A nice idea and good luck! My lawn is dead as our local data center took all the water (I'm kidding!).
We do home property and inventory services using AI on photos as well and the key thing we've found so far is that the biggest rival to those features is just people dragging photos into chatgpt and asking away. So the key here is differentiating from that and making something better and more accurate. What we did was to basically build a better and deeper prompt and history, e.g. context is king in a vertical. So that means the other info the user has put about the property, the memory of previous things asked or seen, combining with publicly available property info we already gather - this would make the information more valuable than straight gpt usage.
So what more can you bring to the bare prompt on the photos to help? What can you build in terms of info about the zip, so you do more 'vertical stuff' before the api call.
Great callout, I saw something similar on another site. Essentially instead of just entering your ZIP, you can also enter your estimated lawn size as well as various other parameters to generate a better prompt. I think that would be helpful to build in the next iteration, both to strengthen the results and also differentiate ourselves from GPT. Another thought I had was to potentially build a database of specific, authoritative lawn care information, such as industry journals, textbooks, university extensions, etc from which the GPT must draw upon and reference, rather than using context clues from the internet to try and invent some kind of analysis or treatment.
Nice idea!
fwiw, false positives and all, but the Windows 11 default Windows Security doesn't like it:
`leakless.exe: Operation did not complete successfully because the file contains a virus or potentially unwanted software.`
Will be interesting to see which side of $999 it drops. I'll buy it regardless but the optics (heh) on the high RAM cost issue and the unit price might temper demand a bit.
I'm not sure I agree, but I can see it used as an open-ended interview question. "Is English the new programming language?" It would be a good test if someone gravitates towards pedantry (AIs can speak another common language just as well!) or if they actually get into the difference between prompting and programming, or whether it's at its core an LLM or just an AI based on transformer architecture. Extra points for having it be part of an async interview and interviewees using an LLM to write the answer, and interviewers using LLMs to grade them.
Strange, it works here in Firefox on Linux using the internal GPU (I don't use primusrun for the browser). Normally I'm the first to notice a particularly heavy website or bad FPS in even pretty old games! Wonder how they managed to make it sluggish on a very expensive GPU but get my crappy setup to run it nicely
Odd. My laptop seemed to do fine with a 'NVIDIA GeForce GTX 1650 Ti Mobile [Discrete]' using CachyOS. It could have been a little smoother but it rendered fine. There were a couple spots where it was a little herky-jerky-laggy that maybe needs optimization.
It choked on my M5 Max MBP w/ 128GB of Unified Memory connected via a TB5 dock over 10GbE directly to my router, which is backed by 5Gbit symmetric fiber. My measured average latency to nasaforce.gov was under 13ms.
Also, is there a way to stop the 'Satellite Inspect' dialog from coming up when clicking around the globe? It obscures a bit of the map.
reply