In partnership with

gpt2 is back in the arena in two versions

The gpt2-chatbot is making a comeback, boasting capabilities that surpass those of GPT-4, Gemini 1.5, Claude, and any other existing models. User experiences shared online reveal that "im-also-a-good-gpt2-chatbot" also solved a complex physics problem that other advanced models, including GPT-4 Turbo, could not handle. Additionally, it uniquely succeeded in a specific coding challenge and an intellectual puzzle, indicating a possible edge in certain analytical tasks. This has sparked comparisons to other AI model relationships, like Opus and Haiku, suggesting "im-also-a-good-gpt2-chatbot" may be the more robust of the two in some respects.

It's exclusively available through a unique testing mode known as 'battle mode' at Chatbot Arena. Here’s how you can try your luck:

1. Visit chat.lmsys.org

2. Switch to Arena (battle) mode.

3. Submit a brief prompt for quicker attempts.

4. After receiving a response, the chatbot's model name will appear at the bottom of the chatbox.

5. Keep trying if you don't immediately encounter 'im-a-good-gpt2-chatbot' or 'im-also-a-good-gpt2-chatbot'. Refresh the page and repeat the process until you do.

6. Check out this video where I got lucky and found both models side-by-side within just 5 tries!

Remember, if you land on one of these gpt2-chatbot models, do not refresh your browser. If you do, you'll have to start the process over to get lucky again.

Related: gpt2-chatbot just created Flappy Bird clone in one-shot

Invest before this company becomes a household name

What if you had the opportunity to invest in the biggest electronics products before they launched into big box retail, would you?

Ring changed doorbells and Nest changed thermostats. Early investors in these companies earned massive returns, but the opportunity to invest was limited to a select, wealthy few. Not anymore. RYSE has just launched in 100+ Best Buy stores, and you're in luck — you can still invest at only $1.50/share before their name becomes known nationwide.

They have patented the only mass market shade automation device, and their exclusive deal with Best Buy resembles that which led Ring and Nest to their billion-dollar buyouts.

Learn how you can become a shareholder.

Microsoft Creates Top Secret Generative AI Service for US Spies

Microsoft has launched a new generative AI service designed specifically for US intelligence agencies, featuring an AI model based on GPT4 that operates in a completely isolated, air-gapped cloud environment. This development marks the first instance of a large language model running entirely offline, as confirmed by William Chappell, Microsoft’s chief technology officer for strategic missions and technology. The aim is to allow the spy agencies to safely process and analyze highly sensitive information without the risk of data leaks or cyber-attacks.

Chappell explained that this project was a challenging endeavor that involved revamping an existing AI supercomputer in Iowa, emphasizing the necessity of keeping the system secure from internet threats. The AI is configured to handle classified data but is designed not to learn from it, ensuring that it cannot inadvertently reveal sensitive information through its outputs. This setup limits access to approximately 10,000 authorized personnel, maintaining stringent control over who can interact with the AI.

The service was activated recently and is currently undergoing a series of tests and accreditations to ensure it meets all required security standards. The CIA and the Office of the Director of National Intelligence, which supervises all 18 US intelligence organizations, are yet to provide feedback on the deployment. Chappell highlighted that the system is operational and capable of functions such as answering questions and writing code, illustrating the practical applications of this specialized AI in intelligence operations.

Meta’s AI tools for advertisers can now create full new images, not just new backgrounds

Meta is enhancing its generative AI tools for advertisers, allowing for the creation of complete new images, not just altered backgrounds. Initially introduced last October, the updated features now enable advertisers to generate entire image variations, which include changes to the product or subject of the ad. For example, a coffee cup in an ad can now be displayed from different angles or in various colors, showcasing possibilities like a lush coffee farm backdrop. However, Meta acknowledges that these tools could be misused to misrepresent products, and has implemented robust guardrails to ensure the integrity and appropriateness of the generated content.

Additionally, Meta has introduced the ability to add text overlays to AI-generated images, using a selection of popular fonts. This feature complements the existing image expansion tool, which adjusts images to fit different aspect ratios for platforms like Facebook and Instagram Reels and Feed. These tools are designed to streamline the creative process for advertisers, helping them adapt their content more efficiently across various media formats. Meta plans to make these AI features globally available to advertisers by the end of the year through its Ads Manager via the Advantage+ creative suite.

Furthermore, Meta is expanding its subscription service, Meta Verified, to several new markets including Latin America and parts of Europe. This service offers benefits like a verified badge, account support, and anti-impersonation measures, with higher tiers providing additional tools for profile enhancement and customer engagement. The expansion also aims to include WhatsApp, enhancing connectivity and support for businesses using the platform.

Other stuff

• OpenAI says it’s building a tool to let content creators ‘opt out’ of AI training

• OpenAI Says It Can Now Detect Images Created by Dall-E—Most of the Time

• Apple announces M4 with more CPU cores and AI focus, just months after M3

• Sperm whale ‘alphabet’ discovered, thanks to machine learning

• OpenAI Is Readying a Search Product to Rival Google, Perplexity

• Consistency Large Language Models: A Family of Efficient Parallel Decoders

• Red Hat unveils RHEL AI and InstructLab to democratize enterprise AI

• Meta AI is obsessed with turbans when generating images of Indian men

• AI Doesn’t Threaten Humanity. Its Owners Do

• Audible deploys AI-narrated audiobooks. Can it replace the human touch?

• Bedrock Studio is Amazon’s attempt to simplify generative AI app development

• Crypto? AI? Internet co-creator Robert Kahn already did it… decades ago

Hope you enjoyed today's newsletter

⚡️ Join over 200,000 people using the Superpower ChatGPT extension on Chrome and Firefox.