously thought. They believe that the chatbots' political biases are likely a result of the data they were trained on. For example, OpenAI's chatbots were trained on a dataset of internet text that is known to be more liberal, while Meta's LLaMA was trained mostly on a dataset of books that is known to be more conservative.

🙆🏻‍♀️What else is happening?

• Canada's Trudeau announces package of AI investment measures

• Microsoft to launch AI hub in London

• US and EU commit to links aimed at boosting AI safety and risk research

• Meta overhauls rules on deepfakes, other AI-generated media ahead of election season

• OpenAI transcribed over a million hours of YouTube videos to train GPT-4 

🐢Interesting facts

• In August 2012, Knight Capital Group experienced a flash crash due to an AI-driven trading algorithm glitch. In just 45 minutes, the glitch executed millions of orders, resulting in a staggering $440 million loss. This incident underscored the susceptibility of AI systems to errors and their capacity to trigger substantial market upheavals.

• In 2018, a robot built by researchers at Nanyang Technological University in Singapore assembled an IKEA chair in a mere 20 minutes, highlighting the remarkable progress made in the field of robotics.

• In 2018, a Japanese man got married to an AI hologram pop star named Hatsune Miku. The relationship led to a traditional wedding ceremony with vows and a real-life priest. The gesture raised questions about the boundaries of emotional attachment with AI.

👩🏼‍🚒Discover mind-blowing AI tools

1. Banner GPT - An AI tool that generates banner images for blog posts (Free)

2. AliceCam - An AI-powered tool that allows users to create professional-looking headshots without a real photoshoot ($11.99 one-time payment) 

3. Virtual CMO - A tool that helps solopreneurs solve any marketing problem in just one minute (Free)

4. Colorify.ai - An AI-powered tool that adds color to black-and-white images (Free)

5. ScriptMe - A transcription tool that allows users to transcribe audio and video files quickly and accurately ($45/month)

6. Taption - Offers automated transcription, translation, and subtitling services for audio and video content ($12/month)

7. eCold.ai - An AI powered tool for automating cold email personalization ($39/month)

8. CardZap - An app that allows users to create and share professional business cards (Free)

9. Launch Teddy - An AI assistant designed to provide specialized support for product launches (Exclusively available to ChatGPT Plus users)

🐣Mythbuster 

Myth: AI can easily transition between different tasks 

Fact: The myth that AI can effortlessly switch between different tasks overlooks some crucial nuances. Transfer learning, a technique where a pre-trained model is used as a starting point for a new task, has shown remarkable potential in adapting AI to different domains. However, the idea that minimal effort is required for seamless transitions is overly simplified.