Wow.
-
Wow. The new Loops For You algorithm is working better than expected!
I still need to tweak a few weights and parameters, we may need a whole documentation page to explain it, but I'm very impressed with the initial results.
Best part: the magic is SQL, no ML/AI is used (yet), meaning it won't require any additional setup or dependencies for admins.
Stay tuned
-
Wow. The new Loops For You algorithm is working better than expected!
I still need to tweak a few weights and parameters, we may need a whole documentation page to explain it, but I'm very impressed with the initial results.
Best part: the magic is SQL, no ML/AI is used (yet), meaning it won't require any additional setup or dependencies for admins.
Stay tuned
@dansup "yet"....?
-
J jwcph@helvede.net shared this topic
-
@dansup "yet"....?
@jwcph yeah, I'm leaving the door open for improvements like content embeddings and more.
It will be optional of course, the foundation of the For You algo is SQL based.
-
@jwcph yeah, I'm leaving the door open for improvements like content embeddings and more.
It will be optional of course, the foundation of the For You algo is SQL based.
@dansup OK - just a thought, though: Considering the audience, I would be very careful with wordings that can in any way be (mis)understood as "keeping the door open" to AI.
Yes, people around here know it's a market-appropriated term for something that has existed for a long time & may have many useful applications, but they're also deeply wary of the creep that is happening from the techbro version to everything else.
-
@dansup OK - just a thought, though: Considering the audience, I would be very careful with wordings that can in any way be (mis)understood as "keeping the door open" to AI.
Yes, people around here know it's a market-appropriated term for something that has existed for a long time & may have many useful applications, but they're also deeply wary of the creep that is happening from the techbro version to everything else.
@jwcph that is a great point and I will keep that in mind going forward when choosing how i word this! I appreciate the feedback.
-
@jwcph that is a great point and I will keep that in mind going forward when choosing how i word this! I appreciate the feedback.
-
@grosser_mensch @dansup - or even better, describe the benefit & how it works, along the lines of "We give you X by doing Y".
Almost nobody cares if it's technically categorized as machine learning or artificial intelligence or whatever - they just want to know what the value is & that it's achieved without screwing them over.
-
@grosser_mensch @dansup - or even better, describe the benefit & how it works, along the lines of "We give you X by doing Y".
Almost nobody cares if it's technically categorized as machine learning or artificial intelligence or whatever - they just want to know what the value is & that it's achieved without screwing them over.
@jwcph @grosser_mensch @dansup Been struggling with that too. My product has a heuristic learning model and describing that without it soundling liek 'AI' nowadays is tricky. I see it as ml(non model)->ml(model like distill / bert) -> llm (traditional 'ai') but the market sees only the last as any kind of intelligence and doesn't grok ML AT ALL never mind the gradations.
-
@jwcph @grosser_mensch @dansup Been struggling with that too. My product has a heuristic learning model and describing that without it soundling liek 'AI' nowadays is tricky. I see it as ml(non model)->ml(model like distill / bert) -> llm (traditional 'ai') but the market sees only the last as any kind of intelligence and doesn't grok ML AT ALL never mind the gradations.
@scottgal @grosser_mensch @dansup True - but nor should they have to. People shouldn't need to understand the engineering of a combustion engine to trust that their car won't suffocate them to death with fumes or explode while driving down the highway - or indeed to opt for an electric.
