|
I implemented a rule-based expert system inside one of our products for identifying postal code information within address text, for the purpose of printing that information as the appropriate bar code.
Of course, this was the 1990's so it's not the neural network / machine learning stuff all the cool kids are doing nowadays.
Software Zen: delete this;
|
|
|
|
|
I saw "spam", but nothing about real E-Mail filtering and prioritizing.
That one screams out at me; as I estimate that something like two or three percent of my E-Mail messages are actually stuff that I should read.
|
|
|
|
|
Office365 does here a really good job
|
|
|
|
|
... but other people in the company just prevent it.
For AI, you need data. Data. Data. And again: data.
So, I added some component which would save lots of data to disk. Of course, our system would have to receive an extra external hard disk for storing all those data. Which would have to be sent by snail mail to the office later on. Next, some one would have to annotate all those data, and then some one could try to create the models etc.
But ... well, I need the help of other guys in this company to get that first step done: fit that extra disk into our system, and activate the data collection component.
Futile.
Instead, we continue to rely on some voodoo algorithms for detecting confounding disturbances.
Oh sanctissimi Wilhelmus, Theodorus, et Fredericus!
|
|
|
|
|
I plan (hope?) to use it for robot vision and optimization functions.
|
|
|
|
|
We have an effort going with Carnegie Mellon University to create a missing data replacement AI model for our application. For one set of our data we have about 28K records that we get from an external source. Each record has about 20 data values. For many of the records, some of those values are missing. The AI model will train itself using the known data and then use various algorithms to predict what the missing values in the incomplete records will be. We get more data from the external source daily, so hopefully this AI will keep helping to fill the data gaps.
|
|
|
|
|
If it could be used to create more efficient SQL statements than Cognos, I'd be all for it.
I’ve given up trying to be calm. However, I am open to feeling slightly less agitated.
|
|
|
|
|
If I did work in these fields, I would totally use AI or ML, no question.
|
|
|
|
|
I recently suggested that we add some AI to our web product's customer support. Basically, customer support ranks people based on the intelligence and frequency of calls. So people that call to ask "where is the any key" (yes, this happened recently), well, you know what bucket they go in.
Then the AI determines the length of their hold time.
Of course, like most things, this doesn't require an AI, but we can call it AI.
|
|
|
|
|
I like it! When I used to work in tech support many moons ago, we kept notes on various "problem children". Those people were put at the back of the queue. If you were polite and nice you got bumped up to the front.
"Computer games don't affect kids; I mean if Pac-Man affected us as kids, we'd all be running around in darkened rooms, munching magic pills and listening to repetitive electronic music."
-- Marcus Brigstocke, British Comedian
|
|
|
|
|
It seems appropriate: AI (artificial intelligence) to counteract NS (natural stupidity).
Software Zen: delete this;
|
|
|
|
|
We could use AI to improve local weather forecast;
In general AI could help to analyze large amounts of data to find similarities or patterns -- for scientific or healthcare use
|
|
|
|
|
AI has a big potential, but take care about the boundaries.
|
|
|
|
|
Indications and contra-indications should be based on logic, not 90% certainty of AI. Worked for a pharmacy.
Now, we program for it; we say which pills mix, and which not to take due to diseases like pregnancy. We already know which pills and diseases don't match.
Now, to add insult to injury I pay 5 euro for all new meds, because the ass-istant has to read the label and repeat it to me, to compensate them and that's sold as cross-checking your meds. No, you dumb cow, that's what the PC does for you, and there's a big fat "NOT COMPATIBLE" on your screen.
As for AI handling that; would you prefer rule based, on proven rules, or an AI that's still learning if your wife in that bed?
Bastard Programmer from Hell
"If you just follow the bacon Eddy, wherever it leads you, then you won't have to think about politics." -- Some Bell.
|
|
|
|
|
I worked on contaminant detection in food and pharmaceutical products and ruleset is always favored due to repeatability, possibility to tune it in and less degeneration of results.
GCS/GE d--(d) s-/+ a C+++ U+++ P-- L+@ E-- W+++ N+ o+ K- w+++ O? M-- V? PS+ PE Y+ PGP t+ 5? X R+++ tv-- b+(+++) DI+++ D++ G e++ h--- r+++ y+++* Weapons extension: ma- k++ F+2 X
|
|
|
|
|
|
AI could be a very useful tool in the medical field if used as a tool and not as a driver. Understanding the human body is worlds more complex than driving on the interstate, so we could use an AI to bring all known factors into play to suggest particular testing/questioning/investigating. This and keeping track of drug interactions, allergies/sensitivities, and lifestyle choices (alcohol, driving frequency, heavy machinery, exercise, etc) would be the limit for me.
And all of this presupposes complete ownership of our own data (as in PODS).
|
|
|
|
|
Cpichols wrote: AI could be a very useful tool in the medical field if used as a tool and not as a driver
Wishfull thinking. They cut me open twice. I allowed it because humans, not AI.
