A lot of people have noted a gap between what LLMs seem capable of doing in influencer-demos and what they're able to get out of LLMs in real-world productive settings.
What we've spent the past year developing is a system that bridges the gap, so you can use LLMs for targeted use cases that require specific libraries and packages. Here's a concrete example of this, with me using @SourcegraphCody to create a pedestrian image detector in 5 minutes that uses a specific library, supervision from the good folks at @roboflow, rather than OpenCV
Technological advances are essential for BNSF to maintain its leadership in the industry, and Roboflow is helping us immediately realize value from state-of-the-art computer vision technology" said Asim Ghanchi, AVP of Technology. "Achieving positive results using AI in a lab environment is easy, but the real challenge comes when scaling the solution across a network like ours without disrupting day-to-day operations. Our partnership with Roboflow is allowing us to do just that.
This is pretty legit - the @roboflow team created a "visual search engine" for NYC.
You can search for anything and the search engine uses AI to find it across all NYC traffic cameras.
It won "The world's shortest hackathon" and they open sourced all the code (🧵). Legends