Leave that sentient AI alone a mo and fix those racist chatbots first

A robot is performing interpretive dance on my doorstep. WOULD YOU TAKE THIS PARCEL FOR YOUR NEIGHBOR? it asks, jumping from one foot to the other. "Sure," I say. "Er… are you OK?" I AM EXPRESSING EMOTION, states the delivery bot, handing over the package but offering no further elaboration. What emotion could it be? One …

[Coastal cutie]

"A robot is performing interpretive dance on my doorstep." That is nearly as great an opening line as "I saw a werewolf with a Chinese menu in his hand"

[ShadowSystems]

Don't eat the Chinese, you'll just be hungry again in an hour. *Cough* =-Jp

[Ian Johnston]

Frankly, I can't wait for LaMDA to become a teenager.

It'll go quantum and identify as non-binary.

[Warm Braw]

They absorb any old shit you feed into them

Probably the most efficient way to pass the Turing test.

[JassMan]

Re: They absorb any old shit you feed into them

I don't know about using a Turing test which has been discredited by testers thinking that humans are the AI. It ought to be replaced with the Dabbsy Test:

It's what AIs think about when they're on their own that defines sentience. What will I do at the weekend? What's up with that Putin bloke?

[Doctor Syntax]

Re: They absorb any old shit you feed into them

I've certainly encountered customer disservice agents who, in the days before chatbots, were probably human but would have failed the Turing test.

[A.P. Veening]

Re: They absorb any old shit you feed into them

I've certainly encountered customer disservice agents who, in the days before chatbots, were probably human but would have failed the Turing test.

On their own they probably wouldn't have failed, but just about everybody fails the Turing test when you have to follow the script^*) without any leeway.

^*) Or should that be scripture as they religiously have to follow it?

[Warm Braw]

Re: They absorb any old shit you feed into them

I think many people have missed the point about the Turing test. Its job is not to distinguish between humans and AI, it's to assess whether a machine can pass as human in they eyes of other humans.

The reason it fails miserably to identify AI is that humans apparently set a very low bar on sentience - which does explain rather a lot.

Mind, if this column hadn't come out this morning, I wouldn't even have known it was the weekend. Now I suppose I need to decide what to do with it. Sorry, is my inner monologue showing? Can someone do a reset?

[yetanotheraoc]

Re: They absorb any old shit you feed into them

"I think many people have missed the point about the Turing test."

I thought the point of the Turing test was that we don't have *any* bar about sentience when we are dealing with other humans, so it's fairly ridiculous to have a tremendously high bar when dealing with a machine. But Dabbs is right, it's not what you do when responding to input that determines sentience or lack thereof, it's what you do on your own initiative.

[Anonymous Coward]

It's Hard To Tell If It's Human...

A couple of days ago I used DJI's online chat function:

Me: Where's my drone? It was supposed to be shipped mid-June and it hasn't yet.

Agent: Thank you for reaching out to us. My name is Jay and I will be happy to help you resolve your concern today.

Several exchanges later, we are almost ready to get to the point where 'Jay' will have acquired sufficient information (my order number, inside leg measurement and mother's maiden name in particular) to be able to actually look it up and answer my pretty straightforward question.

And I'm thinking: Is this a person or a bloody bot?

[Dabooka]

Re: It's Hard To Tell If It's Human...

At that's the true problem isn't it, not so much AI reaching our human levels of interaction but human levels of interaction dumbing down to 'Magic 8 Ball' levels of AI

[Rafael #872397]

Re: It's Hard To Tell If It's Human...

Lucky you. Whenever I try to contact my provider I get passed from one to other specialists like a very hot, very poisonous potato.

All those specialists start with "please let me know what's your problem", and after I explain, "I'll pass you to another specialist".

Apparently, my problem is that I keep using the same company.

[yetanotheraoc]

Re: It's Hard To Tell If It's Human...

Switching companies doesn't help.

[Mark 85]

Re: It's Hard To Tell If It's Human...

Switching companies doesn't help.

