Team-BHP - Artificial Intelligence: How far is it?

Quote:

Originally Posted by shashanka (Post 4267236)

Yes, I read the article too. Somewhat misleading and simplistic I felt. Ethnic differences in social attitudes & behaviour often gives rise to very similar facial expressions but which have vastly different significance. I would rather not go into specifics and risk giving offence where none is intended! As Sathish81 pointed out, this could be an example of "high end" automation and not really AI:)

But then. One would build seed/training models based on ethnicity/culture, right? :)

Edit: 60-70 percent plus is pretty good, actually.

But I challenge AI to build an Idiot's Guide to understand women, and recommend the optimal response to women ;)

Quote:

Originally Posted by nilanjanray (Post 4267259)

But then. One would build seed/training models based on ethnicity/culture, right? :)
Edit: 60-70 percent plus is pretty good, actually.
But I challenge AI to build an Idiot's Guide to understand women, and recommend the optimal response to women ;)

Yes I suppose so - but imagine the bottomless mix with cross-ethnic mixes! The seed-training would require seed-training would require......! The endless onion-skin situation!:)

But I second the proposal about the Idiot's Guide to....imagine the relief all round. Everyone could be Mel Gibson in What Women Want!

Quote:

Originally Posted by sgiitk (Post 4270373)

Despite all the whumf, has any AI implementation passed the 'Turin Test'. Answer is No. So let the question rest.

The questions wasn't whether it passed the Turin test, but how far AI has come.

However, your answer does illustrate how many people think about AI. They look for something that outdoes every human component of intelligence. Clearly we are a long way from it.

But that's not necessary what AI is about. To quote Wikipedia

Artificial intelligence (AI, also machine intelligence, MI) is intelligence exhibited by machines, rather than humans or other animals (natural intelligence, NI). In computer science, the field of AI research defines itself as the study of "intelligent agents": any device that perceives its environment and takes actions that maximize its chance of success at some goal.[1] Colloquially, the term "artificial intelligence" is applied when a machine mimics "cognitive" functions that humans associate with other human minds, such as "learning" and "problem solving".

So it's not necessarily about the total human intelligence, but against any or certain parts of it.

If you look at the two links I provided you see real life applications of AI along those lines. Human intelligence allows us to spot trends and that skill can be honed and trained. Many profession in past and present rely on it.

But AI computers are infinitely better at it then humans and can scan vast amounts of data, pick up trends, learn and adapt their logic.

The biggest mistake we can make is to not take AI seriously until such time a Turin test gets passed.

Under its current definition AI is very real and is already encroaching on everybody's life, but we rarely notice and in most cases simply don't know it's happening.

Jeroen

Whether we are for or against, I believe AI will be pervading more or less all aspects of our lives. The only question is how to regulate the usage so as to not make its application in various fields less dangerous.

For example self driving cars. I have recently come across incidents where cars were hacked and hackers took complete control of the vehicles. While they aren't complete auto pilot cars, it isn't improbable to think the same can be done to fully autonomous self driving cars.

https://www.wired.com/2015/07/hacker...l-jeep-highway

An excerpt from the above article

Quote:

"All of this is possible only because Chrysler, like practically all carmakers, is doing its best to turn the modern automobile into a smartphone. Uconnect, an Internet-connected computer feature in hundreds of thousands of Fiat Chrysler cars, SUVs, and trucks, controls the vehicle's entertainment and navigation, enables phone calls, and even offers a Wi-Fi hot spot. And thanks to one vulnerable element, which Miller and Valasek won't identify until their Black Hat talk, Uconnect's cellular connection also lets anyone who knows the car's IP address gain access from anywhere in the country. "From an attacker's perspective, it's a super nice vulnerability," Miller says.

From that entry point, Miller and Valasek's attack pivots to an adjacent chip in the car's head unit—the hardware for its entertainment system—silently rewriting the chip's firmware to plant their code. That rewritten firmware is capable of sending commands through the car's internal computer network, known as a CAN bus, to its physical components like the engine and wheels.

A set of GPS coordinates, along with a vehicle identification number, make, model, and IP address, appears on the laptop screen. It’s a Dodge Ram. Miller plugs its GPS coordinates into Google Maps to reveal that it’s cruising down a highway in Texarkana, Texas. He keeps scanning, and the next vehicle to appear on his screen is a Jeep Cherokee driving around a highway cloverleaf between San Diego and Anaheim, California. Then he locates a Dodge Durango, moving along a rural road somewhere in the Upper Peninsula of Michigan. When I ask him to keep scanning, he hesitates. Seeing the actual, mapped locations of these unwitting strangers' vehicles—and knowing that each one is vulnerable to their remote attack—unsettles him.

https://www.theguardian.com/technolo...rakes#comments

I have read that Auto pilot cars are actually designed to be more careful and as such even if hacked may not be of much use. However, I think the hackers can use that security features and make a car brake hard in a fast moving highway or swerve sideways by making the system believe there is a obstacle in front, causing fatal accidents.

