AI Newsletter: Latest Developments in AI and Tech

Uncensored Models, AI-Powered Hardware, and the Evolving Landscape of Artificial Intelligence

In this week's AI newsletter, we're covering a range of exciting developments across the AI landscape, from new language models and image generators to innovative hardware and ongoing legal battles. Let's dive into the latest news and advancements shaping the future of artificial intelligence.

Grok 2: xAI's Uncensored AI Model

xAI, the artificial intelligence company associated with Elon Musk, has released Grok 2, an updated version of their large language model. Grok 2 is now integrated directly into Twitter (now known as X) for premium subscribers. According to xAI's press release, Grok 2 performed impressively in the LM SIS chatbot Arena, outperforming models like Claude 3.5 Sonnet and GPT-4 Turbo.

Key features of Grok 2:

- Completely uncensored, with minimal content restrictions

- Integrated image recognition capabilities (coming soon)

- Coding abilities

- Image generation using the Flux One model

One of Grok's standout features is its lack of censorship. Users can ask it virtually anything without encountering the usual ethical guardrails found in other AI models. This extends to its image generation capabilities, allowing for a wide range of creative possibilities.

Grok's image generation, powered by Flux One, is reportedly comparable to or better than DALL-E 3, and at a lower cost than competitors like Midjourney. At $8 per month for X Premium subscribers, Grok offers a compelling package of AI features at a competitive price point.

However, the uncensored nature of Grok has sparked discussions about responsible AI use. Examples of controversial images generated by the system, such as political figures in compromising situations, have already circulated online. This raises important questions about the balance between creative freedom and potential misuse of AI technology.

Google's AI-Powered Hardware and Software Updates

Google held its "Made with Google" event this week, showcasing a range of new hardware and software innovations with a strong focus on AI integration.

Pixel 9 Phone

The star of the show was the new Pixel 9 phone, which Google is positioning as an AI-first device. Here are some of the standout AI features:

1. Gemini Nano: The Pixel 9 is the first phone to launch with Gemini Nano, Google's large language model designed specifically for mobile devices. It can generate outputs at an impressive 45 tokens per second.

2. Gemini Live: This feature aims to provide a conversational AI experience similar to OpenAI's GPT-4 with voice. While the demo had some initial hiccups, it showcased the potential for natural, voice-based interactions with AI assistants.

3. Call Notes: This on-device feature can generate transcripts and summaries of phone conversations, all while maintaining privacy by keeping the data local.

4. Pixel Screenshots: An AI-powered organization tool that helps users easily find and recall screenshots based on their content.

5. AI-Enhanced Camera: The Pixel 9 boasts what Google claims is the world's first AI-powered camera, with features like:

- Automatic upscaling of zoomed-in photos and videos

- Object removal from backgrounds

- "Add Me" feature for seamlessly including the photographer in group shots

Pixel Buds Pro 2

Google also announced the Pixel Buds Pro 2, earbuds designed to work seamlessly with Gemini. These earbuds allow users to have ongoing conversations with the AI assistant throughout the day, reminiscent of the AI companion concept portrayed in the movie "Her."

Google Search AI Overviews

Google has made some updates to its AI-powered search features:

- A new save feature for revisiting specific AI overviews

- A "simpler" button to generate more accessible explanations

- Testing of a new right-hand link display for AI overviews

Imagen 3

Google has rolled out Imagen 3, its latest image generation model. Currently available to US users through the AI Test Kitchen platform, Imagen 3 shows promise in generating high-quality images from text prompts. However, it maintains stricter content policies compared to more permissive models like Grok.

Midjourney's New Web Editor Feature

Midjourney, the popular AI image generation platform, has introduced a new feature in its web editor. Users can now directly edit generated images within the platform, offering a more streamlined workflow for refining AI-generated artwork. This update allows for easier adjustments to prompts, selective erasing, and aspect ratio changes without leaving the Midjourney interface.

OpenAI Updates

New GPT-4 Model in ChatGPT

OpenAI has quietly rolled out a new GPT-4 model to ChatGPT users. While the improvements are described as "bug fixes and performance improvements," the update was significant enough to top the leaderboard on the LM SIS chatbot Arena. The new model shows particular strengths in areas such as math, coding, instruction following, and handling long queries.

Free DALL-E 3 Images in ChatGPT

OpenAI has made DALL-E 3 image generation available to free ChatGPT users, albeit with a limit of two images per day. This move allows more users to experience AI image generation without a subscription, while ChatGPT Plus subscribers retain unlimited access.

SWE-Bench

OpenAI has released a new benchmark called SWE-Bench, designed to better evaluate the performance of large language models as they continue to improve rapidly. This benchmark aims to provide a more future-proof method of assessing AI capabilities.

Hermes 3: New Open-Source LLM

Nomic AI has released Hermes 3, a new open-source large language model available in 7B and 40B parameter versions. Hermes 3 is designed to be more aligned with individual users rather than the company that created it, potentially offering less censorship and more customization options. Early comparisons suggest it performs similarly to Meta's Llama 2 model.

AI and Copyright: Ongoing Legal Battles

New AI Image Lawsuit

A previously dismissed lawsuit against Stability AI and Midjourney has been amended and refiled, with the judge now allowing it to proceed. The new allegations focus on Midjourney's use of a "style list" featuring 4,700 artists' names, which the plaintiffs argue implies false endorsement. This case could set important precedents for AI-generated art and copyright law.

SAG-AFTRA AI Strikes and Deals

The Screen Actors Guild - American Federation of Television and Radio Artists (SAG-AFTRA) continues to navigate the implications of AI in entertainment:

1. Game Performers Strike: Motion capture actors for video games are now striking over concerns about AI potentially replacing their work using data from their past performances.

2. Voice Actor Deal: SAG-AFTRA has reached an agreement with a company called Narrative to develop a platform for fair compensation when AI replicates actors' voices.

These developments highlight the ongoing challenges in balancing technological advancement with fair treatment and compensation for creative professionals.

Universal Music Group and Meta's AI Agreement

Universal Music Group (UMG) and Meta have announced an expanded global agreement focusing on AI monetization and creator protections. While the full terms weren't disclosed, the deal aims to ensure fair compensation for artists and songwriters, particularly in relation to AI-generated content and short-form videos. This agreement could pave the way for more creator-friendly policies on social media platforms regarding music use and AI-generated content.

Exists: AI-Powered 3D Game Creation

A new company called Exists has launched a platform that claims to create 3D games from text prompts. The system purportedly allows users to generate game environments, characters, equipment, and gameplay mechanics through natural language descriptions. While the concept is exciting, it's important to approach such claims with a degree of skepticism until the technology can be independently verified and tested.

Conclusion

This week's AI developments showcase the rapid pace of innovation across the industry. From more powerful and less restricted language models like Grok 2 to AI-integrated hardware like Google's Pixel 9, we're seeing AI capabilities become more deeply embedded in our daily lives and creative processes.

However, these advancements also bring new challenges. The ethical implications of uncensored AI models, the potential impact on creative professionals, and the need for updated legal frameworks are all pressing issues that the industry and society at large will need to address.

As AI technology continues to evolve, it's crucial to stay informed about these developments and consider their broader implications. Whether you're a tech enthusiast, a creative professional, or simply curious about the future of AI, these advancements will likely shape the technological landscape for years to come.

Stay tuned for more updates as we continue to track the latest in AI innovation and its impact on our world.