All the squawking in local AI

Open source, local AI is gaining massive traction. As it stands, AI reminds me of an (ageing) African grey parrot.

-It's impressive

Thanks for reading Notes from Jan 👋! Subscribe for free to receive new posts and support the content.

-It demonstrates remarkable mimicry and response capabilities

-Models can have some impressive contextual awareness

In the end, it's all learned patterns and it might be losing its sh*t.

Credit L.Miguel Bugallo Sánchez

So what can you do with a senior citizen parrot? In practical terms, open source AI has shown great promise in *specific* applications. A notable example?

The implementation of PII switching mechanisms for maintaining data privacy.

AKA ”I've scanned this PDF and discovered something that could be leaked- let's move systems to private”

For businesses, open sourced, local AI offers the opportunity to develop proprietary tech. The journey to commercially viable IP has had years shaved off it. This does come with a caveat.

Overlooking the importance of

-data curation (feeding your hungry parrot🦜)

-organisational readiness (*if your team can look after your new pet*)

-the teams’ ability to shift from dev to production mindset

is fatal.

So while open source AI holds significant potential for innovation and creating proprietary technologies, we have a responsibility to approach its integration with a clear understanding of its current limitations as well as the commitment required to navigate its complexities.

No magic bullets.

Do you have an interest in local AI? Have you checked out the best opensource ai yet?

Thanks for reading Notes from Jan 👋! Subscribe for free to receive new posts and support my work.