Chat GPT and Stable Diffusion are really cool, but outside of Alpha Fold, AI hasn't, to me, yet brought a ton of value to humanity.<p>I have been thinking about what would really be a game changer to people's lives to have automated processes.<p>Would it be food production, medical care, or education? Insulin production, maybe?<p>Putting aside if the tools are ready or not, what would be a high-leverage area that, if there were automation at some level, would have a huge positive impact on people's lives?
Financial products. Start with cheaper advice (which mortgage/insurance/pension product should I get given all this context about my life?)<p>But then give an AI access to your bank statements to get personalised suggestions on how to make your money go further.<p>Maybe eventually something like recommending home improvements — if I want to get my kitchen and bathroom redone, what’s the best order to tackle them in? Does the answer change if we’re going away for two weeks in July? And can it organise builders to come in and get the work done?
Food ordering that makes a custom meal for me. “I’m hungry” and based on my dietician/nutritional information it will send an order to some ghost kitchen (maybe it has like 20 standard ingredients used in various dishes) that makes a meal tailored for my requirements.<p>Running low on fiber today? Here’s more veggies than usual. Idk.<p>Or something in the sustainable space. It would be awesome to have AI help us promote reuse and reduce waste by analyzing the stuff we have and also somehow enabling a sharing culture within the community by intelligently allocating used resources.
Up until recently it was software production with ever increasing demands for developers. I don't expect this to stop, software is eating the world (for better and/or worse). This parallels the adoption of telephones where there was so much demand for operators to connected calls that "everyone would have to become a switchboard operator"--then phones got dialers.<p>Software should be customizable by its users to a much greater degree than we produce today.
Software Packaging: creating packages for software products is painful and tedious. Packaging for Linux requires a lot of work. (See snaps, flatpaks or appimages)<p>CICD pipelines<p>Robustness of software
We should stop going on about this sham of “AI”, and discuss Behavioral Intelligence.<p>The question you might ask is “what intelligent behaviors do we need?”<p>I think modern ML is suitable for building a tricorder like devices. Color test paint, check wear through resonance, interpret x-rays or infra-red patterns.<p>A self balancing hammer?<p>Novelty in application is endless. Knowing what you want before seeing it is hard.