NOCTIVAGOUS AI Facilitation of Work Vs. AI Takeover of Human Jobs Discussing What Should Be the Philosophical Aim for AI Technology, Using The Example of a Vegetable-Cutting Kitchen Appliance Driven by AI noctivagous.github.io A vegetable-cutting appliance for the kitchen that will cut or dice any vegetable placed inside its box would be of immense value to the cook. This type of device would have been challenging to construct before recent advances in AI, as the machine has to be able to recognize every type of vegetable placed inside it and, on its own, know how to peel or cut accurately with its cutting equipment. Many aspects of this are achievable in a straightforward way by training a neural network. Whether it is the recognition of the vegetables or how they are to be cut, a neural network would be trained for the device to process all common fruits and vegetables. We can use this hypothetical device to explain how technologists should view AI more often, which is to solve individual problems as they currently exist inside the activity rather than seek to overtake entirety of a human's role in a job (e.g. no more need to produce videos, as the AI does it all). A full-sized home cook robot won't be affordable to the average consumer for a long time and it will require lots of testing to consistently perform each cooking stage. In the long term a humanoid is versatile, but a vegetable-cutting device is already known to be achievable as a consumer product and it would be utilizable by the public in some initial form in a few years on a scale similar to a microwave oven, cutting across all social strata. Engineering specifics aren't needed immediately, but it could incorporate a tumbling mechanism, used currently by a certain industrial vegetable peeling machine. The dimensions of this product would be close to a large microwave oven with a price that is affordable for the consumer. It would be usable by anyone interested for convenient, at-home cooking, including people who rarely cook today but might just want to put a chopped tomato on a salad or slice a potato to make some air-fried chips. Unlike products that place AI in the kitchen for the heating stage of cooking, a category that is already quite satisfied without the inclusion of AI functionality, the purpose of this appliance is for a less glamorous stage of preparing food: the cutting, slicing, and peeling of vegetables. It would be harder to put this appliance together to process every type of vegetable without making use of AI, unlike the cooking appliances that currently perform well for baking and cooking without it, as they don't need to know the exact form of the food placed inside the oven nor is the food coming into direct contact with mechanisms. An appliance with this goal, once completed, would take the world by storm like the microwave oven, it would be driven by AI, but it would not be a robot doing everything to make a meal. It's a category of kitchen appliance that might seem arbitrary to those who are unfamiliar with home cooking. But many others, who cook regularly and also manage other responsibilities, know that cutting vegetables is frequently an involved process, even if the person has no objections to doing it as a chore. Whatever is happening in life comes to an abrupt halt and becomes tedious vegetable peeling and cutting. Nowadays, the food preparation stage is more in need of being addressed by a kitchen appliance than the heating stages of cooking, as air fryers and pressure cookers are now well-established and working well for anyone who uses them. The microwave oven is now basically indispensable to modern life but it can be used much less if regular people can get cut peeled and vegetables on demand. Addressing the issue of feasibility, this consumer device surely could be engineered to cut just about anything placed into it in a few minutes when there has been so much development for processing food in industrial settings. Our device would provide the convenience of a microwave oven but for the task cutting vegetables— put a vegetable or two inside and out comes a chopped, sliced, or diced result. It would not necessarily focus on quantity to begin with but instead addressing the chore itself. It isn't difficult to chop a bunch of celery. But if it were possible to place the celery inside the appliance and wait a few minutes to have chopped pieces, this would be a big deal for home cooking and later commercial restaurant scenarios. When the vegetables are taken out of this AI-driven kitchen device they will be ready for cooking meals but also assembling cold dishes, such as fruit salads. A potato placed inside can be chopped into the shape for fries, then placed into a nearby air fryer. The machine also cuts up the scraps and places them to the side of the cut vegetables for easy composting, and because of this the device does not need to dispose of anything through a plumbing attachment and it is self-contained. It would just have a cleaning mode after being provided soap and water, also executed by AI. All of this would be far-fetched in a time before recent advances in AI. But a larger point is that the device does not attempt to take over the entire process of cooking, pushing the human out of the kitchen. It instead directly addresses an issue that cooks encounter: vegetable cutting is the main obstacle to regular cooking at home, because even if people enjoy doing it or they don't mind the chore, it can become a hassle that cannot be easily overcome at all times. Professional chefs are no always working in the restaurant kitchen cutting vegetables, either, so of course this could reduce the time to prepare food in other industries. That very fact points to how valuable such a device would be across many environments. The extra work that vegetable cutting creates