Or, maybe it could observe that when you sit down on your couch after 5pm a signal tends to be emitted that turns on the TV. If the AI can duplicate that signal, it'll be able to turn off your lights before you hit the switch. Maybe the AI could listen/look in and see that you getting up and hitting a hue lightbulb switch is correlated to a signal getting sent. Get it to predict the EM signals that are going to be emitted and send them out.įor example, if you have any home automation stuff going on, it is probably sending wireless signals. Here's something I think might be a major stretch, but it might at least be worth look at: a battery powered computer that rides along with you and has a video feed basically equal to your eyeballs and can listen to the surrounding EM environment might have some fun applications. For example idea of an always-on companion AI seem really interesting, particularly if you control it locally - I wouldn't want to give even Siri a constant stream of all my actions, and Apple is a relatively trustworthy company (compared to other major players like Google/Amazon/Facebook). Since you are an experienced coder who is mostly doing this for fun, I think you are right that there are possibilities that you might be able to go after. There's a lot of fiction around JARVIS but I believe there are things that "could" be possible. But, you've got experience and have done research so that doesn't really apply here. IMO it is better to give people an idea of the scope of the difficulty of their tasks so they can decide what they want to do more realistically. People who are just getting started tend to be really optimistic in general, so they pop in with really high expectations. We should probably be more generous with our assumptions, sorry. Honestly, we get all sorts questions here, so some of the responses you are getting might be not be assuming that. Of course I am not that idiot to want to build a military-grade AI robot at home. This is fun programming which I really really enjoy. This is a side/fun project, something just like "I want to be a pilot" for some people. I would also want to program this system to write its own small programs/scripts so it can expand on itself.
I don't care about fancy stuff, just make something VERY basic work. Then I want this service to find patterns in information (possibly NLP here) and the most difficult/impossible part could be to reply to questions.
Jarvis ai doing math pdf#
from various text files, websites, pdf ebooks or images. I want to find a way to feed this machine information e.g. a Raspberry Pi with a core service (brain/heart) that controls subsystems/services (openCV for vision etc.). What I want to accomplish is to setup a dedicated device e.g. I can see the pattern and the algorithm that improves behind that. For example, (I guess that's supervised learning) finding interesting personalized news to show to me, movies or music. Voice recognition, text to voice, some neural network pattern matching for each field but very specifically.
I've researched solutions like Siri and open source ones like M圜roft AI. and ML (finished some books and uDemy courses about Tensorflow).
I've been an engineer for the past 12 years and I've programmed in a few languages and big, demanding web apps. Thanks for your serious and detailed reply. On the bright side, the failure-paths of each of these steps will probably still be pretty good. Maybe Apple would like a general machine intelligence, they don't have as much data as Google, so they might want a machine that is better at abstracting on their limited data. This will be hard, because many of the major machine learning players are major players because their workloads are well suited to the current paradigm.
Jarvis ai doing math software#
So, the steps would be:Ģ) Come up with a revolutionary theory that upends the field.ģ) Become employed by a major company that has the software development resources to implement that theory. It isn't at all clear that the currently popular strategy of convolutional neural networks will ever be able to form a general intelligence - they just compact old knowledge, they don't generate new thoughts (but maybe that's all there is to intelligence. Sometimes very clever pattern matching, but it isn't anywhere near the type of general intelligence that J.A.R.V.I.S. Current "AI" is mostly just pattern matching.
0 kommentar(er)
