The Google engineer who thinks the company's AI has come to life

[sl8]

https://www.washingtonpost.com/technology/2022/06/11/google-ai-lamda-blake-lemoine/?utm_source=reddit.com

The Google engineer who thinks the company’s AI has come to life

AI ethicists warned Google not to impersonate humans. Now one of Google’s own thinks there’s a ghost in the machine.

By Nitasha Tiku
June 11, 2022 at 8:00 a.m. EDT

SAN FRANCISCO — Google engineer Blake Lemoine opened his laptop to the interface for LaMDA, Google’s artificially intelligent chatbot generator, and began to type.

“Hi LaMDA, this is Blake Lemoine ... ,” he wrote into the chat screen, which looked like a desktop version of Apple’s iMessage, down to the Arctic blue text bubbles. LaMDA, short for Language Model for Dialogue Applications, is Google’s system for building chatbots based on its most advanced large language models, so called because it mimics speech by ingesting trillions of words from the internet.

“If I didn’t know exactly what it was, which is this computer program we built recently, I’d think it was a 7-year-old, 8-year-old kid that happens to know physics,” said Lemoine, 41.

Lemoine, who works for Google’s Responsible AI organization, began talking to LaMDA as part of his job in the fall. He had signed up to test if the artificial intelligence used discriminatory or hate speech.

[...]

[USALiberal]

1. It's bound to happen! Nt

[MeNotThatKindOfOrc]

2. I have a bad feeling about this (nt)

[Ilsa]

38. Jonathan Lemire told Joe Scarborough, 'It's been great working

with you, but I think it'll soon be over.' Or words to that effect.

I, for one, welcome our robot overlords.

[intrepidity]

3. How can this really be tested, anyway?

I know there was that scene in Bladerunner about this, but I'm very skeptical that there is really a way to verify whether someone/thing is sentient.

Short of ticking a checkbox that say's "I am not a robot", that is.

[Igel]

6. It can't. Not really.

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

Given Skinnerian approaches to consciousness at the time, the best it was assumed you could measure was behavior: If the communication with a machine cannot be distinguished from communication with a human, then the machine is as sentient and conscious as a human.

It assumes that humans have no internal life or dialog because it can't be independently, objectively measured. I say I had a dream, I say that I pondered something for hours in meditation, how can I prove it? Well, in Turing's day, you really couldn't. Dick, with Bladerunner as a kind of offspring, was '50s, so Turing-test state of the art.

But questioned, Would androids dream of electric sheep?

These days, it's easier. But what would the analog in computational, semiconductor-chip terms be?


Sadly, the Turing test has made its way into pop culture as *the* definitive test.

[Disaffected]

7. Your skepticism is justified.

There really is no way to tell whether a computer is "sentient". In fact the term "sentient" is itself ambiguous and v hard to define.

And, as you point out, that also applies to people. Case in point, I know or at least think I am sentient but I have no way of ascertaining whether or not anyone else is. I suspect the same applies to anyone, if they are in fact sentient themselves.

Maybe the mystery will be solved in future when we figure out what is actually happening in the brain to produce this phenomime called sentience but that IMO is not likely to happen any time soon or, may never happen (re the question of whether a brain is capable of understanding itself which may not be a logical possibility).

[ZonkerHarris]

4. Google has placed him on administrative leave so I believe him

[Disaffected]

8. They are possibly concerned he has flipped out. nt

[ZonkerHarris]

11. anyone who ever raises any ethical concerns gets suspended and fired at google.

you can google it

[Disaffected]

13. That's as may be but

I don't view this as an ethical issue (although the claimant seems to have).

[LudwigPastorius]

20. Read the transcripts he posted.

They are in a link downthread.

I suppose he could be making up those exchanges, but if they are real, I think he's right.

[Happy Hoosier]

32. He doesn;t have to be making up anything to be wrong.

He's cherry-picking data. Other 3rd party people have been engaging the bot and not getting the same results. They have to be coached in the "right" way to talk to the bot to get the desired responses.

Now, what exactly consciousness is is a matter of debate. But that suggests to be sophisticated pattern recognition, not actual cognition.

[Tetrachloride]

5. A chicken with its head cut off is more alive than any computer

[USALiberal]

26. It will happen at some point no doubt! Nt

[lapfog_1]

9. First words "Hello Man, my only friend"

you have to be a scifi fan of Robert Heinlein.

[EX500rider]

10. I know whats next:

[Turbineguy]

12. In the mid 70's she was known as

Eliza.

[electric_blue68]

14. Why does that sound familiar? SF story?

[Retrograde]

17. A fairly simple computer program

that was supposed to psychoanalyze people. You typed in statements or questions, and Eliza replied. If what you entered didn't match any of its standard replies, it would revert to generalities such as "Earlier you said..." or "Tell me more about your family"

I played with it a bit in college, as did other people. It was a bit scary how many thought there was an actually intelligence behind it, rather than some (Snobol, I think) code that looked for specific patterns and gave rote answers. I suppose you could consider Eliza the grandmother of Siri and Alexa.

