News:

Look, I haven't mentioned Zeus, Buddah, or some religion.

Main Menu

All things AI

Started by Dave, May 21, 2018, 04:03:59 PM

Previous topic - Next topic

Dave

Getting enough stuff in the media and elsewhere to think this needs a dedicated, but generalised, thread of its own - the digital version of Silver's brain thread!

Here are a few links to start it off:

Quote
Artificial Intelligence.
Should we beware the machines? Professor Stephen Hawking has warned the rise of Artificial Intelligence could mean the end of the human race. He's joined other renowned scientists urging computer programmers to focus not just on making machines smarter, but also ensuring they promote the good and not the bad. How seriously should we take the warnings that super-intelligent machines could turn on us? And what does AI teach us about what it means to be human? Helena Merriman examines the risks, the opportunities and how we might avoid being turned into paperclips.
https://www.bbc.co.uk/programmes/b05372sx

QuoteUsing artificial intelligence to fight terrorism

At the end of a week in which we have been looking at the development of artificial intelligence (AI), we examine how effective computers can be in tracking suspected sex offenders - or terrorists.

The head of MI5 told us this week that the terrorist threat is at its highest level since 9/11. Our security correspondent, Gordon Corera, has been hearing about one British company - founded by people with a background in government and intelligence - to see what its AI technology can do.
https://www.bbc.co.uk/programmes/p032y5cj

QuoteArtificial Intelligence
In Our Time

Melvyn Bragg and guests discuss artificial intelligence. Can we create a machine that creates? Some argue so. And is consciousness, as we are, with headaches and tiffs and moods and small pleasures and sore feet - often all at the same time - capable of taking place in a machine? Artificial intelligence machines have been growing much more intelligent since Alan Turing's pioneering days at Bletchley in World War Two. Its claims are now very grand indeed. It is 31 years since Stanley Kubrick and Arthur C Clarke gave us HAL - the archetypal thinking computer of the film 2001: A Space Odyssey. But are we any nearer to achieving the thinking, feeling computer? Or is it just a dream - and should it remain as one?With Igor Aleksander, Professor, Imperial College London and inventor of Magnus - a neural computer which he says is an artificially conscious machine; John Searle, Professor of Philosophy, University of California and one of only two people in the world to invent an argument, the Chinese Room Argument, which destroys the plausibility of the idea of conscious machines.
https://www.bbc.co.uk/programmes/p00545h7

The above are for discussion only.
Tomorrow is precious, don't ruin it by fouling up today.
Passed Monday 10th Dec 2018 age 74

Dave

Tomorrow is precious, don't ruin it by fouling up today.
Passed Monday 10th Dec 2018 age 74

Dave

Either the BBC are either using computers to read news items or they have found a reader with intonations and cadences that sound very like a better, female, version of Steven Hawkin's voder. It was a description of sentences handed out to rebel soldiers in Turkey.

https://www.bbc.co.uk/programmes/w172w4drhtr0f8b

Tomorrow is precious, don't ruin it by fouling up today.
Passed Monday 10th Dec 2018 age 74

Dave

Discrimination in recognition systems:

QuoteThe Observer
Interview
'A white mask worked better': why algorithms are not colour blind

Joy Buolamwini is a graduate researcher at the MIT Media Lab and founder of the Algorithmic Justice League – an organisation that aims to challenge the biases in decision-making software. She grew up in Mississippi, gained a Rhodes scholarship, and she is also a Fulbright fellow, an Astronaut scholar and a Google Anita Borg scholar. Earlier this year she won a $50,000 scholarship funded by the makers of the film Hidden Figures for her work fighting coded discrimination.
https://www.theguardian.com/technology/2017/may/28/joy-buolamwini-when-algorithms-are-racist-facial-recognition-bias

The problrm goes beyond that into the automatic recognitiin of skin cancers etc.