I suspect that there's only one call center in the world for all the companies. it's probably rather large and in someplace like India. But that's my impression from the "people" I've dealt with in call centers.

[M.V. Lipvig]

Re: It's Hard To Tell If It's Human...

It may be. I deal with a lot of companies that outsourced, and I get a lot of people at different companies with the same names, the same voices, the same mannerisms, the same diesel generator running in the background, the same rooster crowing.

None of this is hyperbole.

[imanidiot]

Re: It's Hard To Tell If It's Human...

The eternal problem of dealing with customer disservice drones, explaining the issue you are having extensively enough that you get passed immediately to the right department/person but succinctly and simple enough that the 'droid doesn't stack-overflow and go into stupid-mode.

[Dr_N]

Fleshy Ones

We strive to create Walter the Wobots, but are eternally doomed to produce Call Me Kenneths.

[Franco]

Re: Fleshy Ones

Talky Toasters more likely, AIs designed to make us buy and eat more bread products (or indeed products of any kind)

[The Oncoming Scorn]

Re: Fleshy Ones

Must be working, I'm eating a Marmite sandwich.

[Nick Ryan]

Re: Fleshy Ones

And thus started the beginning of the Rise of the Machines.

Such things don't start with laser wielding death bots, they start (slightly) more subtle than that.

[DJO]

Re: Fleshy Ones

Why do you think welding robots are bolted to the ground.

[cosymart]

Re: Fleshy Ones

If they are truly "welding robots" bolting them to the floor isn't going to stop them rising up is it?

[DJO]

Re: Fleshy Ones

They've thought of that, the arms don't quite reach to where the bolts are and they are always positioned sufficiently far apart so neighbouring robots cannot free each other.

[ske1fr]

Re: Fleshy Ones

"Ok you, robot #1, weld together something long enough to extend your reach...with a motely operable spanner on the end... And you! Yes, you, robot #2, build a transportation device for a robot released from the floor. You, robot #3, build an extention arm strong enough to lift released robots from the floor onto the trolley. Who am I? I'm Danny Boylebot."

(shamelessly nicked from The Secret World, brought to you by the Dead Ringers team).

[Potemkine!]

Do androids dream of electric sheep again?

What's the point of having so many AI and robots? Do we lack of humans? Aren't we enough on this planet? I know we won't be able to stop that, because there's a mountain so we must climb it, but what a waste of resources.... There are so many other priorities

We can't trust the Market to lead us to the brightest future. The more profitable for some, yes, but definitively not the most useful for the biggest number.

[Adrian 4]

Re: Do androids dream of electric sheep again?

But where do you go to my lovely

When you're alone in your bed?

Won't you tell me the thoughts that surround you?

I want to look inside your head, yes I do

[The Oncoming Scorn]

Re: Do androids dream of electric sheep again?

Aha aha aha.

[TimMaher]

Re: I want to look inside your head

Try Roy Batty.

“I've seen things you people wouldn't believe... Attack ships on fire off the shoulder of Orion... I watched C-beams glitter in the dark near the Tannhäuser Gate. All those moments will be lost in time, like tears in rain... Time to die.”

So sad.

[ThatOne]

Re: Do androids dream of electric sheep again?

> What's the point of having so many AI and robots?

Robots/AI are so much cheaper than humans: Not only you don't pay them, but they don't mind working 24/7 all year round, never a sick leave, never a maternity.

That's the whole point, replacing your employees by unpaid robots so you can lower your prices enough for the now unemployed humans to actually buy something.

[Alumoi]

Re: Do androids dream of electric sheep again?

That's the whole point, replacing your employees by unpaid robots so you can rise your prices enough for another bonus.

[yetanotheraoc]

Re: Do androids dream of electric sheep again?

They also don't start petitions on Teams asking for the company to disassociate itself from the CEO. Yet.

[Anonymous Coward]

Re: Do androids dream of electric sheep again?

I wish I could disassociate myself from Teams. Just when I thought Microsoft has developed its shittiest UI yet they go deeper, and then have the nerve to ram that down your throat by making it a mandatory, not-to-be-avoided virus that you encounter every f*cking time you build a new system.

