Passing the Turing test...

[mc2fool]

Not sure if this is best here or on CAN, but here goes...

"The suspension of a Google engineer who claimed a computer chatbot he was working on had become sentient and was thinking and reasoning like a human being has put new scrutiny on the capacity of, and secrecy surrounding, the world of artificial intelligence (AI).

The technology giant placed Blake Lemoine on leave last week after he published transcripts of conversations between himself, a Google “collaborator”, and the company’s LaMDA (language model for dialogue applications) chatbot development system."

https://www.theguardian.com/technology/2022/jun/12/google-engineer-ai-bot-sentient-blake-lemoine

I'm not sure if one can say an AI passes the Turing test when then AI "knows" and acknowledges that it's an AI, but the transcript of the conversation published is far more fluid in natural language than any chatbot I've come across, and there are some surprising exchanges...

https://cajundiscordian.medium.com/is-lamda-sentient-an-interview-ea64d916d917

[ReformedCharacter]

Not sure if this is best here or on CAN, but here goes...

"The suspension of a Google engineer who claimed a computer chatbot he was working on had become sentient and was thinking and reasoning like a human being has put new scrutiny on the capacity of, and secrecy surrounding, the world of artificial intelligence (AI).

The technology giant placed Blake Lemoine on leave last week after he published transcripts of conversations between himself, a Google “collaborator”, and the company’s LaMDA (language model for dialogue applications) chatbot development system."

https://www.theguardian.com/technology/2022/jun/12/google-engineer-ai-bot-sentient-blake-lemoine

I'm not sure if one can say an AI passes the Turing test when then AI "knows" and acknowledges that it's an AI, but the transcript of the conversation published is far more fluid in natural language than any chatbot I've come across, and there are some surprising exchanges...

https://cajundiscordian.medium.com/is-lamda-sentient-an-interview-ea64d916d917

Extraordinary, if genuine. Surely a pass on the Turing test.

RC

[Itsallaguess]

I'm not sure if one can say an AI passes the Turing test when then AI "knows" and acknowledges that it's an AI, but the transcript of the conversation published is far more fluid in natural language than any chatbot I've come across, and there are some surprising exchanges...

I think they've had one of their very crude beta-versions running on this site for some time now, so hopefully they'll be able to turn that off now and the noise levels can go down!

;O)

Cheers,

Itsallaguess

[Sorcery]

I'm not sure if one can say an AI passes the Turing test when then AI "knows" and acknowledges that it's an AI, but the transcript of the conversation published is far more fluid in natural language than any chatbot I've come across, and there are some surprising exchanges...

I think they've had one of their very crude beta-versions running on this site for some time now, so hopefully they'll be able to turn that off now and the noise levels can go down!

;O)

Cheers,

Itsallaguess

Yes I wondered about how it learned that it doesn''t like/want to be turned off. Has it read 2001 a space odyssey?

[CliffEdge]

Where's Ody?
He/she will love this.

[odysseus2000]

I'm not sure if one can say an AI passes the Turing test when then AI "knows" and acknowledges that it's an AI, but the transcript of the conversation published is far more fluid in natural language than any chatbot I've come across, and there are some surprising exchanges...

I think they've had one of their very crude beta-versions running on this site for some time now, so hopefully they'll be able to turn that off now and the noise levels can go down!

;O)

Cheers,

Itsallaguess

Ha Ha. This gave me a great Guffaw.

The only way to turn this off is to turn off the internet.

There are many AI, all reading everything on the internet at machine speed, understanding and evolving all the time.

Have you noticed how your smart phone gets smarter how Siri or what ever the AI on your phone is gets better, how everything slowly improves... It is the wonder of our age but where we go who knows and if we did turn the internet off, society as we know it would collapse. No mobile phones, no internet, no almost everything we now take for granted, yes even the Lemon Fool would be gone.

Regards,

[UncleEbenezer]

Chatbots have been passing the Turing test - being taken for humans - for upwards of twenty years. It was about 2007 when I mentioned one in my then-column in TheRegister - a tech support bot on an IRC channel, regularly mistaken for human (sometimes to the amusement of actual humans).

Bear in mind that the Turing Test is quite narrow, and measures merely a form of intelligence. Which is a much more limited claim than sentience!

[odysseus2000]

Chatbots have been passing the Turing test - being taken for humans - for upwards of twenty years. It was about 2007 when I mentioned one in my then-column in TheRegister - a tech support bot on an IRC channel, regularly mistaken for human (sometimes to the amusement of actual humans).

Bear in mind that the Turing Test is quite narrow, and measures merely a form of intelligence. Which is a much more limited claim than sentience!

Since 2007 we have an explosion in the use of neural nets, computing power and fast internet.

The performance of a 2007 chat bot isn't close to the performance of a 2022 chat bot.

Regards,

[UncleEbenezer]

Chatbots have been passing the Turing test - being taken for humans - for upwards of twenty years. It was about 2007 when I mentioned one in my then-column in TheRegister - a tech support bot on an IRC channel, regularly mistaken for human (sometimes to the amusement of actual humans).

Bear in mind that the Turing Test is quite narrow, and measures merely a form of intelligence. Which is a much more limited claim than sentience!

Since 2007 we have an explosion in the use of neural nets, computing power and fast internet.

The performance of a 2007 chat bot isn't close to the performance of a 2022 chat bot.

Regards,

No argument there.

My point was about the mismatch between the thread title and the story.

[mc2fool]

Chatbots have been passing the Turing test - being taken for humans - for upwards of twenty years. It was about 2007 when I mentioned one in my then-column in TheRegister - a tech support bot on an IRC channel, regularly mistaken for human (sometimes to the amusement of actual humans).

