Tweets on Sale for Oncology AI

Plus: Bold moves from Meta and Microsoft

April 21, 2024

Hello Tensor Black Fans!

Want to buy some mean tweets for your AI?

Elon is now selling one million tweets for $5000 on X. Read below to learn what this portends for AI and oncology content in the very near future.

Level up today:

  • Meta Approaches ChatGPT4

  • Mean Tweets on Sale

  • Video Prompting is Here!

  • Gingko Training Healthcare AIs

  • Microsoft’s Post-Oil Investment

Meta released their latest open source model Llama 3 that is now on par with ChatGPT4, even faster when run on Groq servers (try Llama3-70B-8192).

Why it Matters

Many argue that open source models will outpace proprietary AI, simply because more smart people can contribute in an open community than within one organization.

Meta gave this argument a boost this week by releasing an LLM with 70 billion parameters, clearly competitive with ChatGPT4.

Mark Zuckerberg believes the value will come in use of models, not the models themselves. He’s fearless in giving away models that arguably cost Meta tens of millions to train. They’re very difficult and costly to replicate.

Pro Subscribers to X (formerly Twitter) can now download one million tweets for $5000.

Source: Jimmy Kimmel Live

Why it Matters

Generative AI is running out of data.

AI creators need to pull data from YouTube, Twitter, TikTok and other online sources to continually improve and keep up to date.

Elon’s move could cost AI providers upwards of $2.5M per day for a complete feed, nearly $1B per year.

Former Googlers released Reka, a new AI that can accept video prompts and return answers in text and images.

Why it Matters

Video — perhaps immersive video like Apple’s Vision Pro — is the ultimate, high bandwidth interface for AI.

Reka accepts video as part of their prompt, allowing you to ask questions and get summarized answers with locations in time. Imagine doing the same for a colonoscopy, operation, or other patient encounter.

While this is static video today, the results are impressive, just second to ChatGPT in an independent AI ranking.

The cofounder of Gingko announced that they’ll happily train your healthcare AI using their vast database of 8 billion proteins.

Why it Matters

Phenomic AI models need to understand a broad array of data, from proteins, to electronic healthcare records, to blood markers and more.

Much of this data is held in silo’s today, such as Gingko’s massive collection of eight billion proteins and their DNA sequences.

Gingko is establishign precedent for allowing 3rd parties to train on their data, which can greatly enrich a model for a fraction of the cost of acquiring the data independently.

We hope to see other data providers follow suit.

Microsoft announced a 1.5B investment in G42, the AI arm of UAE that drives many of the largest AI projects in the region.

Why it Matters

Middle East countries have a bold vision for their future, where healthcare, AI and other services start to replace oil revenue.

Microsoft is using their treasury as a powerful tool in the competition for regional AI leadership.

Large funds flow from Microsoft, to a third party, who in return commits to buy AI services. This also gives UAE a boost with Microsoft’s technical know-how.

Watch this space as competition heats up!