Aleo on BBC:
QuoteRacist AI
Business Daily

Can artificial intelligence and face recognition technology be racist? AI is increasingly being used in all aspects of our lives but there is a problem with it. It often can't see people because of the colour of their skin. Zoe Kleinman speaks to Joy Buolamwini founder of the Algorithmic Justice League, Suresh Venkatasubramanian from the School of Computing at the University of Utah and Calum Chase, an AI expert and author about what is being done to overcome this problem.
(Photo: Facial recognition system, Credit: Getty Images)

https://www.bbc.co.uk/programmes/w3cswgkg

Some of this is because the big data collectors, Google etc,  statistically get more images of white people to train their AI systems and sell to others.
Tomorrow is precious, don't ruin it by fouling up today.
Passed Monday 10th Dec 2018 age 74

Arturo

I think that is a far cry from racism though. It's not like they intentionally picked out white people only to recognize faces from. I think it's good that this is being brought forth but I don't think it's racism. I think it would be racism if they were intentionally trying to keep people of color out of their market and eliminate them from doing so. And it would only cement their racism if they said that their reason for doing so is if they were irreversibly, fundamentally, and/or inherently flawed in some way.

I wear glasses and often times camera filters have facial recognition and they don't recognize my face correctly. Do I go out and say that people are discriminating against me because I wear glasses? No I think that would be silly.

Because of the things I've seen change when going back and watching old 90's wrestling, I think it's gotten a lot better. The 90's wrestlers of color were always boo'd. And anyone who would fight them, if they were not a person of color, would always be cheered. Even if they were a bad guy. A white person beating up a person of color was always cheered for. Now when I see white people act, (namely people on this forum and elsewhere), I see racism being shot down and people of color being stuck up for when they can't do it themselves. And I think that's a good thing because it gives us a bigger voice.

"I'm interested in one thing Neo, the future. And believe me, I know, the only way to get there is together."

It's Okay To Say You're Welcome
     Just let people be themselves.
     Arturo The1  リ壱

Dave

I don't think they ever said it was "racism", Arturo, but 'statistical bias".  Taking any bunch, using "blind" techniques (every third person say) of people at random out of a crowd that is not congregated for a specific purpose (anti-racialist rally for eg) and the chances are you will not get a representative spread of the national racial mix. In some areas you might get more of any racial type. That forms a statistical bias and may not be useful for working out marketting strategy for goods that might have a cultural (food, clothing etc) influence.

There are other methods, of course, and the above is simplistic. But it seems "bias" does creep in without any sinister motive. But makes the findings untrustworthy.
Tomorrow is precious, don't ruin it by fouling up today.
Passed Monday 10th Dec 2018 age 74

Arturo

No no, you're right. They didn't say it was racist. But they asked the question. I was just giving my piece on it.
It's Okay To Say You're Welcome
     Just let people be themselves.
     Arturo The1  リ壱

Tank

" I think it would be racism if they were intentionally trying to keep people of color out of their market and eliminate them from doing so."

Does that make any difference to the person who is subject to elimination? No, of course it doesn't. Unconscious racism and conscious racism can have the same result. AI and selection algorithms can and will have significant impact on our future and if those algorithms have unconscious biases how will we know if we don't have free and ready to them?
If religions were TV channels atheism is turning the TV off.
"Religion is a culture of faith; science is a culture of doubt." ― Richard P. Feynman
'It is said that your life flashes before your eyes just before you die. That is true, it's called Life.' - Terry Pratchett
Remember, your inability to grasp science is not a valid argument against it.

Arturo

Quote from: Tank on May 28, 2018, 04:53:38 PM
" I think it would be racism if they were intentionally trying to keep people of color out of their market and eliminate them from doing so."

Does that make any difference to the person who is subject to elimination? No, of course it doesn't. Unconscious racism and conscious racism can have the same result. AI and selection algorithms can and will have significant impact on our future and if those algorithms have unconscious biases how will we know if we don't have free and ready to them?

I don't understand your last question there. I think you made a typo or omitted something by mistake.