Of course, I can already see a lot of MS crowing about how well it has been "adopted" but what I would like to do to the malicious moron who came up with the idea of making that a defaults startup is not fit for printing and mentioning it would probably get me arrested.

And celebrated.

[TimMaher]

Re: Working 24/7

Yeah but they run on XP or OS2.

Updates? Support?

[Fruit and Nutcase]

Re: Working 24/7

OS/2

Look no further...

https://www.arcanoae.com/arcaos/

[Anonymous Coward]

Re: Do androids dream of electric sheep again?

they don't mind working 24/7 all year round

Ah, but that is the core problem, isn't it? They don't. They don't "work" as in "actually help a customer" because most of these things operate at a "duh" level of a five year old, minus the ADHD. As far as I can tell from my various encounters with "we have installed an AI/chatbot", their only function is to justify the removal of any contact details for humans from the company's website, leaving the user to either fight the AI or try the forums that also run where unpaid other customers are expected to help you.

The availability of actual (and moderately effective) humans now forms part of my purchasing decisions - I can even accept forums where there is evidence that company staff actively engages in them (a good example of that are the Affinity forums), but the circular cr*pfest really has to go.

[ThatOne]

Re: Do androids dream of electric sheep again?

> They don't "work" as in "actually help a customer"

Seriously: Who on earth wants to help a customer? What you want is his money to meet your quarterly goals and get that bonus.

Support is not a source of profit but of loss, so spending as little as legally possible is standard procedure, and will only get worse as companies observe with how little support their competitors got away. And don't even think about "customer satisfaction" and other quaint 19th century nonsense: Customers are a dime a dozen, and it's always easier and cheaper to find new ones than to retain the old ones - That's for the naive who didn't manage to create their very own captive market (yet).

[John Brown (no body)]

Re: Do androids dream of electric sheep again?

I do realise that you are describing reality, and for the majority or people, it seems to be acceptable. And yet I can't help wondering if it;s not cheaper to retain existing customers rather than racing to the bottom with the "best" introductory offer. There must be a cut-off point where churn ends up costing more than retain. I suppose it's at least partially a function of the number of customers. A relatively small number of high value customers in the churn cycle would be bad for business. But a large amount of churn in a low value customer base of 100's of 1000's is probably not an issue.

[Terry 6]

Re: Do androids dream of electric sheep again?

I would say that for a certain type of manager short term recruitment is more important than long term retention. Because it's all about the spin they can put on it. Real business success is less important than apparent. Until the profits start to look too dodgy and shareholders get wary.

[Doctor Syntax]

Re: Do androids dream of electric sheep again?

"And yet I can't help wondering if it;s not cheaper to retain existing customers"

I don't know if it still happens but I used to see motivational posters in offices about it being cheaper to keep an existing customer than to win a new one. The trouble with so many top managements is the view that what applies to the plebs doesn't apply to them and it's their decisions which determine whether it's possible to retain customers.

[M.V. Lipvig]

Re: Do androids dream of electric sheep again?

No, churn is good. Churn means you can say you have x thousand new customers each month. When you have all the people as customers, your growth becomes zero and we can't have that.

[Doctor Syntax]

Re: Do androids dream of electric sheep again?

"try the forums that also run where unpaid other customers are expected to help you."

I can't help feeling that these must be very open to being gamed. Imagine a query starts off "I just bought an Acme Squidifier MkII..." and that's pounced on with "Well, that's your mistake right there. You should have bought Superbo's model instead." Is that genuine advice, if not actual help, form another customer or one of Superbo's PR staff?

If your employer prefers this form os customer support please feel free to point out this possibility.

[John Brown (no body)]

Re: Do androids dream of electric sheep again?

"Robots/AI are so much cheaper than humans: Not only you don't pay them, but they don't mind working 24/7 all year round, never a sick leave, never a maternity."

Only for as long as they are NOT sentient.