No VB6 app that gonna cut into my intestines. Ever.
Bastard Programmer from Hell
"If you just follow the bacon Eddy, wherever it leads you, then you won't have to think about politics." -- Some Bell.
|
|
|
|
|
No to "unsupervised" AI on healthcare, I would agree.
But in most domains, AI as a help to decision can be very useful, if supervised. AI being mostly dumb, its best performance it when coupled with a human expert. And that also tends to improve the expert performance.
Starting with reading scans of all sorts, they can point out the area of interest "instantly" which can lead to faster decisions, especially in case of positive findings. With the condition that the human superviser also knows to look beyond that, especially when it is negative.
|
|
|
|
|
Elrond wrote: No to "unsupervised" AI on healthcare, I would agree. What use is it to supervise an AI messing up?
Elrond wrote: But in most domains, AI as a help to decision can be very useful, if supervised. I need proof, because I never experienced that. I use Google; but that's not an AI, it's vetted by humans, marketing as AI. Find me one, and I'll show you a liar.
Elrond wrote: Starting with reading scans of all sorts, they can point out the area of interest "instantly" which can lead to faster decisions, especially in case of positive findings. With the condition that the human superviser also knows to look beyond that, especially when it is negative. That means a human supervisor has to check the automated guess every time. Meaning, that you pay for automation unused (because they need check all outcome, and do the same amount of work as without AI), or for f***ups (those not checked).
Or let me explain it simpeler; if you in pain, abdomen, cause unknown.. Would you like a chatbot, or a nurse to evaluate what is wrong? Keep in mind, AI doesn't tire, has no bad hairday, doesn't do PMST, is never pissed. Eventually management will go there.
I prefer 'Olga', my fat German nurse. She commands like no AI does, no one challenges a tank. The AI I'd just tell to "byte me". You don't do that will Olga.
Bastard Programmer from Hell
"If you just follow the bacon Eddy, wherever it leads you, then you won't have to think about politics." -- Some Bell.
|
|
|
|
|
Wow, such a narrow view of AI! It is not only for use in chatbots you know? And I agree with you, a chatbot is NOT where I want an AI, at least on healthcare.
By supervising AI, I meant essentially a cooperation to help on decision, and facilitate work, as opposed to an AI that makes the decision alone and rolls on the result without human input/overview.
For scan for example, when checking for cancer, if the AI can point you directly to where it thinks there are cancer cells. If it is correct, it saves a lot of time for the specialist, in that he does not have to review the entire scan and can quickly say yes, there is cancer. And there are a lot of other diseases where it can point directly to areas of interest, saving time for positive diagnosis.
That's only for the negative ones that you may want the human to check the scan anyway, depending on the level of confidence you have. I guess if your confidence is low enough then yes, it may save no time at all. But only if the AI finds nothing.
Having an AI showing the specialist where areas of interest are can also reduce errors by the human who sometimes can pay less attention, glass over too many scans to review. That's where a good cooperation between human and AI is more productive and efficient than either taken separately.
|
|
|
|
|
Better yet
Human and AIs both scan ALL images independently.
If they disagree on diagnosis, then you go to a tie breaker. Worse case, it might require follow up imaging.
|
|
|
|
|
Elrond wrote: Wow, such a narrow view of AI That's your petty opinion.
Elrond wrote: And I agree with you, a chatbot is NOT where I want an AI, at least on healthcare. No you don't. Chatting is where they do least damage.
Elrond wrote: By supervising AI, I meant essentially Oh, dear..
Elrond wrote: For scan for example, when checking for cancer, if the AI can point you directly to where it thinks there are cancer cells. If it is correct "If". Humans hit that if more often.
Elrond wrote: That's only for the negative ones that you may want the human to check the scan anyway, depending on the level of confidence you have. I guess if your confidence is low enough then yes, it may save no time at all. But only if the AI finds nothing. You a lot of faith. Not only, always.
Elrond wrote: Having an AI showing the specialist where areas of interest are can also reduce errors I pointed out the area of interest; they made pictures without AI.
Your confidence and trust are admirable, but if you were a son o' mine I'd hit you. Rather hard, so you'd never forget. You can be naive all you want.
Bastard Programmer from Hell
"If you just follow the bacon Eddy, wherever it leads you, then you won't have to think about politics." -- Some Bell.
|
|
|
|
|
I do think I was naive now. But for a different reason.
Quote: I'd hit you. Rather hard,
That sums it up.
|
|
|
|
|
That's a nice selective quote.. People raise kids with different values, and that has nothing to do with this topic at all
Bastard Programmer from Hell
"If you just follow the bacon Eddy, wherever it leads you, then you won't have to think about politics." -- Some Bell.
|
|
|
|
|