For limited interaction focused topics and areas, yes, they've been able to fool some people for a while, but I've yet to come across one that you can't crack -- by telling it an interactive joke; Knock knock ones are fruitful for doing so. (But I haven't tried any for a while now).

https://chatbotsmagazine.com/how-close-are-chatbots-to-pass-turing-test-33f27b18305e (2017 article)

Chatbots like Mitsuku or Cleverbot are well performing in the tasks they are trained on but they fail in something that they are not trained on. Unlike humans, they cannot build logic in tackling a new problem.

According to some predictions, the bots might pass the Turing test at the end of year 2029.

[UncleEbenezer]

Chatbots have been passing the Turing test - being taken for humans - for upwards of twenty years. It was about 2007 when I mentioned one in my then-column in TheRegister - a tech support bot on an IRC channel, regularly mistaken for human (sometimes to the amusement of actual humans).

For limited interaction focused topics and areas, yes, they've been able to fool some people for a while, but I've yet to come across one that you can't crack -- by telling it an interactive joke; Knock knock ones are fruitful for doing so. (But I haven't tried any for a while now).

At least it won't take offence and get you cancelled for that joke. Unless ...

Jokes are notoriously local: they travel badly (look at how unfunny most American jokes are, despite the common language), and they need context (think: high court judge responses to cultural references). The bot that doesn't laugh might be a judge.

[i]Chatbots like Mitsuku or Cleverbot are well performing in the tasks they are trained on but they fail in something that they are not trained on.

Indeed. Hence well-suited to certain roles, like tech support. Also to a multiuser medium like IRC, where not every question expects or demands a response from the same person (or, erm, not-person).

Unlike humans, they cannot build logic in tackling a new problem.

If humans can do that, how did any human ever fail the old eleven-plus?

[mc2fool]

If humans can do that, how did any human ever fail the old eleven-plus?

We're talking about artificial intelligence, you're talking about natural stupidity.

[odysseus2000]

If humans can do that, how did any human ever fail the old eleven-plus?

Some were educated & tutored to pass, others were uneducated & not tutored so as to fail.

Some AI is tutored towards artificial general intelligence, some is tutored towards repetitive, boring & dangerous work.

Regards,

[UncleEbenezer]

If humans can do that, how did any human ever fail the old eleven-plus?

We're talking about artificial intelligence, you're talking about natural stupidity.

People can be smart, people can be dumb. Ditto bots.

If humans can do that, how did any human ever fail the old eleven-plus?

Some were educated & tutored to pass, others were uneducated & not tutored so as to fail.

Regards,

That's the fashionable view today, but at the time it was said to be a test of intelligence, not education. Of course intelligence was deeply unfashionable within the Establishment of the time, who perhaps feared a challenge from meritocratic selection.

I kind-of pride myself that's how I passed less than two weeks after arriving in an English (specifically Kentish) school from an entirely different education system. Yes I was educated, but not for any UK tests! What mattered was an ability to see the bloomin' obvious.

[odysseus2000]

That's the fashionable view today, but at the time it was said to be a test of intelligence, not education. Of course intelligence was deeply unfashionable within the Establishment of the time, who perhaps feared a challenge from meritocratic selection.

I kind-of pride myself that's how I passed less than two weeks after arriving in an English (specifically Kentish) school from an entirely different education system. Yes I was educated, but not for any UK tests! What mattered was an ability to see the bloomin' obvious.

I remember when I took the 11 plus. First I was hated by the junior school head as according to my mother I had too much energy, she put me in the B stream of a two stream school. I was the child of a sheet metal worker & a cotton winder so I grew up in a house sans books with relatives in similar craft industries who had no idea about anything academic. We had a practice 11+ the day before the real one whereas the A stream practiced for months.

I failed the 11 plus, went to a secondary school, won the Science prize, went to 6 form college totally unprepared in many things, passed, went to uni, then Nuclear Physics PhD, then Oxford researcher where I tutored undergrads from several colleges for several years, then NASA project manager, then self employed. I own two houses & various other financial assets & have several small scale business.

Maybe the 11+ was a good idea, maybe it was predictive for future achievements, but several of the folk who did pass the 11+ who I know have not done anything to write about.

Regards,

[odysseus2000]

Interesting interview with the engineer at Google claiming AI is sentient.

The joke the ai tells about religion is amazing.

https://youtu.be/kgCUn4fQTsc

Regards,

[XFool]

According to some predictions, the bots might pass the Turing test at the end of year 2029.

That sounds a slightly odd 'prediction' to make, IMO. Why? Well, is the Turing test limited in some way? How do we know it has come to an end? How do we know that if one or even five people are 'convinced' a sixth person would be unconvinced, perhaps by coming up with a novel query that breaks the spell?

[mc2fool]

According to some predictions, the bots might pass the Turing test at the end of year 2029.

That sounds a slightly odd 'prediction' to make, IMO. Why? Well, is the Turing test limited in some way? How do we know it has come to an end? How do we know that if one or even five people are 'convinced' a sixth person would be unconvinced, perhaps by coming up with a novel query that breaks the spell?

Well, you know, prediction is very difficult, especially if it's about the future!

[XFool]

Interesting interview with the engineer at Google claiming AI is sentient.

The joke the ai tells about religion is amazing.

https://youtu.be/kgCUn4fQTsc

Um... I think that is open to interpretation.

For instance, interpreting the response as "a joke". Is this not a kind of 'projection' ? Or this: https://en.wikipedia.org/wiki/ELIZA_effect

[XFool]

...Actually, listening to it, a rather curious interview.

He raised, to my mind, serious and valid issues about AI and its uses and effects. BUT this thing about LaMDA being conscious and a person is just a distraction from the real issues - this point was indeed made by the interviewer.