I'm not programmer or expert on AI, but from what I can tell is that AI can only be programmed to know what we tell it to. It might take that off into other directions like Google's deepmind did with the game "Go" in Korea, ultimately beating the champion, instead of breaking like they thought it would when it went in the other direction that they saw as flawed.

But yeah, for me it's tomato, tomato. Corporations are in the business of making money. And in order to make the most money, they have to appeal to everyone. So it's in their best interest to correct this imperfection. That's why I say it's good that this is brought up. Because the implications can lead to positive outcomes. The same as when the same thing happens in social situations.
It's Okay To Say You're Welcome
     Just let people be themselves.
     Arturo The1  リ壱

Dave

Quote from: Arturo on May 28, 2018, 06:14:07 PM
Quote from: Tank on May 28, 2018, 04:53:38 PM
" I think it would be racism if they were intentionally trying to keep people of color out of their market and eliminate them from doing so."

Does that make any difference to the person who is subject to elimination? No, of course it doesn't. Unconscious racism and conscious racism can have the same result. AI and selection algorithms can and will have significant impact on our future and if those algorithms have unconscious biases how will we know if we don't have free and ready to them?

I don't understand your last question there. I think you made a typo or omitted something by mistake.

I'm not programmer or expert on AI, but from what I can tell is that AI can only be programmed to know what we tell it to. It might take that off into other directions like Google's deepmind did with the game "Go" in Korea, ultimately beating the champion, instead of breaking like they thought it would when it went in the other direction that they saw as flawed.

But yeah, for me it's tomato, tomato. Corporations are in the business of making money. And in order to make the most money, they have to appeal to everyone. So it's in their best interest to correct this imperfection. That's why I say it's good that this is brought up. Because the implications can lead to positive outcomes. The same as when the same thing happens in social situations.

Yeah, I was wondeting about Tank's last line as well!  :grin:

With regards to them pesky algorithms,  true, they are a list of instructions shich the computer cannot deviate from. But, to get the right redult you have to compose exactly the right instruction, no room at all for unexpected factors to cause a single hitch in thousands of lines of code. Misspell a crtical word, forget to zero some  sriabke . . .

The world is full of tiny errors that have big results. Ask a stupid question and you will get a stupid answer. Oh, just remembered good old GIGO - Garbage In Garbage Out!

Where is Davin when you need him?
Tomorrow is precious, don't ruin it by fouling up today.
Passed Monday 10th Dec 2018 age 74

Tank

Algorithms are usually propitiatory and a) not available for public inspection and b) probably too complex to understand by the average person. So we won't have access to them 'physically' and intellectually. We will have to trust the commercial entity that  created them and as we know that never goes wrong :D
If religions were TV channels atheism is turning the TV off.
"Religion is a culture of faith; science is a culture of doubt." ― Richard P. Feynman
'It is said that your life flashes before your eyes just before you die. That is true, it's called Life.' - Terry Pratchett
Remember, your inability to grasp science is not a valid argument against it.

Tank

There is also the issue of AI learning being uninterpretable and chaotic in nature. Two AI systems subjected to identical stimuli from the point of birth (switch on) will not develop identically because of quantum fluctuations during the learning process. They are by their very nature unpredictable.
If religions were TV channels atheism is turning the TV off.
"Religion is a culture of faith; science is a culture of doubt." ― Richard P. Feynman
'It is said that your life flashes before your eyes just before you die. That is true, it's called Life.' - Terry Pratchett
Remember, your inability to grasp science is not a valid argument against it.

Dave

#12
Quote from: Tank on May 28, 2018, 07:08:29 PM
There is also the issue of AI learning being uninterpretable and chaotic in nature. Two AI systems subjected to identical stimuli from the point of birth (switch on) will not develop identically because of quantum fluctuations during the learning process. They are by their very nature unpredictable.

Bit like some people, eh!