for cleanup often takes too much time for a home cook to balance other chores, but at times a bowl of two chopped tomatoes might be all that is needed. Following the conventional approach for making a kitchen appliance that incorporates AI for cooking, it will be an expensive robot that has moving arms (moving around outside of a contained appliance box) and it will be able to cook on a typical stove what it cuts too, perhaps roaming around the stove and kitchen on a track. Apart from requiring far more engineering and AI technology than can be affordable for the average consumer today, this robot setup is unlikely to be manufactured for the consumer within a normal timeframe, and it will be an experimental product for a long time, limited to restaurants, some of whom might not even want such a robot until it is well-established. The size of the device will be large, it will take up a lot of space, the assembly will likely require a professional's assistance, safety may be an issue with its large rotating parts and arms, and it will not necessarily fit in an average household kitchen. This is not even addressing the software side, which is that there isn't AI advanced enough at the current time to drive an autonomous domestic robot for multiple tasks. Will the humanoid move with a person when he changes living settings? Will it fit in the back of a car? It may not be considered worth it by most homeowners for a long time, just to have an extra cook, because of all the downsides. Thus this second approach, this takeover of the whole process by making a partial humanoid that roams across the kitchen, takes a lot more development time and engineering resources and is an inferior way to utilize AI in a kitchen appliance for our times-- if we want AI to be of use to us in the near future for real life. A chef robot would take a speculative research track, promising to change everything while always needing more time and money to be finished, with its success contingent on breakthroughs that are not guaranteed to be timely or satisfactory. It also stirs a person inside, as the entire cooking activity has been taken over completely-- supposedly for the purpose of making everyday life easier. Though it makes life easier in that way, it doesn't have to do that, as we just demonstrated, because only one part of the cooking process is currently causing significant obstacles for people at home. The approach of what to use AI for matters: the chef robot may not solve problems for regular people whereas the vegetable-cutting appliance certainly will, today. It is still not known yet how the public will receive the self-driving car, whether people will want this immediately, in a few years, or whether it will be avoided and result in the video of the scared person with his back against the back seat. Will people get used to this? So much is unknown because so few people have tried this technology. No one can quite say how a product that is so unusual like this will be received outside of the early adopters and enthusiasts and this is the most important part, that it requires investigation into audience reception and then revision in response to how people perceive it and what they want it for. If it is just placed into the public with the assumption people will want it, it is hard to say what will happen. The AI community can stray into areas where the technology does something very powerful, like generate an image from a text prompt, but in the situation the technological gain isn't something the average person cares about for long because it stopped at the initial demo stage and user research and revision wasn't undertaken. The goal of an AI device should not be to make life easy in itself, to replace an activity to the point that no human effort has to be expended but instead fix specific aspects of existing processes that carry complicated aspects, especially those that are causing problems because they are missing pieces to their puzzle. When modernity developed it left gaps in certain areas of the chores that it made simple or convenient and so those valuable conveniences everyone enjoys can make the rest of the chore difficult and sometimes imbalanced. For home cooking, it's the vegetable cutting stage that needs to be made a cinch and then everything else will fall into place. precision and consistency that it can provide at this stage, reducing the fatigue impacting workers for the other stages. Whatever the case, the AI community would start its road construction product by trying to do everything from start to finish, with a very large machine, and it would remove all workers from the situation as a goal. This would be an open-ended research effort as well, like self-driving cars. Instead, it is almost certain that many more workers could be kept while being provided one or more AI-driven machines that take care of the hassles of specific stages that are causing unnecessary delays and wear on the workers. This is the discussion of AI's complete takeover of an entire task versus AI's facilitation of certain stages of the work process. AI takeover is undesirable and is the mindset at AI companies, whereas AI facilitation is basically an absent topic of discussion. Devices made for AI facilitation can exist today, delivered on schedule like the vegetable-cutting machine that incorporates realistic engineering goals, whereas takeover AI, like self-driving cars, will always take a long time and it involves long-term issues of speculation of what people can achieve with the technology. Self-driving cars were supposed to be driving many people around by now with AI (a regularly stated prediction around 2016), at least from the airport and back. Takeover AI machines make lots of promises that cannot be guaranteed to be fulfilled on time, because too much that is unknown is