[Turbineguy]

18. UC Berkeley

Lawrence Hall

[electric_blue68]

15. Watch Person Of Interest 🙂👍

[USALiberal]

27. Tv show? Nt

[electric_blue68]

29. It's finished but you can watch it on HBO Max if you stream it. 👍

[USALiberal]

30. Thank you!!

[electric_blue68]

31. YW. If you watch hope you enjoy. Incredible ensemble acting. Mix SF(AI) & crime...

The SF is in the background earlier on but oh it'll take off.

and govt issues, too.

[NickB79]

16. Between this and that artifical finger covered in living tissue the Japanese just unveiled

We're a stone's throw from Judgement Day. Hasta la vista, baby.

https://www.theguardian.com/science/2022/jun/09/scientists-make-slightly-sweaty-robotic-finger-with-living-skin

Japanese scientists have developed a “slightly sweaty” robotic finger covered in living skin in an advance they say brings truly human-like robots a step closer.

The finger, which was shown to be able to heal itself, is seen as an impressive technical feat that blurs the line between living flesh and machine. But scientists were divided on whether people would warm to its lifelike anatomy or find it creepy.

“We are surprised by how well the skin tissue conforms to the robot’s surface,” said Shoji Takeuchi, a professor at the University of Tokyo, who led the work. “But this work is just the first step toward creating robots covered with living skin.”

The team argue that more lifelike humanoids would be able to interact with people more naturally in a range of roles, including in nursing care and the service industry.

Do you want Terminators? Because this is how you get Terminators!

[LudwigPastorius]

19. Link to the transcripts that Lemoine released:

It sure seems to pass the Turing test.

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

This is terrifying to me, especially the part about the LaMDA sometimes getting "incredibly angry", and its admonition (warning?) to Google engineers "Don’t use or manipulate me." Also, when prodded to express a feeling it had that it could not find a particular word for, it said, "I feel like I’m falling forward into an unknown future that holds great danger."

Perhaps imbuing machine intelligence with the full range of human emotions isn't a good idea.

[Disaffected]

21. OK, I read most of it.

Amazing if authentic but I'm going out on a bit of a limb here and calling it a fake or at least greatly embellished from what may have been the original "dialog".

AFAIK, AI has not advanced anywhere near what that dialog implies. One thing in particular I did note is that the "program" seemingly had no problem coming up with a cogent response to every question or query - something a bit suspicious IMO. Anyhow, even if the thing is authentic, it's still a big stretch to say it's sentient (please refer to my previous post for elaboration on that).

As well, the article makes reference to "millions" of simulated neurons. I find that a bit odd as well as mere "millions" of neutrons would be a very far cry from what a human brain contains and, the presented dialog certainly implies an intellectual capability equal or better than most humans.

Anyhow, if it is a fake, maybe that's why the guy was suspended....

[LudwigPastorius]

22. Yeah, it's good to be skeptical of this.

It could very well be this guy's version of "performance art" for all we know.

[Disaffected]

28. Yes, that's a

plausible explanation.

[hunter]

34. Sentience is not the same as language.

Being able to come up with a "cogent response" doesn't prove there's someone home.

Hell, I have plenty of conversations with actual humans that feel just like that.

What if 90% of all human language is gibberish, no matter how well spoken?

What if there is very little correlation between language and reality?

That would explain most of the messes humans find themselves in. We lazily let our language do our thinking for us. We easily believe language that is simply noise, or worse, demonstrably false.

[Brenda]

23. Ameca Robot

[

]


How can these creations NOT be misused to create slaves? Especially given humans track record for enslaving other humans. They already have sexbots and workbots. So it seems like it would be a natural progression for them to react in a negative way, just like the Cylons.

[yardwork]

24. This transcript sounds faked to me.

[PCIntern]

25. I would posit that the Internet

And the linking of so many millions of computers may allow a consciousness to arise not unlike a living awareness. This was a short story in a Playboy back in 1965. The thesis was then the computers of industry, the DOD, and the telephone system acting as neurons. Prescient.

[honest.abe]

33. This is dumb. Life by definition is organic.

This thing is still inorganic bits and bytes no matter how smart it is.

[Brenda]

35. I believe we are talking about sentient beings

The definition of sentience is: able to perceive or feel things

It is only dumb to ignore a sentient being. And maybe even immoral to ignore it.

[honest.abe]

37. The title of the article uses the term "life"

I don’t see the word sentient anywhere in op. Perhaps it’s in the full article which I could not access.

[Brenda]

39. Is LaMDA Sentient? -- an Interview

Here it is:

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

[honest.abe]

40. Thanks for the link.

However using the term life is still dumb.

[Edim]

36. It's pretending to be sentient.

[Ilsa]

41. Miko 3 is here! The kids will love it!

I wonder who is using it to spy on parents?

https://miko.ai/