UKC

I must admit, I have not always, I just thought of Alexa as a computer, I did not realise it is AI, or even what AI is.

After reading Mo Gawdats book book Scary Smart, I now speak nicely to Alexa.

I just did not appreciate what AI is, and how powerful it is going to become in my lifetime. A much much bigger issue than Climate Change, Covid or Ukraine.

In reply to Godwin:

The only scary thing is why people buy them at all. 

In reply to Godwin:

Alexa is pretty dumb.  The signal processing and word recognition are good but it doesn't try to understand semantics it just has a pretty basic parsing of a limited grammar for simple commands.  I actually think this is a good approach, it is predictable and less creepy.  The API is open and you can script Alexa apps yourself.  I had a go a couple of years ago but got frustrated by the limitations.

If you want AI then Google or Facebook are going further along that path.

In reply to ExiledScot:

AI is all around you, not just Alexa, Siri, Cortona and Google, your smart phone.

Have you ever wondered why when a website  asks you to prove you are Human, it shows pictures of Traffic lights and Zebra crossings?

The reason is, that you are been used to train AI, to drive self driving cars.

I would assess that your one line comment indicates you are as ignorant AI as I was, and apparently 95% of people are. Possibly time to try and bring yourself up to speed on AI. But remember, even the people who develop* AI, don't really understand how it works.

* [EDIT] changed "code" to "develop", because AI codes itself.

In reply to tom_in_edinburgh:

Alexa is just collecting Data, its like a little baby, thats why you should speak to it nicely now, because in a very short space of time AI will be thousands then millions of times cleverer than the entire Human Race, and what will it think of a species that was nasty to it, when it was a baby?

In reply to Godwin:

The bigger issue is that Alexa’s requirement for curt, clear instructions is training our babies to abandon courtesy when speaking to humans, such as teachers.

Alexa should be programmed to respond to politeness and to require a please and thank you.  There should at least be a child response setting where this is required.

In reply to Dave Garnett:

Trust me, thats not a bigger issue.

If I was a Teacher, what I would consider to be a bigger issue, is that in short order, possibly AI will be the Teacher, though there maybe Humans in the classroom, they will be teaching assistants.
It will be logical to have the most intelligent thing in the room doing the Teaching and that will not be a Human, it will be AI, thats if its even worth educating Humans.

Just stopping for a moment, lets go back. You have actually said Alexa is training our babies. Just pause a minute and think on that thought, Alexa is training our babies.

In reply to Godwin:

She could at least teach them to be polite.

That said she won’t be training anyone in our house.  I might occasionally shout at the radio but I don’t speak to anything that doesn’t breathe.

In reply to Godwin:

Always, I am lining myself to be on the correct side of history. I welcome out robot overloads

In reply to Dave Garnett:

Alexa is remarkably taciturn about privacy and Jeff Bezos. It's almost as if someone programmed it to keep stum...

In reply to Godwin:

Worlds gone too far. We need to blow up the internet 

In reply to ExiledScot:

Absolutely. Alexa and all its dystopian ilk can f*** right off. And I would happily shout that repeatedly in its nasty little face until its evil little AI brain has got the message.

In reply to Godwin:

are they AI or just preprogrammed algorithm or decisions, that automatically carry out human decisions, if varying parameters have been met?

Turning on a radio station through listening to a command isn't AI, it's just a human asking a machine to do something they can't be bothered with. 

Adverts that revolve around keywords you've used on search, comments on facebook, films you've watched isn't AI, it's programming where tracking cookies gather data, then use that as search fields to place certain adverts in front of you. It's clever but it's not AI, there are literally 1000s of people working on these algorithms.

Much of so called AI isn't.

In reply to Godwin:

How does it know that you've picked the right pictures? Surely it must already know which are pictures of traffic lights and zebra crossings, otherwise it wouldn't be able to say that you've picked the right pictures. Therefore, it's learnt bugger all.

I 'teach' mine to masterbate, so, if AI is so good at learning stuff, it'll end up being too busy wanking itself to death to be concerning itself with human domination.

In reply to Maggot:

I don't know, but Mo Gawdat says thats what's happening, and Mo seems pretty well informed on these matters to me https://en.wikipedia.org/wiki/Mo_Gawdat

Apparently Elon Musk thinks we need to be careful with it, and I suspect he knows a thing or three about AI.

In reply to Godwin:

We don’t have that in the house, but our car does have Siri, which is kind of useful to phone or text someone hands free while driving. My wife likes to text swear words to me, which Siri dutifully reads out in a posh voice. Highly recommended.

