A question about what is possible with a computer

[Planobilly]

I assumed that my home computer would be a lot more useful at this point. Perhaps because I don't know how to use it and perhaps the software is just not advanced enough.

Perhaps this is sort of stupid but I want my computer to do these things.

Recognize me when I walk in the room.

Say good morning and ask how did you sleep. Log my answer

Check my vital signs and log them

Remined me of the day of the week and month

Advise of any appointments that I have logged and incoming information from my medical team.

Read my email and sort through the crap and say you may want to take a look at this and get smarter based on what I flag as useful.

I don't want to type or move a mouse around. The computer could send me a text but I probably would not respond.

I assume all of this is possible but have no idea of the programing involved.

I am a little to old to be learning C+...lol

Billy

 
 

[bob91343]

Perhaps if you learned what computers actually can do, rather than what you wish they could do, you might get closer to your goals.

First, a computer is an electronic machine.  For it to know your vital signs or ask you questions, it must have hardware inputs and outputs.  For instance, body temperature measurement requires a temperature sensor attached somehow to your body.  Blood pressure is more complex but of course it can be done with the appropriate hardware.  Speech recognition and synthesis are possible but require programming and microphones and speakers.

What you are looking for, perhaps, is the WiFi that has become popular.  But it takes a lot more than just a computer.

[guenthert]

I assumed that my home computer would be a lot more useful at this point. Perhaps because I don't know how to use it and perhaps the software is just not advanced enough.

      There's been a widening schism ever since the smartphone was introduced.  You, like most people, seem to be better served by one of those; applicability of PCs, regardless of their CPU's relative performance to the one in the 'phone, seems to be stuck in what they were applicable for in the '90s.

Perhaps this is sort of stupid but I want my computer to do these things.

Recognize me when I walk in the room.

     While not impossible per se, I haven't seen software for PCs which does that.  They might recognize (e.g. via Bluetooth) the phone you're carrying.  Smartphones recognize your fingerprint (or one very similar at least) or face.

Say good morning and ask how did you sleep.

     You mean, plain mechanically?  That's possible for fifty years or so.  Or do you mean actually recognizing you and recognizing you as a person and greeting you as such?  Not yet, not in our lifetime and quite possibly never.

Log my answer

     That's surely possible, but I'm unaware of software which does it.  Shouldn't be all that difficult and I wouldn't be surprised if it already exists.

Check my vital signs and log them

      At least for smartphones / smartwatches there is something approaching such.

Remined me of the day of the week and month

     A calendar application?  It's standard on smart phones, but MS Windows and Google (and I'm sure Apple and others) offer similar .

Advise of any appointments that I have logged and incoming information from my medical team.

     With any luck your "medical team" sends appointment in ICS format, which gets added automatically into aforementioned calendar.

Read my email and sort through the crap and say you may want to take a look at this and get smarter based on what I flag as useful.

     Well, we have spam filters since the nineties ...

I don't want to type or move a mouse around. The computer could send me a text but I probably would not respond.

     Is the computer supposed to read your mind?

I assume all of this is possible but have no idea of the programing involved.

I am a little to old to be learning C+...lol

Billy

     Few people program their own computers.   
   

[T3sl4co1l]

A lot of these are possible, or a reasonable facsimile; it's more that there's not nearly enough market to serve it.

For starters, consider accessibility tools like text-to-speech and voice recognition.  These are available in the operating system, or with affordable add-ons.

Voice recognition has historically been very janky (interestingly, it's been possible since the 90s at least), but in the last decade or so it's reached accuracy good enough for mainstream use.  Though the best ones (Siri, Alexa, Google) still "phone home" to a remote server ("the cloud") -- where extra processing power is available to generate the best results ... and also log your conversation for posterity.

I'm not sure what capability is available on PC, without all those privacy issues, and give or take at what price point.  Probably something still usable and not very frustrating?


