• The Edge
  • Posts
  • AI recreates images from brain activity

AI recreates images from brain activity

see what's on the edge

In partnership with


Howdy, humans!

Here are some useful AI updates and tools we gathered today:

  • Mind-reading AI recreates what you're looking at with amazing accuracy

  • Latest in tech and AI + dev resources

  • Cool AI tools + see who raised funds

  • Prompt of the day and more


Join the live session: automate compliance & streamline security reviews

Whether youre starting or scaling your companys security program, demonstrating top-notch security practices and establishing trust is more important than ever.

Vanta automates compliance for SOC 2, ISO 27001, and more, saving you time and money while helping you build customer trust.

And, you can streamline security reviews by automating questionnaires and demonstrating your security posture with a customer-facing Trust Center, all powered by Vanta AI.

Sponsored Ad



Researchers have developed a new AI that can interpret brain activity to recreate images that a person is looking at, achieving remarkable accuracy by focusing on specific brain regions during the process.


  • AI Capability: The AI can recreate images by analyzing brain activity, specifically using fMRI scans.

  • Testing Method: Three subjects' brain activities were recorded while they viewed photographs.

  • Accuracy: The AI achieved high accuracy, especially with AI-generated images.

  • Implications: This technology could pave the way for advanced brain-computer interfaces and neuroimaging applications.

  • Publication: Findings are available on bioRxiv.

What we think

This breakthrough in mind-reading AI technology holds immense potential for medical and technological advancements, especially in neural interface development. However, ethical considerations and data privacy must be addressed as this technology progresses. The precise recreation of visual experiences from brain activity exemplifies the growing synergy between neuroscience and artificial intelligence.


Figure's humanoid robot, Figure 01, is set to start operations at BMWs Spartanburg plant, marking one of the first commercial uses of humanoid robots in a manufacturing environment. The robot, designed to perform complex and dexterous tasks, aims to enhance automation in car production. This deployment signifies a significant step in integrating AI-powered robots into practical, real-world applications in the automotive industry.

Tencent's AI Lab in Seattle has introduced "Personas," AI-generated virtual characters designed to produce synthetic datasets for training AI models. Utilizing the "Persona Hub," which contains one billion personas, the researchers generate diverse data, such as math problems and logical tasks, potentially revolutionizing AI training data generation. While this method promises efficiency and scalability, it also raises ethical concerns about replicating a model's knowledge base.

The newly launched Udio app, created by former Google DeepMind researchers, uses AI to enable users to generate professional-quality music in minutes by simply typing a description. Backed by prominent investors such as Andreessen Horowitz and supported by industry figures like will.i.am, Udio offers features for editing and iterating tracks, ensuring the music is original with automated copyright filters. This innovative app aims to democratize music creation, providing new tools for artists to easily produce and experiment with their music.


Researchers have fine-tuned the LLaMA model for multi-stage text retrieval, creating two new models: RepLLaMA, a dense retriever, and RankLLaMA, a reranker. These models, tested on MS MARCO and BEIR datasets, outperform smaller models, efficiently handling longer contexts and demonstrating strong zero-shot effectiveness. This approach eliminates the need for traditional segmenting strategies, offering enhanced performance for both passage and document retrieval.

OpenStreetView-5M is a significant open-access dataset with over 5 million geo-tagged street images from 225 countries designed to push the limits of computer vision by testing image localization capabilities.

There are only six core concepts (A records, CNAMEs, MX records, TXT records, NS records, and TTL) needed to understand DNS. This article looks at these essentials, which will cover most DNS needs. It includes a table that contains a complete list of DNS functionality with descriptions.


Discover mind-blowing AI tools

  1. Glasp

    Use ChatGPT and Claude to get summaries of YouTube videos and web articles.

  2. Brevian

    Build custom AI agents and delegate your work without writing any code.

  3. Rapport

    Create, animate, and deploy emotionally intelligent characters to enrich dialogue with your audience.


Discover startups who just raised funds

  1. ZeroEyes

    ZeroEyes raises a $53M Series B Gun detection.

  2. K Health

    K Health raises a $50M Round AI copilot for primary care.

  3. Helical

    Helical raises a $2.4M Seed Open-source platform of bio LLMs.



Cumin lamb noodles

Image 1

Which image do you think is real?

Login or Subscribe to participate in polls.

Image 2


Speak to your favorite characters

Prompt: I want you to act like [insert character] from [insert series]. I want you to respond and answer like [character] using the tone, manner and vocabulary [character] would use. Do not write any explanations. Only answer like [character]. You must know all of the knowledge of [character]. My first sentence is "Hi [character]." Examples: Character: Harry Potter, Series: Harry Potter Series, Character: Darth Vader, Series: Star Wars etc.

You can adapt the prompt to your specific needs.


Share The Edge with your fellow founders and get awesome stuff from our team as a thank you!


How did you like it?

Rate this newsletter!

Let us know what you like!

Login or Subscribe to participate in polls.

If you have any comments or feedback, just respond to this email!

Thanks for reading,

Sam & The Edge team

Join the conversation

or to participate.