taken on by the engineering research and scope of the machine is too large. Though AGI (artificial general intelligence) is something some people want that supposedly would solve many issues mentioned, it is not available after all and experts vary in their predictions on when it would appear. That is what should be the topic of discussion, that researchers should aim for AI-driven products that address what people need today, not what supposedly futuristic world they wish existed through open-ended, science fiction research goals. To all commuters, road construction takes a frustratingly long period of time, but if AI is supposed to be the means to do it all at once in a single machine we may find ourselves in the same situation as self-driving cars, which still bring with them a lot of unsolved problems and remaining ethical concerns. No one should shy away from what matters in the short-term, in other words, and this is a more responsible use of investment money anyway. This can benefit humanity more today compared to those oversized undertakings that are so speculative. It would be beneficial right now to produce devices that assist road construction for individual problems that make the work take too much time. In the case of the vegetable-cutting appliance that can sit next to the microwave oven, the barrier to AI providing something helpful is extremely low compared to what is extremely high: a big, expensive robot requiring complicated assembly screwed into the wall or placed on a track that roams across the kitchen. It doesn't demand that scientific breakthrough occur at some unknown point in the future, the development of a full AGI that can drive a domestic humanoid safely and autonomously. The purpose of AI has not been confined by the single purpose of a vegetable-cutting device but rather directed into a more rational objective, to deal with what has been identified as inadequate for people's process in life, as it is lived today. This is an AI philosophy that does not aim do it all for people, such as to write the entire paper, to make the entire illustration for them, to prepare and cook the entire meal for them from a predefined setting, to make every aspect of the entire road work, to drive the car from point A to B without a person touching the steering wheel. The consequence is that the outcome will be higher quality, feasible, and controllable right now. This involves going through the effort to identify what is missing first, though. As we said, when conveniences came along for home cooking, they left a gap at the stage of vegetable cutting. Thus the researchers will have to do more than set out to work on technology side and they have to investigate. What is the obstacle in a given activity, such as how the task of home cooking is psychologically and practically hindered by the friction of having to peel and cut vegetables and then clean up the scraps afterwards? What is really holding people back in any given workplace? By comparison, the current long-term aspirations of popular AI enthusiasts are limited, as they aren't willing to examine what is wrong in any given situation, just make a one-size-fits-all AI mega-machine that takes it all over and has been trained on everything people do. What follows is a job that is commonly too big for AI engineering, which is a machine that has to do everything for something and also needs more advanced physical engineering to match the AI software. The key topic is speculation and gambling as to whether the necessary breakthroughs will occur and if some do they will be adequate, in time, and affordable for production. When it is recognized that achieving these types of AI projects is too big of a job, then it becomes a matter of speculation, that there will be AGI (artificial general intelligence) that will server as a kind of panacea for all of the domestic life issues people have, which is not a certainty. Instead what might occur is something uncomfortable and domineering, carrying with it many negatives, something inescapable and harmful to people. People might find themselves more unhappy, in other words, when the chef robot finally exists. They are likely to encounter undesirable side effects when there is an AI machine that makes the entire road. Of course in theory a versatile, personal domestic robot would take care of everything from house cleaning to cooking would be valuable because of how complex life can be, but in practice, given the actual state of science and technology today, that will be expensive for a long time and will not appear as a finished consumer product for many years to come. Because while it may be possible to simulate these tasks in software, it may not be easy to translate to robot hardware. So the first goal is to recognize that there exists this excessive optimism in the tech industry which can never deliver such products like self-driving cars on time (is that a worthwhile goal in the first place?) and admit that for AI to truly serve people it has to be applied to existing situations such that it lubricates the gears of life, not takes by force entire jobs of everyday life. If something helpful makes people uncomfortable, it is following the wrong track and its end point will make people unhappy. Had ChatGPT existed for one purpose, to write code, and OpenAI centered its efforts around that, it would have made a complete product, which would do the job much better than today. It would push AI to develop on the code editor side that presents the code, even placed in the compiler. Though ChatGPT (GPT4 currently) is remarkable in how it is able to write code, and the quality is frequently high, there are often errors anyway. When the devices are too broad in purpose there results a lack of goal clarity, for something like ChatGPT, because it takes a lot of effort