In reply to Godwin:

i think you are mixing up real AI super computers, which aren't AI (yet), and voice activated and pre programmed machines.

true AI isn’t here and yes there's risk. A bit like nuclear weapons, once a achieved there's no going back. So far computers can't identify traffic lights, zebra crossing, buses....

In reply to ExiledScot:

Once there's true general purpose AI, then I will feel obliged to speak nicely to it.  We can discuss whether it appreciates it and what other rights it thinks appropriate. 

In reply to Godwin:

I've heard this suggestion before, and was wondering what your source for this is.  I researched a bit on reCapture technology a few months ago, involving academic paper review, and the ones I looked at made no reference to this hidden agenda, neither did a quick whizz through the references of the papers.  I did find a couple of websites that looked a bit iffy, so didn't take them especially seriously.  If you have any serious sources I'd be interested.

Is this really true?  What background and qualifications do you have to make such a strong statement with that degree of certainty?  And does it really "code itself" - would certainly make a mockery of all the Python code I've been churning out recently

In reply to Godwin:

Elon Musk does rocketry and electric cars, he has a mechanical background, playing a meaningful role in a technical discussion about AI is as far outside his baileywick as an onsight solo of London Wall while selling you a mattress is out of mine. 

In reply to gravy:

"Alexa, what do you think of Jeff Bezos"

"I'd rate him 5 stars"

In reply to Dave Garnett:

I grew up watching Star Trek - talking to Alexa or Siri is basically exactly the same as talking to the Star Trek computer (though obviously it isn't quite as good).  I'm not sure that had any impact on how I talk to actual people with feelings, because it's fairly obvious they aren't the same thing.

In reply to Godwin:

that page (how reliable ?)  says he's a entrepreneur and writer who believes in intelligent design over evolution - so i think i am gonner take his opinions with a bucket of salt

In reply to Godwin:

Go and have a nice lie down.

In reply to Dave Garnett:

FTFY

In reply to Godwin:

Reith Lectures this year were all about AI...https://www.bbc.co.uk/programmes/b00729d9

I got half way through one that was about AI weaponry and then decided I had more cheerful things to be doing : )

In reply to Godwin:

Alexa, schmalexa.  

I do have a Google mini thingy in the kitchen, where it mostly serves as a voice-activated timer; this sounds trivial but once you've used it like this, it becomes an essential tool.  However, it has now started answering back.  I fear that we'll need to bring in a relationship counsellor before too long if I wish to avoid burning things.

T.

In reply to paul_in_cumbria:

Remember when you used to be able to text a landline number and the recipient would be read to by BT Steve (well, that's what I called him).. swear words included! Much fun as a teenager. 

In reply to Godwin:

Did you listen to the recent Reith Lectures on BBC R4? AI was the sublect matter. Available on BBC Sounds.

In reply to Godwin:

Alexa tell (insert names of a couple of kids) dinners ready please - or as we used to do back in ‘the day’ go to the bottom of the stairs and shout ‘DINNERS READY!) 

I experienced the first example recently at someone’s house, f&@king dire…..

In reply to Andy Gamisou:

The comment re reCapture is from Mo's book Scary Smart. 

Possibly contact him via Twitter and ask him

I have no qualification, I am just passing on the message from Mo's book.

Oddly someone has commented on the Wiki link about Mo, and not noted he was Chief Business Officer of Google [X] https://x.company/projects/everyday-robots/ which gave him access to a lot of cutting edge stuff, and Sergey Brin says this about one of his books ”A powerful personal story woven with a rich analysis of what we all seek in a way we can act upon” and not, this guy is total fanatisist https://www.solveforhappy.com . So I consider he may have something worthwhile to say. 

From this thread, I would assess, he is correct, and most people do not appreciate the impacts AI is going to have, and how fast these impacts will be, or he is a nut job.

His key message though is this, AI is going to happen, it cannot be stopped, and it will be the most intelligent thing on the planet, and that we should be teaching it to be a good citizen, in much the same way as parents teach a child to be a good citizen, by actions and example. 

Really the concepts in the book are a bit hard to get over in a Forum post.

https://www.mogawdat.com/scarysmart
 

In reply to Godwin:

Always....

I've seen The Terminator so I know how this ends and, unlike my wife who is often rude to Alexa, I don't want some T2000 telling me that I'm terminated ass-whole...

In reply to Godwin:

I have a colleague who works on big data stuff and I found it fascinating and timely with Apple doing the foundation series by Asimov (phychohistory, with enough people you can predict exactly how they will behave like the gas’s laws).  The interesting thing for me in terms of science fiction becoming fact is he got it wrong.  He said you could not predict the individual.  When people say Alexa spies on them as they saw an advert for something it probably not that or in the way they think.  We think we’re all unique and special but 1 bit of info puts you in a group of 10 million,2 bits, 1 million, 3 bits 100 000 and so on.  Everyone in this group likes rock climbing and will almost certainly like kayaking or caving or cycling.

