> Being a "software engineer" consuming APIs and updating database rows ...
> Only an elite few get to touch these machines.
But lately many can run quite a lot of AI models at home. Doesn't require too crazy of a setup.
Why not build something software fun at home that doesn't involve a DB? Maybe using some free AI model?
I did experiment lately: automatically "screenshot" a browser and ask an AI to find the URL and ask if the URL and site looked like a phishing attempt or not. Fun stuff (and it works).
I tried installing one of these "photo gallery" in a Docker container (where you can put all your family/travel pics and let anyone on your LAN [or on the net] browse them). I saw some of these have "similarity" searches features. I also saw that SAM / SAM2 (Meta's Segment Anything Model) was plenty quick: some people are using these to analyze video frames in real-time. So I was thinking about sending all my family pictures through SAM2 (or a similar model: I saw some modified SAM2 to make it even faster) and then augmenting the "similarity search" by using the results of SAM2. For example finding all the pictures about "pool", etc.
And why limit myself to pictures? I could do family vids too: "Find all the vids where that item can be seen".
Possibilities at the moment seems endless: times are exciting if you ask me.
> Only an elite few get to touch these machines.
But lately many can run quite a lot of AI models at home. Doesn't require too crazy of a setup.
Why not build something software fun at home that doesn't involve a DB? Maybe using some free AI model?
I did experiment lately: automatically "screenshot" a browser and ask an AI to find the URL and ask if the URL and site looked like a phishing attempt or not. Fun stuff (and it works).
I tried installing one of these "photo gallery" in a Docker container (where you can put all your family/travel pics and let anyone on your LAN [or on the net] browse them). I saw some of these have "similarity" searches features. I also saw that SAM / SAM2 (Meta's Segment Anything Model) was plenty quick: some people are using these to analyze video frames in real-time. So I was thinking about sending all my family pictures through SAM2 (or a similar model: I saw some modified SAM2 to make it even faster) and then augmenting the "similarity search" by using the results of SAM2. For example finding all the pictures about "pool", etc.
And why limit myself to pictures? I could do family vids too: "Find all the vids where that item can be seen".
Possibilities at the moment seems endless: times are exciting if you ask me.