to figure out how to use it for everyday life, even though it can do a lot and is very useful for many tasks. It then puts someone in a confused state as to how to use this extremely useful machine that doesn’t have a stated purpose. Usually something like ChatGPT would be shipped as a developed media editor software application, with many tools, categories of use, and menus, so that there is some kind of guidance on how to use it beyond typing everything out inside a box. Because that isn't present, OpenAI has struggled to cross over into the mainstream with its AI product. On the surface ChatGPT would exist to write papers for people from start to finish so that they don't have to do anything, or write entire computer programs for them from start to finish that compile. That is at first glance what its purpose would be for many chores. What is ChatGPT for though? "Whatever you want it to be." Lots of things like that, with some major exceptions added to the discussion. When there is no order in the situation and it is an anything machine, the utility of the machine is reduced to a population of those who are willing to come up with ways to use it all the time. It takes a lot of work to use ChatGPT sometimes, even though the output can be of value. People are not really discussing this. The author is often of the view that he can get ChatGPT to do things in spite of the organization that made it because this is not a finished product. From the technological side this is partly the case as well, as there is the concept in AI research of "emergent capabilities" of a large language model, that many capabilities of the AI were unexpected and not predicted by the research organizations like OpenAI as they went about training neural networks. Their internal workings are sometimes inscrutable too, with the concept of "interpretability" in AI research being a topic of concern of how the AI actually arrived at its conclusions. These subjects remind a person of how sometimes technology companies gloss over what should be in the foreground, such as how VR devices usually cause discomfort after 15 to 30 minutes. Instead, there is basically no discussion of that by VR companies, as if ignoring it in public will make it less problematic for the technology to be of value. AI can write code well because it comes from the same thing, computer code itself. It will be a long time before code is directly compiled into finished programs from AI without error, and in the first place it would then be a new culture of AI-directed app programming (much like how it is today on a small level) instead of truly replacing regular programming. To replace regular programming with something new takes much more than having the AI make chunks of computer code; it requires revising what programming is and modernizing its form. That is the issue which is this is a matter of product and not just engineering. The AI companies say that the GPT is a tool for all kinds of tasks, as long as you can figure out what they are and how to get it to do them. That doesn't sound so great when re-examined, but we can use the AI they have made nevertheless, of course. An LLM certainly can serve quite capably and much of the output is remarkable, but that wasn't really how the ChatGPT was set up, so there is a misalignment of stated product goals and the form that the product takes, which is open-ended and places the burden of using it on the consumer. The consumer should know what the product is for. In technology there is the crucial objective of achieving sensible, beneficial, and intended gains, which are a result of a well-designed and proper course of research, planning, and investment that is aware of what regular people need. These types of products make a useful impact on people's lives and so they catch on, such as the microwave oven or perhaps the touchscreen smartphone. Actual problems that people face are being addressed by them and genuinely useful benefits are provided to the consumer. This is in contrast to enterprises and technologies that end up providing incidental and accidental gains, such as today's VR, as they didn't bother to examine what people want to make life better. VR was predicted by many to take over the computing environment, to be the replacement for everything eventually, but it never attempted to do anything that people do with computers today, like edit 2D videos or 2D photos.  It was going to break off into the future and drag everyone into it and ignore what people are doing every day. VR didn't actually have technical conditions to take anything over that people do on the desktop computer and what it ended up doing was providing an alternative device for immersive entertainment that can be worn for 30 minutes just before discomfort sets in. Along the way it found some side use cases, like training employees, which is the incidental and accidental part. When product failures like this occur it is often a result of ingrained complacency stemming from a false concept that there is such as thing as technologies that don't need to be closely shaped and directed towards benefitting everyday life in a useful way, a kind of "hands off" or free market type of mentality that is actually rooted in laziness, one that will let everything unfold without proactive participation on the part of the organization as long as the fundamental engineering is there. It's a common character flaw found in Silicon Valley in particular, which expects everything to be a success once the hard engineering has been completed.  The thinking goes that OpenAI will be fine because it has the advanced engineering behind the models that drive AI and it doesn't need to make any finished or developed product on its own because surely others will do this with what they make. It remains to be seen if that will occur.