Image recognition is definitely there.  Recognizing humanoid figures in video feeds for example has, what, >95% classifier accuracy?  With a bit more training, it can probably identify you personally, too.  It's... not trivial to set up, but can technically be done in the comfort of ones' own home, given the skill to do it; the tools are free and out there(!).  Perhaps there's a precooked tool that can do a more streamlined/automatic build given a training set?  The catch is, I'm not sure you can train such a system on say, just a few portraits -- usually it takes thousands of examples.  (Or minutes of video -- amounting to as many still frames!)

Or something of a shortcut can be used, like biometrics (fingerprints, facial dimensions, eye scan, etc.), if you don't mind.  There are tools already on the market, for example fingerprint scanners for automatic login.  Perhaps this would be an adequate substitute -- your computer might not recognize you on sight because it's blind, but it can recognize your touch (distinguish fingerprints), much as a similarly blind human might, eh?


The rest is much as others have said above.  Note that, getting meaningful interactions, is very, very far from trivial; what happens if the camera misidentifies you?  Should it identify at high frequency (i.e. categorize each individual frame of video -- that may seem overly aggressive, but this is the lazier way to program it, y'see), or build a list of possible candidates ranked by confidence?  Should that list be updated when so notified ("good morning Bob!" "I'm not Bob I'm Jerry." "Oh, I'm sorry Jerry. How are you this morning?")?  What if someone else walks in and starts talking, should it mind its own business, should it welcome them too, --- you see, it very, very quickly spirals into unimaginable complexity, so it's not at all easy to do.

On the matter of convincing conversation, computers are getting there -- but it's still very much based on machine learning as far as I know, and as soon as you stray from the beaten path, they make very jarring errors, such as obvious mischaracterizations of the objects of sentences, or coherence errors seemingly trailing off into irrelevant subjects.  And if you expect them to initiate small-talk, too, good luck...  Grammar at least is pretty well solved I think, no small feat given the sheer complexity of natural language.


As for spam filters, probably a lot of this is already going on transparently and you don't even realize it!  Mail systems have quite a lot of filtering already.  This is a tricky one, because, do you really trust an algorithm -- or worse still, an utterly inscrutable machine-learning system -- to identify important emails, and discard fluff, at nearing 100% accuracy?  What if you get an important medical document, or job notice, or legal service or something like that, and it drops it?

Actually, let's not fool ourselves... humans are far from 100% accurate at most things they do.  I wonder what miscategorization rate would be considered embarrassing for a secretary?  Does it matter that the computer can't accept blame, nor give an apology, when caught?


To summarize, you basically want a computer secretary, right?  There are definitely some things you can do right now, which will make things easier in that regard.  We're certainly not there yet, overall; so if you truly demand the full suite of capabilities that only a human can provide, I'm afraid you have little choice but to hire a real one.  Which, you definitely can, they're available for rent in your nearest classifieds...

Or, you can hire a modest sized team of very smart humans, to make your very dumb computer, somewhat less dumb -- but I'm guessing that's even less financially viable.  Again, the ultimate problem is: not so much whether it's possible, but whether it's worth creating.  Alas, it seems this particular combination of problems just hasn't been judged profitable yet.

Tim

[james_s]

Is this a joke?

Shows like Star Trek are great entertainment, but they are pure fantasy. Computers only do what you tell them to do, some of these individual tasks can be performed to some degree but your vision of what a computer can or should do is totally unrealistic. Interactive assistants have been all the rage for a few years now, but they are totally reliant on powerful server farms to process the requests and in my experience they are still not really very useful, you don't really interact with one, you have to learn exactly the right commands to give it to get it to do what you want it to do, pretty much like any method of interacting with a computer. They are not intelligent or self aware, code executes in predetermined paths.

[Nominal Animal]

At this point in time, it is much more cost effective to hire a human personal assistant or two or three to do this.

[Bicurico]

Why not just use Google Home with a Google Nest Hub?

They will do most of what you are asking for, plus you can just talk to it and give orders, like turn on the light, show me surveilance CAM at the front door, play some relaxing music, schedule the restaurant for tonight, etc.

They also do sleep analysis right now.

