SurvAIval Insights #44_24

Good Saturday, my ro-bro fellas!

Fast weekly summary

• OpenAI releases ChatGPT search, a new competitor to Google and Perplexity?

• Meet Botto, an Autonomous AI artist making millions. Will this be our new future or only a passing trend?

• James Cameron warning about AGI. Kind of scary…

🚨 Scroll down for more good stuff!

The chart of the week: OpenAI's new SimpleQA benchmark study reveals troubling accuracy rates among top AI language models.

The findings challenge the growing trend of using AI as standalone research tools, especially among students. With a test comprising 4,326 questions across various fields, the study shows these models should be used as information processors with reliable external data rather than independent knowledge sources.

Their best model achieved only 42.7% accuracy, while Anthropic's Claude-3.5-sonnet reached just 28.9%, though smaller Claude models showed better judgment by declining to answer when uncertain.

The study also found that AI models consistently overestimate their capabilities. When rating their confidence in answers, the models gave inflated scores about their accuracy, highlighting a significant gap between perceived and actual performance.

Read the whole article here.

One way to address inaccuracies and hallucinations is by providing the model with specific data to ground its reasoning. This is where a tool like NotebookLM makes a difference. Watch the tutorial in the video below:

My News Picks

• OpenAI has launched ChatGPT Search. The war is on, comparison here.

• Meet Google’s AI agent: Jarvis

• Botto, an Autonomous AI artist just made $351,600 at Sotheby's

• Universal Music teams up with AI audio company

• Meet the Guys Dating A.I. Bots

• I own my LLM chat history, and so should you

• Inside Elon Musk's Colossus: xAI's World-Leading AI Supercomputer

• New AI image generator Red Panda outshines rivals while origins remain mysterious

My Video Picks

• James Cameron: Special Video Message about AI and AGI

• Bill Gates on possibility, AI, and humanity 

• Wonder Dynamics now lets you go straight from multi-camera video to fully animated 3D scene

Tools and Education

• Dream Lab by Canva is a new powerful text-to-image app powered by Leonardo’s Phoenix model. (Canva acquired Leonardo recently)

• ChatGPT’s web app now has an option to search old chats.

• Suno has rolled out Personas, enabling users to capture and reimagine the essence of songs. Now Taylor Swift has competition!

I hope these insights, news, and tools help you prepare for the future!

Have a really nice weekend and week.

Stay kind.

Rafa TV

🚨 Your feedback is valuable 🚨 Respond to this email and tell me how you think we could add more value to this newsletter. I really appreciate it!