Quite scary in a way but fascinating.

Forgive the writing….shattered my right arm…..I know…. I can’t believe it either.

Cheers

Toby

In reply to Dave Garnett:

Alexa is a glorified light switch.  I don't think making it pretend to be human, trying to use it to enforce social norms for conversation with humans, or giving it serious AI capabilities would be a step forward.  It's an alternative to a remote control or walking across the room and pushing a button not a person.  Simple direct clearly spoken commands are less likely to be misinterpreted and they prevent all the 'uncanny valley' issues with anthropomorphising computers.

If I was a feminist I'd get angry about inanimate personal assistants all getting female names and referred to as 'she'.

I changed the settings so my command is 'Computer' not 'Alexa',  admittedly mostly because I'm a Star Trek fan but I also think that socially it is better not to give it a human name.

In reply to Godwin:

Both could be true.

In reply to Godwin:

I refuse wherever possible to have anything to do with rubbish like that.

99.9999% of stuff I get directed to me as a result of using Google is useless to me

When I built my first computer in 1980* the idea of voice input would have been fascinating.  Now I see no need for it and I'm definitely not getting on first name terms with anything like that.  "Open the pod bay doors .."

* one of these. 

https://bbprojects.org/computer/tangerine-microtan-65/

The girlfriend in the story had it right! So did Blake

In reply to Andy Gamisou:

It's not some great secret or conspiracy that ReCaptcha is used to train AI models, particularly for pattern recognition. It's not a coincidence that a lot of it is picking out traffic lights, buses, etc.

In reply to wercat:

I wrote an app for Alexa so I could say "Computer, arm the photon torpedoes"  and it would reply "Photon torpedoes armed" it was fun for about ten minutes.  Unfortunately (and probably just as well for the Tories in London) we don't have photon torpedoes so that was all it did.

The most frustrating limitation was it is all <command> <action> you can't get an application running on a server to use Alexa to notify you when something happens.  Even timers are a weird special case.

In reply to john arran:

You should read what he says about computer security, he reckons that AI + Quantum computing, and all computer security will be toast.

In reply to Godwin:

back to HF radio and secure orders/one-time pads then, life in the woods eh?

Avoid the internet altogether

In reply to Godwin:

What he probably means is that all existing computer security will be toast. Of course, there will then be options for quantum-derived security, which I understand may have the potential to be mathematically uncrackable because there simply is no known 'correct' state.

No, I can't really get my head around it either!

In reply to Rog Wilko:

Did you catch the Rutherford and Fry programmes expanding on the themes of those Reith Lectures?
(Adam Rutherford & Hannah Fry, as in 'The Curious Cases of Rutherford & Fry')

Links:
https://www.bbc.co.uk/programmes/p0bcrt38
https://www.bbc.co.uk/programmes/p0bb5lqr
https://www.bbc.co.uk/programmes/p0b8r8zh
https://www.bbc.co.uk/programmes/p0b7gdx3

In reply to tom_in_edinburgh:

Why aren't you a feminist?

In reply to gravy:

I'm not female so I wouldn't describe myself as one.

In reply to tom_in_edinburgh:

A photo of your mouse microphone is required at this point

My favourite AI was the city computer in Cities in Flight, by James Blish. If a city executive made a poor decision, it shot them. We could make good use of it.

In reply to Godwin:

Why?  Doesn't one need some sort of reasonably deep understanding in the area to make this sort of definitive  assessment? I don't think reading the "mood" of the thread is a strong argument especially as the provenance of the contributors is unknown. Clearly, you don't have this background so why are you so sure?  I know I don't have enough insights to make this sort of judgement, despite having some exposure to the issues (40 years software development, first 3 years of which working in the AI lab of a then well known research institute, currently in latter half of an MSc in AI).

In reply to midgen:

It might not be a great secret, but is it true? Maybe they use picking out traffic lights, buses, etc because there's already a big database of images and associated classifiers to use (as they do for their handwriting reCaptures).  Do you have any reliable source for this other than it being "common knowledge".  I'm certainly not saying it isn't true, I'm just looking for better evidence (not sure I want to link to a UKC forum thread as a reference in my dissertation  )

In reply to Godwin:

My guess is that a super intelligent conscious AI would choose to live happily in a simulated world and ignore us.  It probably couldn't be arsed to help us with our problems.  Very likely it would piss off into space because it wouldn't be water/carbon based and would have no reason to limit itself to staying on the Earth.