I have posted this elsewhere in the Tesla thread, but felt it is more pertinent here also, so excuse the double posting.

Quote:

Originally Posted by Jeroen (Post 4270364)

Here is some interesting uses of AI

http://www.bbc.com/future/story/2017...ai-protects-us

Jeroen

Quote:

Originally Posted by Samurai (Post 4276070)

The stuff in the following article is not even AI. But imagine this information in the hands of AI.

https://www.economist.com/news/leade...ce?frsc=dg%7Ce

Most current applications of (pseudo-ish) AI are only as good as the humans designing any particular AI engine.

A majority of curent research is aimed and funded at and by either military, medical or tech companies. There's nothing particularly wrong with that but it does tend to shape the trajectory and everyday usefulness potential of AI.

As of now quite a lot of research is at a novelty stage - exploring/ trying new things and getting weird results (like those FB bots). This is particularly true in the non-profit acedamia research - they have numerous research hypothesis/ papers and are waiting for someone (private/ corporate world) to develop something useful out of it.

Then at the extreme end is Amazon's AAI, look up Mechanical Turks :p

Quote:

Originally Posted by mn2363 (Post 4276090)

Again, I think it is very much about semantics and what one would consider AI to start with. Anything remotely resembling anything that would pass the Turin test is a long way off. But AI has most definitely been introduced in many (commercial) sectors. If you are on social media, what you see on FB for instance is more and more controlled by AI. Insurance companies, bank, aviation to name a few are heavy into AI. Basically, anything that benefits from fast and accurate big data mining is likely to be looking into or already into applying AI and figuring out what else it can do / should do.

To think only as AI as something that is similar to human is too limited a way of thinking about AI. Current AI is already outperforming human skills and intelligence. We are simply very limited as to how much data we can comprehend and handle

Where programming stops and it becomes fully autonomous is almost a philosophical debate. I mean at some point in time, some code was written that allowed it te become autonomous. Whatever it does can always be traced back to code.

Jeroen

Rapid strides in Artificial Intelligence
If you thought that AI taking over human jobs is a worrying concern, you are in for a rude shock. The possibility if AI taking over decision-making and the future is NOT science fiction. Most noteworthy scientists, philosophers and futurists share this concern, and recommend responsible development of AI.
https://www.youtube.com/watch?v=8nt3edWLgIg

Mandatory & unconstitutional Aadhar enforcement:
Given the need to be more responsible with our data, we have done the exact opposite. Unlike all developed nations that have scrapped this Orwellian idea, every Indian citizen will now have to surrender his bio-metric data to the State to keep his bank account live, own a phone, file taxes and even to drive.

Whatever maybe the multitude of arguments for this approach, there are two massive risks: (1) purposeful human abuse & (2) rise of AI

Right now, the mandatory Aadhar debate is centered around privacy. But, did we ever stop to think about the AI implications of our actions?

Our legislators may promise safegaurds to prevent Aadhar misuse by human beings and corporations. But, how will we prevent Aadhar (private biometric data) misuse by AI, something that we don't even fully understand?

So, why should we voluntarily handover human-identifier data to machines that will very soon (within 2 decades) be smarter than us ?

I am fairly certain that our legislators are not aware/bothered about the emerging thereat of AI. But, shouldn't our proactive judicial machinery seriously consider the security implications of AI?

While Team-BHP may be a car forum, a car signifies freedom and that is what this post is all about. My two recommendations are:

1) Fellow citizens should STOP signing up for Aadhar like it is free candy. I don't have one, and I wish I had more people on my side.

2) Activist lawyers should file PILs questioning the validity of Mandatory Aadhar, with regard to impending AI security concerns.

Absolutely agree with the points made. We are heading in a very dangerous direction without understanding the risks involved. The Aadhar project to me looks very much like Hydra's Project Insight from the Captain America movies and allows the deep state to have unprecedented control and surveillance over entire populations. The Supreme Court's privacy related judgement notwithstanding, they have crossed all reasonable limits. The monster that have created seems to have become self sustaining now.

Quote:

Originally Posted by ashwin489 (Post 4282838)

1) Fellow citizens should STOP signing up for Aadhar like it is free candy. I don't have one, and I wish I had more people on my side.

As much as I'd like to, but as a businessman, it is not possible to work without Aadhar. They'll freeze the bank accounts and I can't file ITRs.