AI applications:

We all know how AI is going to be powerful and drive the world in future. But I don't see any real world applications which can be developed using AI. If you know any real world applications please share here.

For eg. Traffic signals automatically managed by AI as per actual traffic?

Machine learning (which is a subset of AI) is everywhere.

Ex: the recommendation algorithms for Amazon and Netflix.

@DeepHotel yes those are not replacing human. Something in real world where a human is replaced.

@prash32 hmmm

Call centers. Instead of talking to a customer support representative, you talk to a digital agent that uses natural language processing to understand what you are saying

The car GPS manufacturer TomTom lays 10% (500) of its employees off, as advances in automatic map creation made them superfluous (https://glm.io/165839?n; in German).

@DeepHotel as much as i like this idea, i guess that won't really happen.

We're talking about call centers. The kind of resource you use when you can't or don't want to write an email or read the knowledge base article. We're using humans for that just for the sake of it, it's absolutely unnecessary that call centers exist in general.

Also, i think many requests via phone are made to negotiate something out of the ordinary.

I think police and judges can profit from AI in using evidence. You train it to recognize what's going on and the AI is able to make a decision on who's guilty or not.

The best use will be games i think. Where you can train AI to react in different ways to user interaction and the response of the AI is not the same move over and over again but the AI is actually using tactics.

Stop using AI to describe chatbots. Also - Expert information systems, ML, and any "shit that is unable to learn" project.
NLP is not AI. Supervised learning is not AI. NN is most definatly Not AI. Deep Learning is just NN with a hidden layer.

AI can learn by itself. It is self aware, and able to reflect on it own processes. It is multipurpose, and is not limited to one thing like "have you tried turning it off and in again" support calls.

@nitwhiz At the current state of transparency in AI it surely would be "funny".

AI: You are guilty.
Defendant: Why?
AI: 🤷

(Years later)
Human judge: Sorry to have convicted you, but the AI linked a blue shirt with burglary. To prevent future issues like these, please do not wear a blue shirt at court next time.

@nitwhiz i understand what you're saying, but the technology already exists and is being implemented: https://cloud.google.com/solutions/...

https://five9.com/products/...

Also, I think using AI to help judges and police is an interesting idea. How would you approach minimizing bias with a system like that?

@magicMirror Huh?

“Stop using AI to describe chatbots. Also - Expert information systems, ML, and any "shit that is unable to learn" project.”

So expert systems, ML, and production systems aren't under the umbrella of AI? If you take any AI course, they'll tell you that all three of those are considered AI.

“NLP is not AI. Supervised learning is not AI. NN is most definatly Not AI”

How is NLP not under the umbrella of AI? It’s literally listed as one of AI’s use-cases on the wikipedia page for AI. Why aren’t NNs a part AI? Just because you have to use supervised learning to teach them?

“AI can learn by itself.”

So AI is anything that can learn by itself? If I train an NN with a genetic algorithm (unsupervised learning), does that mean that the NN is now under the umbrella of AI?

“It is self aware, and able to reflect on it own processes”

It seems like you are mixing ARTIFICIAL intelligence with NATURAL intelligence

@DeepHotel
Define inteligence. srsly - make an effort.
Then you might understand why some very complicated algorithems, that generate a "thats a banana"/"not a banana" image classfier is not inteligence.

@magicMirror I like this one:
"Intelligence is the ability to achieve specific goals"

@DeepHotel i have no idea whatsoever.:D

But in my naive thinking I'd say "it just works", because it has all law-history with all cases ever occured, their results plus if they got corrected and if so, why.

It would decide just like a human, but with less "wrong" decisions.

anything that is easy for brain to do but difficult for a programmer to code.

@magicMirror you didn’t answer any of my questions. What current software or technology would you define as AI?

Think of times 20 years from now. What AI ML would you develop in real world to make our life better.

@Lensflare Really?
By that definition, a shoe is inteligent. Also, a door knob, a knife, and a leaf.
@DeepHotel
I don't remember you asking. But in any case - there is no such thing at the moment. All of our current tech is only able to solve one question at a time, and really simple questions at that.

@magicMirror of course it needs to be an agent to be intelligent. So, being able to interact with the environment and make decisions and actions. Obviously, agents are not generally intelligent.

@Lensflare Are you perhaps talking about Agency?
A programable agent usualy has a specific task - say, "buy me 20 liters of beer at 2$ per liter", or "buy me 2000 shares of a stock, at 3.5$ or lower".
Able to react to input, and produce an output, is a printer. Able to fix itself, is a virus/microbe.

@magicMirror well, yes.
You can measure its intelligence by how well it is able to accomplish its task.
If it’s a printer who’s task is to print, let’s see what happens if you take away the paper, the ink or the electric power that it needs. An agent that will fail to print then, is less intelligent than an agent that still manages to print by somehow getting the paper, the ink and the power.

@Lensflare Ok.
Inteligence is the ability to solve the following problem:
"get me a print copy of that".
Non inteligent problem is:
"given a printer, digital document, paper, ink, and printing algorithem, get me a printed copy of the digital document".

@magicMirror it’s not about how hard the problem is. It’s about how good the intelligent agent is at finding ways to solve the problem given arbitrary obstacles.

@Lensflare Have you ever used an A*, or a DFID algorithem to solve a maze?
Are you calling those Inteligence?

@magicMirror I would definitely say that search algorithms are intelligent

@magicMirror I have used A*.
And no, algorithms is not intelligence. At least not that kind of algorithms.

Automated warehouses (partial, see amazon)

Automated groceries (theres something like two or three test bed stores for this)

Automated video captioning

Translation is sufficiently good that it will begin to destroy the job of professional translators in the next five years

Mental health diagnosis from text samples, behavior analysis, speech patterns, etc.

Xray interpretation is going to become as common a skill among doctors as legible writing it (give it anout seven years for regulatory hurdles though)

Video editing at the amateur and semiprofessional can and is being automated (film level editing wont be commercially viable in terms of demonstrated projects for another five years though)

Writing tools, in the realm of stories, characters, and dialogue are pretty much there and commercializing as we speak (with major adoption going to take place between now and the next three years)

Circuit design.

Materials engineering research

Quantum chemistry research

Etc.

Some more:
Pharmaceutical research

Structural design

Cooking is about 3-5 years off with large chain restaurants being first adopters (staff will be laid off, and those remaining will be assigned to cleaning and restocking the machines).

Machine supervision (supervising automated machines to report needing material restock, cleaning, or maintenance- which will still be done manually) will itself be automated.

Jobs both high and low will go away, replaced by apps and services. The human element will largely become a "friendly face" put in front of algorithmic decision making and maintenance. Most of maintenance will be spun off into third party companies. While for example the guy who restocked the fridge with fries at mcdonalds? That job will go to the people delivering the fries themselves.

I estimate about 40% of all jobs will be semi or fully automated within ten years.

Tl;dr trust my estimates. I read on the order of 500-1000 pages of actual research per year.

@DeepHotel Ok then.
You are saying the ability to ctrl-f a web page for a word is inteligence.
I now understand your point of view. While I don't agree, it does show why "A"I is so commonly used term.