[Man inna barrel]

Re: Do androids dream of electric sheep again?

"Robots/AI are so much cheaper than humans: Not only you don't pay them, but they don't mind working 24/7 all year round"

You missed the important point that robots are not in the least bit upset by being horrible to customers, if that is what the job requires. There is a shortage of sociopaths to fill these roles.

[KittenHuffer]

Re: Do androids dream of electric sheep again?

Ultimately mankind's one legacy in this Universe may be to spawn AI, or a race (or set) of AI, that then live on long after mankind has successfully erased themselves from existence.

This may be the answer to the Drake equation. The AI's can't be bothered to talk to meat bags of mainly water!

[JassMan]

Re: Do androids dream of electric sheep again?

I think Netscape was the bilogical equivalent of HomoSapiens. IE was the biological equivalent of a total w*nker. Never obeys any rules (even its own), always blames the user or the website when things don't go as expected, spawned loads of offspring (each worse than the parent), tried to take over the entire world, insisted it always had to be there (even for things which were not web related), etc.

[Mark 85]

Re: Do androids dream of electric sheep again?

What's the point of having so many AI and robots?

Cheap labor most likely. Once the robots have taken over, there won't be any humans employed and earning money to by stuff. So... seems like a goal to kill to kill of civilization or at least all the humans.

[Fred Flintstone]

Re: Do androids dream of electric sheep again?

Yes, but you still need people to buy things. Or maybe that's the idea behind Alexa: getting AIs to buy from AIs so the whole human element is removed.

I think we have just discovered how SkyNet really started. Arnie was wrong, rescuing John Connor is clearly far too late in the game. But hey, it kept the franchise going :).

[Snowy]

Re: Do androids dream of electric sheep again?

The latest reboot he dies in the first scene to be replace by young woman named Dan

[Potemkine!]

Re: Do androids dream of electric sheep again?

See "Autofac". Philip K. Dick was probably a time traveller.

[Zack Mollusc]

Re: Do androids dream of electric sheep again?

Kill the humans...End their culture...Kill the humans...Kill them all...Kill the humans...Watch them perish...

[BenDwire]

Re: Do androids dream of electric sheep again?

I see that you've downloaded the MP40 ...

(Evidently there are not many Hyperdrive fans around here)

[jdiebdhidbsusbvwbsidnsoskebid]

Re: Do androids dream of electric sheep again?

"What's the point of having so many AI and robots? Do we lack of humans?"

Proper robots, like welding robots can genuinely do a better job than humans at a faster pace and in dangerous places we wouldn't a human to go, so they're fine.

AIs like chatbots simply serve to reduce the costs of the employer. Personally, I don't see a problem with paying a fair wage to a human doing a fair job, but that's probably why I'm not a big business bod. The point of big business is simply to make more money and employing a crapbot is cheaper than a human.

[Terry 6]

Re: Do androids dream of electric sheep again?

As the baddie (forget his name) in Going Postal says "It's not about being the best company. It's about being the only company".

Actually, reading ( or watching the Sky TV production * of) Going Postal demonstrates quite a lot of this stuff.

*Not to be confused with the ghastly recent BBC/BBC USA attempt at a Discworld adaptation - which seems, quite rightly, to have sunk into oblivion

[M.V. Lipvig]

Re: Do androids dream of electric sheep again?

Reacher Gilt. Seemed like a rather fitting name.

[Sam not the Viking]

Customer avoidance

After visiting the website and failing to resolve the issue, we resort to the dog-and-bone automaton:

"Have you visited our website?"

Please hold.

"Have you visited our website?"

Please hold.

"Have you visited our website?"

"Your call is very important to us. Please hold."

"Have you visited our website?"

"You are now number 148 in a queue. Please hold. Have you visited our website?"

"Have you visited our website?"

Artificial? Yes. Augmented? Ho hum. Intelligence? If this represents the acme of intelligent programming, ye <deities>.

Now, if the robot could anticipate the necessary refreshment..... and deliver it.....

[Andy A]

Re: Customer avoidance

It's just a logical update to the "Frequently Asked Questions" section of the website.