[Ed.Kloonk]

Say good morning and ask how did you sleep. Log my answer

Check my vital signs and log them

For anyone who keeps waking up feeling dusty and are wishing they didn't, consider looking online for Sleep Monitoring or more specifically sleep apnea/ sleep apnoea monitoring.

Newer technology has emerged that allows for personal wearing at home and it connects via bluetooth compiling a report for a doctor to see who can contact you via the phone and direct you toward the next course of action.

The previous well established method of staying the night at a sleep monitoring facility is still quite good, however the results can vary because many people prefer to sleep in their own bed, problems not withstanding. So the in-home solution is a major breakthough.

If you find yourself spending time worrying about your vitals, go see a doctor.

[brabus]

Pretty much any smartwatch can do these things nowadays.

[Planobilly]

Hi Guys,

Thanks for the many responses to the question. No, I was not asking the question in a joking manner. Some of the information I was looking for exist currently. Obviously external sensors have to be connected to the computer. Voice recognition software exist and has gotten better over the years. Remote temperature sensors are being used in some airports as a function of the pandemic. My computer can currently do facial recognition currently but it sometimes does not work very well. My desire to have vital sign information is more research based than need basses. 

Actually my Apple iPhone does several of the functions related to vital signs via Bluetooth from a heart rate sensor and other sensors. It is also very useful to respond to certain types of questions such as convert electronic values.

It appeared that "smart phones" are going more in the direction I have an interest in.

It is also obvious that  that there is a cost benefit question. Humans are much better at most things than computers and much less costly to employ than the cost of developing non trivial software.

In some ways, what I am asking for is under serious development in robotics, particularly in Japan for the purpose of helping to care for elderly people.  We may be a long way from effective email sorting and Scotty is unlikely to to beam you out of here anytime soon but change is coming.  Perhaps more in smart phones than PCs. Most of the information I am asking for exist in a digital form at present and can be acquired with a keyboard and a mouse or by turning on a app on my phone. A few apps are now actually speaking to me on my phone to inform me of the data they are collecting.

Thanks,

Billy
 

[Melt-O-Tronic]

You can get an Nvidia Jetson and a decent camera for it and roll your own facial recognition program to trigger whatever you want.  You'd have to spend some time behind a keyboard, but building is 105% of the fun.   

When I get a new device, I first disable Cortana / Siri / Alexa / whatever, but my employer has Cortana sending us a "daily briefing" every morning.  It reads our messages and includes things in the briefing like "yesterday, you told <person> that you would <whatever>" or "4 days ago, you asked <person> 'do we have a purchase order?'"  It seems to be smart enough to infer whether things were resolved and not remind if they are.

This morning, it didn't find anything that needed follow-up, but it saw my calendar and told me that I only have 8 hours of unscheduled time next week.  It suggested several things to do with that time such as blocking it out for a mental recharge, catching up on work, doing some online learning, etc.

Maybe that's what you need.

[rstofer]

I suspect that each of the items listed can be done today.

I'm just starting to mess around with OpenCV 4 and there are a lot of examples of face and eye detection with a hint of facial recognition.  Google for 'opencv 4 facial recognition learning'

https://www.pyimagesearch.com/2018/09/24/opencv-face-recognition/

I didn't say it was easy, just doable.  Clearly, it's a lot more doable on television.

There's a lot of work being done on the subject of deep learning and there are some pretty fantastic libraries available for Python.

As to machine learning and reading email, most of that already exists.  I get a lot less spam since the ISPs have been filtering out the obvious junk mail.

To think it all boils down to that class in Linear Algebra that was such a PITA in the sliderule days...

[rstofer]

You can get an Nvidia Jetson and a decent camera for it and roll your own facial recognition program to trigger whatever you want.  You'd have to spend some time behind a keyboard, but building is 105% of the fun.   

I am playing with that too!  It's a neat little gadget with 128 CUDA cores for parallel processing.  I can buy a graphics card with over 10,000 cores.  We're talking multiple teraflops of compute power.  The CDC 6400 super computer (of the day) that took us to the Moon could only do 2 megaflops.  We've come a long way.

