• Starter AI
  • Posts
  • ChatGPT’s new feature, Learn more about Tesla’s Dojo, Gemini beats GPT-4

ChatGPT’s new feature, Learn more about Tesla’s Dojo, Gemini beats GPT-4

You can now mention your favourite GPTs like in a group chat!

Here’s to a Monday filled with AI!

Hello, and welcome to Starter AI! A place for people like you and me who are eager to learn more about this technological rollercoaster that we call Artificial Intelligence.

Today’s newsletter will catch you up on the biggest changes since Friday, and - yet again - there are some big changes:

  • ChatGPT has a mention feature

  • Tesla’s Dojo supercomputer heads to Buffalo, NY

  • Gemini Pro surpasses GPT-4

  • Learn about the Meta-prompting method

  • 2024 is the year for pocket AI

  • And more AI news!

OpenAI's plan, in the long run, is to turn ChatGPT into an all-in-one personal assistant, and they're making sure to add the right functions to pursue that goal. Their most recent feature is a great example, allowing you to talk to several GPTs in the same chat. 

It's very simple: just type "@" and then the name of the GPT you want to use. This way, you'll unlock a whole new level of customization in your daily tasks.

Personalization is a key part of OpenAI's roadmap, and Sam Altman has made his intentions very clear in several interviews. What’s a better way to pursue this than having the insights of two specialized GPTs in the same place. Definitely a powerful group chat!

Have you heard about Dojo? This is Tesla's supercomputer designed for processing large amounts of data generated by their vehicles, particularly video data. This capability allows them to improve its AI features for their groundbreaking autonomous driving technology. Overall, it is a significant part of Tesla's efforts to advance in AI.

Elon Musk proudly announced a $500 million investment to build the next generation of Dojo at their Gigafactory in Buffalo, New York. Moving away from Nvidia's reliance, this version is reported to harness Tesla's own AI ASIC chips.

Hugging Face's leaderboard is undergoing unexpected changes as competition intensifies. Google's Bard, powered by the recent Gemini Pro model, is now in second place, surpassing OpenAI's GPT-4 (however, it has yet to reach GPT-4 Turbo's capabilities).

This marks a major shift, as Gemini's powers have been consistently doubted, partly due to a staged video presented during its launch.

What are your thoughts? Will the yet-to-arrive Gemini Ultra manage to surpass everyone's favorite? This is a race worth watching!

📝 A research team from Stanford University and OpenAI has introduced "meta-prompting," an innovative method aimed at improving LLMs' performance by enabling them to break down difficult tasks into smaller, more manageable ones.

📱As the AI landscape continues to grow, ironically, efforts are underway to make it smaller. Reports suggest that 2024 will become the year for on-device training, with names like Apple and Google advocating for models that can run locally on smartphones.

Thank you for reading!