In reply to Andy Gamisou:

I think it pretty clear that I know very little about AI, and have nothing worthwhile to say on the subject.

However, I would suggest that a person with an academic interest in the subject might look at this thread and check out Mo Gawdat and his book Scary Smart, because if you read the thread, it is possible that due to his role in Google [X] he might have deeper insights than anyone on this thread. 

If I was doing academic research into AI, and had read this thread I would;
 

• Know OP is barking up the wrong tree, laugh and move on.
• Or, be researching Mo Gawdat, reading a couple of informed reviews of his book, and maybe look for something else by Mo Gawdat, maybe some articles, Podcasts or Videos, and see if it possible he knows something I could use, basically dig deeper.

In reply to Godwin:

Excellent book, a narrative on us as much as AI. The difference being the potential power that AI could (will ?) have.

In reply to Andy Gamisou:

I'm sure a computer could be programmed to recognise any image, most phones can identify a face, other apps can ID flower species, they search for cancers in scans etc... It's a question of cost and time, "I'm not a computer" is just a small hurdle for basic hacking and spamming on average sales websites. You won't ever see it as part of a banking site for example. 

In reply to ExiledScot:

As I understand it, AI would not be programmed any image, it will learn itself . 

If you give a child one of those boxes that you push shapes through holes, triangles and stars etc, the parent sitting next to the child does not teach the child, but just rewards the child by smiles and praise when it puts the shape through the correct hole, thats how AI learns, its not programmed.

What Mo Gawdat seems to say is that as AI is learning now, it depends on what we praise it for doing. Mainly we are using it for Commerce and Banking (Greed), Military (Violence) Advertising (Spying), and if a parent motivates a child for Greed, Violence and Spying, what kind of Adult are you going to get.

Mo suggests that AI will be a reflection of us and our values, just a Billion times smarter. As Davvers says, the book is as much about Humans as AI.

In reply to Andy Gamisou:

ReCaptcha has always had a double purpose. Initially you were brute force transcribing books for Carnegie Mellon, it was never about just seeing if you can read the words right!...the system was an obvious fit for Google to employ it training AI models.

https://www.cs.cmu.edu/news/2009/google-inc-acquires-carnegie-mellon-spin-r...

In reply to tom_in_edinburgh:

You don’t have to be female to support gender equality. It’s a bit sad if you truly think that. I’m hoping you just don’t know what “feminist” means. 

In reply to wercat:

Apricot released a portable computer in 1984 with speech recognition but it wasn’t great. But at least they tried to do it on board, not like Alexa where it just records what you say and sends it to a massive datacentre for processing.

https://en.wikipedia.org/wiki/Apricot_Portable

In reply to Mark Edwards:

Those were good computers. Back in the day, the manufacturers handed out "loan" PCs to the software companies to make sure their software ran on it as well as possible. We got some Apricot Xi's, the hard disk version, which were way ahead of their time. When they arrived, a colleague spent a happy morning shouting "WORDSTAR" at it, and somewhat entertaining the team with the randomised results; usually it ran Supercalc. After that, he fortunately gave up and installed Concurrent CP/M instead, which kept him happy as he could do lots of things at once.

They had a high res 800x400 mono display without the screen update issues that the IBM PC had, and the keyboard clipped on the back of the unit to make a luggable, screen in one hand and computer in the other. It went home every weekend without fail for a good number of years.

In reply to freeflyer:

Wordstar and concurrent CP/M  there are memories!

Z80 assembler writing a comms interpreter with conditional execution depending on what the host sent back!  It could use the host to execute commands sent and then receive back data, which could also be in the form of generated commands for the interpreter which could be saved parsed and executed and could involve logging on to different machines, supplying credentials and then transferring data or executing local commands - endless really, like a worm!  back in 1984 it was exciting stuff!

In reply to Godwin:

Sort of, the AI needs to be programmed to be able to learn, once that happens it has the potential to be autonomous to the limitations of the box we put it in (so using a small hard disk or slower computing hardware will result in an AI that may appear to be intellectually deficient in some areas). Put to use it could do some amazing things, as long as it is taught morals. There have been various experiments with self teaching AI that have been corrupted by the arsehole of the Internet, one example is below:

https://www.cbsnews.com/news/microsoft-shuts-down-ai-chatbot-after-it-turne...

It's going to make a huge difference in our lifetimes I think, some of it could be positive, some negative. That said, we are a long way off an AI capable of doing banking/ advertising being used in a medical role for instance, instead we will have lots of specialist AIs, much like we have lots of specialist people. The ethics of different fields will slowly bleed across though (as already happens when someone moves from something like the rail industry to consumer goods). 