It all comes down to Linear Algebra and matrices.

There is a 'formal' curriculum at NVIDIA (costs a wee bit of money) along with free blogs that provide guidance on using the Jetson Nano (and the other systems).  Who would have thought that a graphics card builder would be providing the tools for artificial intelligence.  If you need a lot of computing in a small space, NVIDIA has a solution - for specific types of computing.

[Planobilly]

Again, thank you guys for all the feedback and some new ideas. I am retired or retarded or something...lol so I have a lot of free time to experiment. The information that you guys have provided is very useful.

I have a electronics lab and a small CNC machine shop here at home and a bunch of crazy programmer friends. We are always in the middle of some pretty far out projects like life support for human organ transportation. I specialize in high voltage tube electronics...fun stuff if you don't kill yourself...lol


Thanks,

Billy

[R-star]

Billy,

You sound almost exactly as my good friend Randy.
He once told me almost exactly the same thing he wanted.
Like, almost word to word. You're not weird and not alone.

I hate to sound this way, but this is not what the computers are actually for.
On a good side, this is why there will never be a true AI.

[CatalinaWOW]

Surprising how many people have said that all of this is hard, impossible or not marketable.

Google Home and Amazon Alexa do almost all of this now, and do a pretty good job.  The things they don't do is take vital signs (though they can with a tie in to accessories like Fitbit) and go through your email for you.  Your email service does actually do a fair amount of that as anyone who examines their spam folder can tell you.  The mail programs I have used reject about 75 to 80 percent of spam, with a very low rejection rate of "good" email.  What they aren't very good at is sorting the remainder into personal communications, billing notices and the like.

[james_s]

Surprising how many people have said that all of this is hard, impossible or not marketable.

Google Home and Amazon Alexa do almost all of this now, and do a pretty good job.  The things they don't do is take vital signs (though they can with a tie in to accessories like Fitbit) and go through your email for you.  Your email service does actually do a fair amount of that as anyone who examines their spam folder can tell you.  The mail programs I have used reject about 75 to 80 percent of spam, with a very low rejection rate of "good" email.  What they aren't very good at is sorting the remainder into personal communications, billing notices and the like.

We've had an Echo device for a few years now that was a gift from my mother in law. Having spent a good deal of time using it I'd say that calling it "AI" is a stretch. The device itself has very little capability, it is completely reliant on a cloud service to do all of the heavy lifting, and while it does a pretty good job of recognizing speech, there is a limited selection of things it can actually do. You need to learn exactly the right phrasing to use to get it to do what you want, you can't have a conversation with it and it doesn't recognize who you are by your voice. It's a neat gadget but I mostly use it as a kitchen timer, music player and to control lights with my home automation. Even that functionality is somewhat disappointing, you can tell it to turn on a light, but you can't tell it something like "turn on this light for 5 minutes" or "set this light to come on every night at dusk", it can't do any scheduling at all with the automation.

[CatalinaWOW]

I agree, it isn't AI.  All I am really talking about is implementing the specific tasks identified by the OP.  And I omitted a key one, recognizing you when you enter a room, something the XBox with video attachment does pretty well.  So while there isn't a single, packaged device that will do everything the OP asked for, most of the elements are there and could be knitted together reasonably easily with todays technology.

Your complaints about the limitations of Echo are well founded, but again, many of the things you desire are not technically challenging.  Scheduling, setting an on interval and the like don't require invention, just implementation.

You could even program one of the advanced robots to do the morning greeting, complete with batting eyes and a grin.  I doubt that it would be very satisfying and would probably get old in a hurry.

Now if he wants something that detects moods, asks probing questions to determine the cause of the mood and takes or recommends actions to improve the mood, that is AI and is a long ways off.   Having a conversation is kind of a grey area.  Software that can pass the Turing test, at least with some people on the other side of the terminal, is already here.  But good conversation isn't.   A good argument can be made that those problems have not been fully solved by the meat computers we all run around with.