Microsoft's VASA, HIPPA-compliant GPT, Brave's New Search, and Portkey

Tools and industry news in a digestible form.

Apr 19, 2024

Morning y’all!

We made it another week. Whooohooo! I think we can all find something that we’re grateful for and that we’re happy we didn’t miss. Have a good and restful week! I’ll be traveling for the next week and a half but the newsletters should continue to land.

Don’t worry.

※\(^o^)/※

— Summer

A free and open source project, it’s like Neuralink for your AI agents. Cool name.

Microsoft just shared news on VASA-1, a new AI model capable of generating strikingly lifelike talking head videos from a single still image and an audio clip. The results are wild and continue to impress me.

A new study from Cambridge found that OpenAI’s GPT-4 model performs to the level of expert ophthalmologists in analyzing eye conditions and suggesting treatments. This is neat because I have some eye issues and so this hits a bit closer to home.

Jamie Dimon is a slug, but, he runs one of the largest banks in the world so we’re forced to listen to his drivel:

AI is already embedded in JPMorgan’s wealth plans, but soon that technology will be even more sophisticated, he said. “All that is going to do is know more about you, learn more about you, look at patterns and, you know, look at successful things in the past,” Dimon added. “AI is going to be a huge aid to things like that.”

Unsurprising is that he’s made it a big part of their plan since that’s where the money is right now.

I thought that this was an interesting implementation. It helps with HIPAA compliancy and other legal matters when it comes to data.

HuggingFace has a new iOS app out. If you spend a lot of time here and especially if you manage a project, this seems like a no-brainer.

Yes, an interview with Mark. Summary here:

Mark Zuckerberg discusses Meta's advancements in AI research, specifically the new version of Meta AI powered by Llama-3. He highlights the model's open-source nature, larger size, and improved capabilities, including real-time image generation and animations.
Zuckerberg also expresses concerns about potential risks from actors controlling closed models and APIs. He reflects on Facebook's history, including the decision to invest in GPUs for Facebook's infrastructure in 2018 and the importance of Facebook AI Research (FAIR) in the company's development.
Zuckerberg also shares his views on the future of AI, emphasizing its potential impact on various industries and the importance of collaboration and open sourcing to mitigate risks and advance the technology.
He then reflects on his experiences, drawing parallels between ancient Roman emperor Caesar Augustus and his own journey. He discusses the value of open-sourcing models in tech, Meta's approach to open-source models and licensing, and the benefits of custom silicon for handling inference tasks.
Zuckerberg emphasizes the importance of staying dynamic as a company, the potential of open sourcing large models, and the potential impact of open-source projects on the world. He also shares his past experiences, including leading Facebook through the development of custom silicon and the challenges of managing large companies like Google+.
Throughout the conversation, Zuckerberg highlights the significance of young leaders making significant impacts and the importance of addressing harms caused by models.

I laughed out loud a bit when he started the comparison with Caesar, but, you go on and do what you want Mark.

From Foundation Capital:

One thing is certain: service-as-software represents a once-in-a-lifetime shift, with AI at the leading edge. This development points to a future where services are not just delivered as software but are constantly learning and evolving alongside us.

Some pretty graphs on money and AI startups. Also, these charts that look at the generative AI market as a whole: