Premium News Recap

TLDR;

Technology:

🧠 ALS Patient Speaks Again

🫡 Cohere Improves Flagship Model

🤯 Nvidia’s Biggest Rival is Physics

Markets:

💰 About OpenAI’s $100 billion valuation

🚗 Uber Invests in Wayve

🫧 Tu Bubble or Not to Bubble

Product:

🤖 Alexa To Feature Claude Model

🤨 Chatbots Hate Me. So What Now?

OTHER PREMIUM CONTENT

New Notion Content

😟 Putting into Perspective Elon Musk’s Game-changer Colossus 6 billion dollar data center. How large will Grok-3 be? (Only Full Premium Subscribers)

🫢 Understanding Cursor’s unique features (All Premium)

🔐 Personhood Credentials (All Premium)

BRAIN-COMPUTER INTERFACES

ALS Patient Speaks Again

Researchers have achieved a breakthrough in BCIs that allows an ALS patient to speak again five years after the illness began.

Amyotrophic Lateral Sclerosis (ALS) is one of the worst illnesses one can develop, as it slowly but surely leads to progressive muscle weakness, loss of mobility, speech and swallowing difficulties, and eventually respiratory failure. Most patients die within 2 to 5 years, with no cure available.

Now, through an invasive technique, four microelectrodes are implanted into the person’s brain. Using AI to analyze the patterns in neural activity, they decode those patterns into words, achieving near-perfect accuracy and giving that person the chance to communicate.

This is important because…

Validating in open research what closed companies like Neuralink promise is really good news, as you can’t help but assume that many of the claims companies that have raised capital make are overblown.

But here we are, decoding words from a man who can’t clearly express feelings or words anymore, having the opportunity once again to communicate with friends and family.

COHERE

Cohere Updates Flagship Model

Cohere, a rare sight amongst top AI research labs as it does not include the development of AGI in its mission statement (thank god), is a company focused on Generative AI for enterprises. Now, it has updated its flagship model, Command R+, with improved capabilities on RAG, tool use, and reasoning.

Specifically, the model includes a compelling capability where it can ingest different tools it could use to solve a particular problem and categorize the actions to be done on each tool (if required), making the overall experience a much better one.

Additionally, it boasts improved RAG capabilities, allowing to retrieve and cite several documents in one single response to ensure the user can access the original source when required. Citations are notoriously error-prone in LLMs, so this could be a big deal in terms of enterprise “chat with your data” use cases.

This is Important Because…

As we’ve discussed several times, enterprise GenAI has had an underwhelming track record since the industry became top of mind with ChatGPT.

Most companies continue to distrust the technology, mostly due to hallucinations, a misnomer—every prediction is a hallucination—for when the model displays inaccurate statements. While it’s unclear how hallucinations will be eliminated, improving the model’s capability to retrieve information from trusted sources (RAG) appears to be a fundamental step for the industry.

However, if I were a Cohere investor,…

Enjoyed the introduction? Subscribe or sign in to continue with this edition.