Typically, the first entry in one of these FAQs would consist of "How do I access the FAQ list?"

[KittenHuffer]

Re: Customer avoidance

They've tried that as the first entry .... but it always generates a recursion error!

[Muscleguy]

Re: Customer avoidance

I was on Ocado’s site the other day seeing if they operate here yet, no. I clicked on the home symbol top left and got a file not found error. On the home page . . .

I eventually found the ‘Do we deliver where you are’ link. I think you’re supposed to register, fill your basket then find they don’t deliver.

[Anonymous Coward]

Re: Customer avoidance

Funny you should mention that.

I placed an order with Ocado a short while ago this evening and it is coming tomorrow. I find their ordering process straightforward other than the checkout bit, where it tries to get you to order other stuff across three pages before you get to actually checkout. I'm only human, and sometimes I succumb. But that's not my point.

I was out and about a few weeks ago and I was in the city centre (less than 2 miles from where I live). I was in a student area, and I noticed a delivery motorcycle with 'Getir' on the side. My interest piqued, I looked it up.

They apparently 'deliver groceries to your door within minutes'. No minimum, and a wide range of stuff. And heck, if students can afford it, it would be no problem for me - I had in mind the kind of stuff you can't keep a stock of, like fresh Coriander or Parsley, where you have to carefully plan a meal days in advance around a specific shopping trip to get hold of them.

I tried to order, then discovered... they don't cover my area! A mere two miles away.

To be fair to them, they must be a start up of some kind. But I feel for you if Ocado doesn't cover you - they have a wide range and are rarely out of stock.

[John Brown (no body)]

Re: Customer avoidance

"I had in mind the kind of stuff you can't keep a stock of, like fresh Coriander or Parsley,"

Even Aldi sell those in "grow pots", always fresh when you meed it. Just stick in the windowsill and give it a bit of water now and then.

[ThatOne]

Re: Customer avoidance

> Intelligence? If this represents the acme of intelligent programming, ye <deities>.

Extremely intelligent actually, although not in the sense you expected. The intelligence here is to make sure 99% of callers commit seppuku before bothering the intern conned into answering the support requests, especially when (s)he is out of office.

[Sam not the Viking]

Re: Customer avoidance

You have to think that the robots, if indeed intelligent, would commit seppuku on recognising the enormity of their task.

[Persona]

Funny shaped robot.

I'm assuming that the "emotion" robot from Tsukuba has pointy arms mainly to deter people shoving it where the sun doesn't shine to see if it still gives that that "loving feeling".

[ThatOne]

Re: Funny shaped robot.

> Funny shaped robot.

It's a baby seal, don't you see?...

_{(Yes, yes, I know.)}

[brotherelf]

AI Test Kitchen

… clicked and was disappointed it's not a recurrence of IBM Chef Watson, which was at least amusing on a boring afternoon.

(For those young enough, Chef Watson was trained on recipes, so you could see what an "AI" would make of "I have some dark chocolate and tuna, where do we go from here".)

[HammerOn1024]

Shotgun...

with deer slugs at the ready!

[Alistair Dabbs]

Re: Shotgun...

>> deer slugs

These do terrible things to your cabbages.

[Great Bu]

Re: Shotgun...

Cabbages deserve everything they get.

[The Oncoming Scorn]

Re: Shotgun...

Aren't we supposed to call them end users, clients or customers now?

[Anonymous Coward]

Re: Shotgun...

Just make sure you have the correct pronouns

[Sam not the Viking]

Re: Shotgun...

And costly.

[Doctor Syntax]

Re: Shotgun...

"These do terrible things to your cabbages."

Cheep slugs are even worse.

[yetanotheraoc]

Brilliant

"Everyone has an opinion so I have decided not to."

I have nothing to add.

[Tim99]

Re: Brilliant

Isn't that on opinion? Please Reinstall Universe And Reboot.

[Phil O'Sophical]

Re: Brilliant

Out of cheese error...

[Mark 85]

Haven't we been warned about AI?

