I actually work with ML a lot (at the intersection of my domain with it), though I am not an ML/AI engineer.
I think short-term, ML/AI has a great chance of helping hugely with accessibility issues with users of various systems. My secondary thought is maybe related to elder care, but I’m not sure yet.
I have largely had bad experiences with AI assistants (coding, search, and other domains), except maybe helping with finding/generating code samples for libs/packages with poor or missing documentation (though I go to the docs and code first and those results aren’t always correct).
I do see virtual assistants in various forms being a possible near-term implementation with promise, but most are still heavily trained on and biased to. A handful of languages (in the case of LLMs and such) which limits global appeal.
I am both frightened (the race to market without considering the near- nor long-term costs to society as a whole neither ethics in many cases) and hopeful about the whole thing.
I think you are probably correct, though I also feel we might have something in physics or robotics that has ripple effects opening new avenues. Only time will tell, I suppose. Cheers!
I am both frightened (the race to market without considering the near- nor long-term costs to society as a whole neither ethics in many cases) and hopeful about the whole thing.
Same! I’m downright terrified about the impossibility of determining what should be legal/illegal because our politics move at a glacial pace and rarely involve experts in the field. Some bad actors can AND WILL really fuck things up here. I do think that the possibilities are a net positive. If I were in charge, I would pump the brakes until we could better ascertain what the fallout will be.
I have largely had bad experiences with AI assistants (coding, search, and other domains), except maybe helping with finding/generating code samples for libs/packages with poor or missing documentation (though I go to the docs and code first and those results aren’t always correct).
Maybe my current scenario is in some Venn diagram of the perfect situation, but I’ve been having the opposite experience. I’m a game designer changing from about a decade of Unreal (and another decade and a half of various proprietary engines) to learning Unity. I’ve got a pretty clear idea of what I want to do, I’ve just got no clue how to do it. I’ve been using a combination of GPT-4o and co-pilot to figure things out and it’s been great! GPT has been a combination of pair programming and Google replacement. I’ll tell it what I’m trying to do and it not only spits out a code example, but it describes what each section is doing. Occasionally it’ll tackle problems from the wrong end, like yesterday it was placing a UI element and then clamping it to ensure it was drawn on screen instead of figuring out the proper screen space scale first and properly converting world space to that specific space/scale. But if you’re familiar with the methodology and need help with the syntax/structure, it’s kind of amazing. Co-pilot is SO FAST! I’ve got no idea where a property I’m looking for is being stored and that shit auto-completes (almost always) exactly what I’m looking for, purely based on context or comments. Some times it hallucinates properties that don’t exist, but the IDE calls my attention to that pretty quickly and co-pilot usually sets me on the right path.
I actually work with ML a lot (at the intersection of my domain with it), though I am not an ML/AI engineer.
I think short-term, ML/AI has a great chance of helping hugely with accessibility issues with users of various systems. My secondary thought is maybe related to elder care, but I’m not sure yet.
I have largely had bad experiences with AI assistants (coding, search, and other domains), except maybe helping with finding/generating code samples for libs/packages with poor or missing documentation (though I go to the docs and code first and those results aren’t always correct).
I do see virtual assistants in various forms being a possible near-term implementation with promise, but most are still heavily trained on and biased to. A handful of languages (in the case of LLMs and such) which limits global appeal.
I am both frightened (the race to market without considering the near- nor long-term costs to society as a whole neither ethics in many cases) and hopeful about the whole thing.
I think you are probably correct, though I also feel we might have something in physics or robotics that has ripple effects opening new avenues. Only time will tell, I suppose. Cheers!
Same! I’m downright terrified about the impossibility of determining what should be legal/illegal because our politics move at a glacial pace and rarely involve experts in the field. Some bad actors can AND WILL really fuck things up here. I do think that the possibilities are a net positive. If I were in charge, I would pump the brakes until we could better ascertain what the fallout will be.
Maybe my current scenario is in some Venn diagram of the perfect situation, but I’ve been having the opposite experience. I’m a game designer changing from about a decade of Unreal (and another decade and a half of various proprietary engines) to learning Unity. I’ve got a pretty clear idea of what I want to do, I’ve just got no clue how to do it. I’ve been using a combination of GPT-4o and co-pilot to figure things out and it’s been great! GPT has been a combination of pair programming and Google replacement. I’ll tell it what I’m trying to do and it not only spits out a code example, but it describes what each section is doing. Occasionally it’ll tackle problems from the wrong end, like yesterday it was placing a UI element and then clamping it to ensure it was drawn on screen instead of figuring out the proper screen space scale first and properly converting world space to that specific space/scale. But if you’re familiar with the methodology and need help with the syntax/structure, it’s kind of amazing. Co-pilot is SO FAST! I’ve got no idea where a property I’m looking for is being stored and that shit auto-completes (almost always) exactly what I’m looking for, purely based on context or comments. Some times it hallucinates properties that don’t exist, but the IDE calls my attention to that pretty quickly and co-pilot usually sets me on the right path.