• Starter AI
  • Posts
  • GPT-4.5 Turbo leaks, Devin: An AI Engineer, Deepgram’s Aura

GPT-4.5 Turbo leaks, Devin: An AI Engineer, Deepgram’s Aura

Is OpenAI ready to surprise us?

Hello, Starters!

Sometimes a quick search might clear some of our doubts, and other times a simple search might leak announcements ahead of time. Guess which one we're getting close to.

Here’s what you’ll find today:

  • GPT-4.5 Turbo leaks on search engines

  • Cognition presents Devin: The first AI Engineer

  • Deepgram announces Aura

  • Perplexity integrates Yelp suggestions

  • AI is currently used to identify COVID-19 variants

  • And more.

Two search engines have indexed what appears to be an announcement for OpenAI's GPT-4.5 Turbo. Although the page leads to a 404 error, the teaser text reveals enough information to increase expectations.

Apparently, GPT-4.5 Turbo harnesses a context window of 256,000, surpassing its predecessor GPT-4 Turbo. Additionally, it is up-to-date until June 2024, which hints at a potential launch date in June. 

Predictions for the announcement have begun as we're getting closer to the anniversary of the release of GPT-4. Perhaps this little leak is just a strategic move to bring back the spotlight to OpenAI after Anthropic's recent upgrades to Claude.

Cognition is a new AI startup with high ambitions. They have recently unveiled "Devin," the first fully autonomous AI software engineer. Diverging from famous coding assistants such as GitHub Copilot, Devin can manage entire development projects independently, from coding to bug fixing and execution.

This marks a significant milestone for developers, as they can leverage Devin's capabilities to streamline their workloads with the assistance of an AI-powered worker rather than just a coding assistant. Devin is currently rolling out to a selected number of users.

Deepgram has recently launched Aura, a real-time text-to-speech API with lifelike voice models and low-latency capabilities that assist developers in creating conversational AI agents. When powered by LLMs, these agents can serve as substitutes in call centres, enhancing customer interactions while reducing costs.

Aura stands out for its fast rendering of human-like voice models, which prevents the agents from sounding robotic or unnatural, defying the challenges of traditional voice models.

🍽️Following the trend of chatbots serving as search engines, Perplexity has added Yelp's data to its platform to provide users with a wide range of information, including reviews, locations, and details about queries involving restaurants.

🦠Researchers from the Universities of Manchester and Oxford have discovered an AI framework capable of identifying and tracking emerging COVID-19 variants. By integrating reduction methods and a clustering algorithm named CLASSIX, they expect that the framework could assist in managing future infections.

What did you think of today's newsletter?

Login or Subscribe to participate in polls.

Thank you for reading!