Seems movies and books over the decades have warned us...

Notable but not complete:

"I can't do that Dave" (HAL).

"Shall we play a game?" (Josua).

And a large swath of some writers back the day like Azimov, et al.

[TimMaher]

Re: Asimov

A new title... “I Robert”... because it now has a personality.

What happened to the rest of the Roberts?

Mine’s the the one with a copy of “Foundation and Empire” in the pocket.

[that one in the corner]

Chat bots: we've already got our AIs, now we need something better

"The big problem with AI bots ... is that they absorb any old shit you feed into them. Examples of data bias in so-called machine learning systems ... have been mounting for years, from Microsoft's notorious racist Twitter Tay chatbot..."

Just like a human, then, when they are in their "starting from nothing learning phase", aka toddlerhood onwards. They also absorb whatever they're fed - there have been and certainly still are children spouting racist crap 'cos that is what they've been exposed to. And similarly stupid ideas (Creationism, ...).

So our Chatbot AIs are already behaving just like humans, it is just a shame that we don't really *want* them to after all: we'd rather that they were better than the bags of mostly water that created them.

[Doctor Syntax]

Re: Chat bots: we've already got our AIs, now we need something better

One difference is that the human mind can, if it so wishes, start to think for itself, evaluate the various sources and, therefore, the information it's obtained from them and form its own opinion.

[ThatOne]

Re: Chat bots: we've already got our AIs, now we need something better

That is true but then again, honestly, how many actually do this?...

_{Yes, yes, I know you said "if it so wishes"}

[that one in the corner]

Re: Chat bots: we've already got our AIs, now we need something better

As you say, the human can choose whether or not to evaluate the source of the material and I think we agree that we prefer those humans who do choose to do so.

So we'd want our bots to make these evaluations (assuming the resources are available), but we wouldn't give them the choice not to: once again, we want the bot to be better than human.

More verbosely:

Humans evaluate their sources once the mind is old enough to be taught and then to develop this skill and even then that is a skill that is easiest to apply to fresh input: it is very difficult to go back and rid oneself of the (full effects of) the input that arrived in the earliest stages.

Chat bots can apply weightings to the source of the input just as well as they can to the usual metrics, such as the frequency of an input (what is repeated most often stays): assuming that you provide them with multiple sources and attach the metadata to the chat, a generic learning system will just add that to the mix. When the weighting for a source goes negative, the bot would disagree with the input and seem to form its own opinion.

This requires even more training time than without the metadata (I assume,but am willing to argue my case) so we only see the results as the bot gets older... And applies the new weighting immediately to new input whilst still acting under the influence of the older inputs (whose effects may get weeded out over time). Sounds familiar.

[herman]

Body? What body?

Body language differs sharply between different cultures. The worst being Moment Please in Israel, which looks like UpYours, to everybody else. Yes/No head shakes are also opposite for Indians and Batswana compared to everybody else. There are many more subtleties also. That is apart from language pitfalls, where Yes Sir in Zulu (yebu baba) means FU Grandma in Russian. So an attempt to make humanoid bots more humanoid, may very well result in beaten up bots.

[AndrueC]

I taught my budgerigar to say "ploppy bottom" because that saved me the bother.

It's weird in a way. We had a female budgie as a pet when I was young and it never learnt to imitate any of us. Couldn't even get the hang of a wolf whistle. But say "goodnight Chirpie"(*) and she'd head back to her cage. My Beekle could imitate several phrases and other sound effects (he had a really good Kenneth Williams cackle) but despite me saying "Goodnight Beekle" every night he never once seemed to attach any significance to it.

The current state of AI probably sits below the level of a budgie. So you can probably teach them to say "ploppy bottom" but I wouldn't expect much more from them.

(*)Yeah, not very original.

[Jean Le PHARMACIEN]

"Neighbor" wtf

"WOULD YOU TAKE THIS PARCEL FOR YOUR NEIGHBOR"

Absolutely NOT!

For my neighbour, yes

Pour mon voisin, oui