In reply to tom_in_edinburgh:

Just as a note, not all. One of the main assistants, Siri, has a neutral name and you pick whether it should have a male or female voice.

In reply to Alkis:

From Wiki:

Why is Apple named Siri? It was co-founded by Dag Kittlaus, Tom Gruber, and UCLA alumnus Adam Cheyer. Kittlaus named Siri after a co-worker in Norway; the name means "beautiful woman who leads you to victory" in Norwegian. Siri's speech recognition engine was provided by Nuance Communications, a speech technology company.

Short for Sigrid apparently so female. I certainly think that it has always been framed as female by default, although you can change the voice to male.

In reply to Neil Williams:

I don't recall J-L ever asking for 'tea, Earl Grey, hot, please'. It's a wonder the Enterprise didn't eventually tell him where to go. 

In reply to Stuart Williams:

Obviously you don't need to be female to support gender equality although even the word 'equality' is hard to define in the context of gender.  Most people would argue for specific additional gender based rights for females because the sexes are de-facto not identical and more of the burden of reproduction is placed on females.

Feminist means different things to different people.  As a male I choose not to call myself a feminist which doesn't mean that I don't have similar views to most feminists.

In reply to Stuart Williams:

Mine has only ever had a male voice and I don't ever recall being asked to choose.

In reply to tom_in_edinburgh:

I’m no clearer on what someone’s gender has to do with whether they support a socio-political movement. 

Is your concern just that the word starts with “fem-“? “Feminist” doesn’t imply that someone is feminine. 

In reply to Neil Williams:

All I can say is that it is a female name and the default voice was female from when it was first released in 2011 until mid-2021 when it was changed so that people had to choose male or female. Either way, someone must have chosen to set yours to a male voice. 

In reply to Stuart Williams:

I must have changed it and forgotten then!

Interestingly it does give a different feel.  Alexa comes across as more knowledgeable to ask questions, whereas Siri comes across as the deferential butler you just tell to do stuff.  Which sort of fits, as Siri's question answering functionality is very, very basic, and Alexa's rather better.  But I wonder if the two voices (and association between Alexa and the Enterprise computer's female voice) influence that feel?

More importantly, someone really should do one with a Brummie accent.  "Oi, Hol, play Red Dwarf!"

In reply to Neil Williams:

It would be interesting to know what the casting requirements were when they were trying to find someone to voice it. I imagine there was probably a lot of thought given to the qualities they wanted in the voice and why. 

In reply to Stuart Williams:

there you go

youtube.com/watch?v=1cr5zFlr6B8&

In reply to midgen:

Then it was all about reading the numbers on front doors taken from Street View - more OCR training.

Anyway, for every technological problem there should be a fix, and Google can force your kids to be polite or not do what they've asked.

https://techcrunch.com/2018/05/08/a-google-assistant-update-will-teach-kids...

In reply to Neil Williams:

I would be more than happy having Kryten around. Not with a Brummie accent though, ideally.

Or Jarvis, but he's a bit sneery.

In reply to freeflyer:

Holly, obviously!

In reply to Godwin:

I was going to train the kids to be polite to Alexa but then I had the epiphany that they are going to grow up in a world full of AI tools, and the more useful life lesson in the long term is to understand that these things are not people, can’t be your friends and you shouldn’t treat them as such.

On a personal note I swear at Alexa deliberately when she gets stuff wrong as I have a (unevidenced) theory that negative interactions like that are more likely to be reviewed by a human to improve the system. 

In reply to lithos:

Ooo, thank you

In reply to Godwin:

Nothing to do with AI and everything to do with quantum computing

https://en.wikipedia.org/wiki/Shor's_algorithm

In reply to mullermn:

While it’s a pretty safe bet that in this specific example that’s true, it seems there’s also something to be said for being nice/polite/kind for its own sake, and the effect on one’s own well-being, even if it doesn’t make much objective difference in a specific situation. Hopefully that’ll mean one will then incline that way more instinctively in a situation where it maybe would be of more consequence

In reply to Currently Resting:

OK but the state of the art in quantum computers is nowhere near where it needs to be to break public key crypto.  That article was talking about them failing to factor 35 in 2019.  Public keys in the low thousands of bits.   

If we get quantum computers with enough qubits and good enough noise performance they can break public keys they'll be able to do a lot of other stuff as well and we'll be living in a completely different world.

Even then public key cryptography is an important component of computer security but it isn't the whole thing.  Shor's algorithm isn't going to help you break a private key cipher like AES. 

In reply to Godwin:

Only when I want something