| Some more interesting AI results 10:34 - Nov 11 with 863 views | StokieBlue |
In this study the AI was able to determine if people had a chance of dying within a year by looking at their ECGs even when doctors thought they were normal:
https://www.newscientist.com/article/2222907-ai-can-predict-if-youll-die-soon-bu
The issue here as with many AI results is that we have no idea how it's come to these conclusions even though they seem to be correct. This is the next big step - getting the AI to be able to explain to humans how it's reached the conclusion. Until that's possible they are black-boxes in which information goes in and comes out but we have no idea how the conclusion is reached. That could lead to issues in the longer term.
SB |  |
| Avatar - IC410 - Tadpoles Nebula |
| |  |
| Some more interesting AI results on 10:36 - Nov 11 with 851 views | unbelievablue |
Sorry, can't help myself, but as per usual with this sort of thing, this ISN'T AI. Unless the term has morphed to now actually mean the wrongly used definition. |  |
| |
| Some more interesting AI results on 10:40 - Nov 11 with 831 views | StokieBlue |
| Some more interesting AI results on 10:36 - Nov 11 by unbelievablue |
Sorry, can't help myself, but as per usual with this sort of thing, this ISN'T AI. Unless the term has morphed to now actually mean the wrongly used definition. |
Depends on your definition. I guess this would technically fall under machine learning but I just went with the terminology in the article otherwise when people who aren't familiar with the subject read what I've posted and then the article there could be a disconnect and it's confusing.
At the moment it's fairly standard and not particularly bad to use ML and neural nets under and all encompassing AI term. AI doesn't have to be sentient by definition:
"Colloquially, the term "artificial intelligence" is often used to describe machines (or computers) that mimic "cognitive" functions that humans associate with the human mind, such as "learning" and "problem solving".
SB [Post edited 11 Nov 2019 10:41]
| |
| Avatar - IC410 - Tadpoles Nebula |
| |
| Some more interesting AI results on 10:43 - Nov 11 with 818 views | unbelievablue |
| Some more interesting AI results on 10:40 - Nov 11 by StokieBlue |
Depends on your definition. I guess this would technically fall under machine learning but I just went with the terminology in the article otherwise when people who aren't familiar with the subject read what I've posted and then the article there could be a disconnect and it's confusing.
At the moment it's fairly standard and not particularly bad to use ML and neural nets under and all encompassing AI term. AI doesn't have to be sentient by definition:
"Colloquially, the term "artificial intelligence" is often used to describe machines (or computers) that mimic "cognitive" functions that humans associate with the human mind, such as "learning" and "problem solving".
SB [Post edited 11 Nov 2019 10:41]
|
Yeah, true. Interesting results there though as you/the article says.
I think it frustrates me a little having worked in IT, Software etc. for a while. AI is used by consultancies as a buzzword marketing term to sell their products and services, when actually all they were doing was some interesting (or not very interesting) coding and machine learning. Disingenuous.
This is a good piece from the Atlantic too:
https://www.theatlantic.com/technology/archive/2017/03/what-is-artificial-intell | |
| |
| Some more interesting AI results on 10:43 - Nov 11 with 818 views | DanTheMan |
| Some more interesting AI results on 10:36 - Nov 11 by unbelievablue |
Sorry, can't help myself, but as per usual with this sort of thing, this ISN'T AI. Unless the term has morphed to now actually mean the wrongly used definition. |
How are neural networks not AI? |  |
| |
| Some more interesting AI results on 10:48 - Nov 11 with 796 views | unbelievablue |
| Some more interesting AI results on 10:43 - Nov 11 by DanTheMan |
How are neural networks not AI? |
Is this a neural network? | |
| |
| Some more interesting AI results on 10:51 - Nov 11 with 785 views | Oldsmoker |
| Some more interesting AI results on 10:43 - Nov 11 by unbelievablue |
Yeah, true. Interesting results there though as you/the article says.
I think it frustrates me a little having worked in IT, Software etc. for a while. AI is used by consultancies as a buzzword marketing term to sell their products and services, when actually all they were doing was some interesting (or not very interesting) coding and machine learning. Disingenuous.
This is a good piece from the Atlantic too:
https://www.theatlantic.com/technology/archive/2017/03/what-is-artificial-intell |
If they don't feed the actual outcome back to the database to say this person actually had it or didn't then how is the computer "learning".
It might give a result based on parameters such as expected ranges and if a reading is outside the range then if flags it up.
Unless those ranges are modified every time they know for certain that those results gave a positive or negative I don't see AI. |  |
| |
| Some more interesting AI results on 10:59 - Nov 11 with 767 views | DanTheMan |
| Some more interesting AI results on 10:48 - Nov 11 by unbelievablue |
Is this a neural network? |
My mistake, I thought I'd found a linked study but that was in fact a different neural network study on ECGs.
Although usually when people mention training models, it's going to be neural network based, unless they are using something like a genetic algorithm (which could itself be classed as AI).
I think the article you linked is incredibly strict in what it classifies as AI in computing. | |
| |
| Some more interesting AI results on 11:04 - Nov 11 with 755 views | unbelievablue |
| Some more interesting AI results on 10:59 - Nov 11 by DanTheMan |
My mistake, I thought I'd found a linked study but that was in fact a different neural network study on ECGs.
Although usually when people mention training models, it's going to be neural network based, unless they are using something like a genetic algorithm (which could itself be classed as AI).
I think the article you linked is incredibly strict in what it classifies as AI in computing. |
Not sure - can't it just be deep learning without using artificial neurons?
Read a few links over the past half hour. I think I'm behind the curve on how AI is being used as a term. Hopefully a new term emerges which truly defines sentience, though I gather that's particularly hard. | |
| |
Login to get fewer ads
| Some more interesting AI results on 11:10 - Nov 11 with 746 views | StokieBlue |
| Some more interesting AI results on 10:59 - Nov 11 by DanTheMan |
My mistake, I thought I'd found a linked study but that was in fact a different neural network study on ECGs.
Although usually when people mention training models, it's going to be neural network based, unless they are using something like a genetic algorithm (which could itself be classed as AI).
I think the article you linked is incredibly strict in what it classifies as AI in computing. |
Usually genetic algorithms don't need training (well not the ones I've implemented) - they work on the principle of natural selection and randomised evolution and each run is independent to previous results. You could integrate past results into the evaluation but it's not something I've seen before.
Agree with your main point though, usually it will be a neural net that requires training and that the definition in the article posted is very strict.
It's an interesting field.
SB | |
| Avatar - IC410 - Tadpoles Nebula |
| |
| Some more interesting AI results on 11:22 - Nov 11 with 727 views | DanTheMan |
| Some more interesting AI results on 11:10 - Nov 11 by StokieBlue |
Usually genetic algorithms don't need training (well not the ones I've implemented) - they work on the principle of natural selection and randomised evolution and each run is independent to previous results. You could integrate past results into the evaluation but it's not something I've seen before.
Agree with your main point though, usually it will be a neural net that requires training and that the definition in the article posted is very strict.
It's an interesting field.
SB |
Fair point on the Genetic Algorithm, it's been a long time since I've done them and managed to completely forget you don't end up actually training them at all, they just "do". Although they are still, at least from a computer science point of view, AI.
Also get UBs point of view, some of the marketing that comes out from companies saying they are doing AI is obviously nonsense. Same with "we use blockchain", double points if you use both AI and blockchain. | |
| |
| |
By continuing to use the site, you agree to our use of cookies and to abide by our Terms and Conditions.