True, not sure how much in the way of a personality, ethics etc can be programmed in. Such things probably need vast volumes of mrmory tgat have no other functional use - expensive. Asimov's "positronic brain" had, IIRC, a similar structure to the organic brain, paths were established during learning and remained as firmware. The Three Laws being similarly "burned in".

It has to get cheap enough for potential "rogues" to be recognised and culled at an early stage. Then you need clever, and ecpensive, testing to spot those potential rogues!

The whole field is in its infancy; limited function, energy hungry, still pretty unsafe except for basic functions like route folloeing and collision avoidance. Though even the latter will be oicked over until there are enough stats to show that, on average, it is safer to ket a high-end self-drive car loose in city traffic than have a human driver.

I can see pseudo AI systems, stimulus-reaction systems as we have now, acting as visitor recognisers, house control systems, security etc becoming cheaper in the next ten years. Maybe able enough to recognise that little Jimmy has slipped his leash and is headed for the big bad road and turned left out of the the door. Especially if little Jimmy is wearing his GPS tag . . .

True AI systems with truly independent decision making and action selection (even if within parameters) are a whole other bundle of fun!

A lot of this will fall of the back of the Chinese systems I think. I doubt that they will be able to resist marketing the techniques and scooping up even more foreign capital. But some are already concerned about back-doors and logic bombs built into their products. We truly live in interesting times .

Correction: the "positronic brain" was a matrix made of metal alloys and wss volatile and thus required a constant supply of energy to function. Thought I had better check my memory. We can do better than that can't we? Gotta get the energy requirements and heat losses down though.
Tomorrow is precious, don't ruin it by fouling up today.
Passed Monday 10th Dec 2018 age 74

Arturo

Quote from: Tank on May 28, 2018, 07:01:23 PM
Algorithms are usually propitiatory and a) not available for public inspection and b) probably too complex to understand by the average person. So we won't have access to them 'physically' and intellectually. We will have to trust the commercial entity that  created them and as we know that never goes wrong :D

It also is not subject to being impervious to being corrected. Like Dave said the field is still in it's infancy and many of these things are still getting bugs worked out. When a person makes a new product, the goal is to impress the consumer, not create a flop. My stance is to give the customer the best product possible before release.
It's Okay To Say You're Welcome
     Just let people be themselves.
     Arturo The1  リ壱

Dave

#14
Quote from: Arturo on May 28, 2018, 10:31:20 PM
Quote from: Tank on May 28, 2018, 07:01:23 PM
Algorithms are usually propitiatory and a) not available for public inspection and b) probably too complex to understand by the average person. So we won't have access to them 'physically' and intellectually. We will have to trust the commercial entity that  created them and as we know that never goes wrong :D

It also is not subject to being impervious to being corrected. Like Dave said the field is still in it's infancy and many of these things are still getting bugs worked out. When a person makes a new product, the goal is to impress the consumer, not create a flop. My stance is to give the customer the best product possible before release.

My guess is that certain algorithms will acquire "class names", like functions in a washing machine. Sales patter will go along the lines of, "The Able 5 model House Control and Command system has grade A recognition routines in both the audio and visual functions, at least 99% accuracy once trained. It can differentiate between real command phrases and accidental ones from media due to its enhanced voice spectrum analysis and source location determination abilities. Safeguarding and barring routines have been greatly improved. . ."

(Subtext to last, ". . . after the embsrrasing court case due to pre-teens mimicking the parents' passwords for access to the drinks cabinet and a hard porn channel . . .")

Your camera spec does not need to tell you how its face/smile recognition works, just imply that it does so reliably.

But, if a routine requires "this plus this plus this and that" to get it right chosing an ambiguous "that" is potentially gonna foul things up! Like ". . . if there is light then . . ." Without specifying just what kind of light and not fitting sensors to differentiate between any conditions of sunlight and street plus headlights.
Tomorrow is precious, don't ruin it by fouling up today.
Passed Monday 10th Dec 2018 age 74