as predicted

[lucy24]

Scanning my bookshelves for something to read, I landed on The Practice Effect from 1984. Early in the first chapter, before the actual plot gets started, the narrator tells us that it was established “back in 2024” that Artificial Intelligence was a dead end. Later in the chapter, he drops the further detail that the record for an AI conversation is six minutes before the human at the other end realizes they're talking to a robot.

Oh, well.

[mack]

Wow, someone got the timing right :-)

Mack.

[Martin Potter]

I surely hope that it turns out to be true.

[LifeinAsia]

Considering that these days it's getting so easy to realize the human at the other end has no intelligence, recognizing AI on the other end should be a snap.

[engine]

With enough predictions out there, some are going to get it right.
ML has been offered for some time on those stupid chat helpers on websites. Considering the advances, I engaged with one of the real stupid ones at the weekend. If that's the best, we're all going to be fine.

[blend27]

The Practice Effect from 1984: Good Book!

RE: “back in 2024”

Are we there yet?
--------------------------
@LIA(LifeinAsia now u are)> ... so easy to realize it is the human(System Bless The Bot), recently I got bugged down from a chut-bot(misspelled on purpose ) by a Hosting Company that I've been dealing fro the past 10+ years ......SAyING, let me get a hUMAN to help you out....

Now something is wrong with my KeyBoard...(Darn Sweet Tooth, Oh wait It is Blue now days ain't it?)

[ronin]

I've increasingly arrived at the conclusion that, although popularly referred to (since 2022, if not before) as AI, Large Language Models (LLMs) barely really qualify as Artificial Intelligence at all.

LLMs are pattern recognition / pattern reproduction engines, trained to analyse the input of a stream of symbols and then output an appropriate stream of symbols.

They do this, more often than not, extremely impressively. But they do it with zero semantic comprehension.

It's tempting to infer emergent properties - modest ones at the very least - which have (somehow) resulted from the colossal datasets on which the LLMs have been trained.

But I don't think there are any.

It's simply that we have a tendency to anthropomorphise everything. We see faces in clouds, don't we?

In this article from May 2023, Michael Wooldridge describes LLMs:

Large language models (LLMs) like ChatGPT are essentially a very sophisticated form of auto-complete.

Source: [bigthink.com...]

We might argue that, although novel, LLMs are nothing more to language than calculators are to arithmetic.

Using a calculator, we can multiply three six figure numbers in fractions of a second, but we don't then refer to calculators as artificial intelligence, do we?

Perhaps (by analogy with the term calculator) rather than referring to LLMs as Artificial Intelligence, we might regard these tools as articulators.

I suspect the most useful function of articulators will turn out to be as a User Interface (UI) for human / computer interaction.

We've had smart speakers with Voice UI for a couple of years now. But, to date, the language recognition software within a smart speaker may comprehend only a fairly limited set of instructions, and often even these need to be explicitly formulated.

That software cannot act as a interpreter, translating free-form human-spoken instructions into sophisticated computer interactions and translating computer responses / outputs back into natural human language output.

LLMs can do this.

In conclusion, perhaps one of the primary functions of articulators in future (once we've all got over the novelty) will not be enabling computers to talk to humans, but enabling humans to talk to computers.

[lucy24]

Darn Sweet Tooth

Well, what do you expect when you eat ice cream directly above the keyboard.

[Martin Potter]

rather than referring to LLMs as Artificial Intelligence, we might regard these tools as articulators.

@ronin I believe you have hit the nail exactly on its head.

(Also your suggestion that many people have a tendency to anthropomorphize everything should be seriously considered by everyone. I see it all the time in various sciences, especially biology.)

[lucy24]

I see it all the time in various sciences, especially biology.

Biology gets tricky because one can just as easily fall into the opposite trap of assuming that some given property is unique to Homo sapiens, and then if you suggest it is more widely distributed you are accused of anthropomorphizing.

:: patting myself on the back because my fingers got it right on the first try ::

[Martin Potter]

then if you suggest it is more widely distributed you are accused of anthropomorphizing.

Yes, that too. And it seems to happen too frequently.

(Congratulations to your fingers. Give them a cold glass to hold.)

[explorador]

Later in the chapter, he drops the further detail that the record for an AI conversation is six minutes before the human at the other end realizes they're talking to a robot.

Oh, well.

It's funny... on another forum there is a lot of noise around AI, in the sense of "it's going to help me to post on the forum, and read/whatever other things people post", I couldn't even understand why regular people would need AI assistance to read or post on a forum...

There is plenty of AI examples around of pretty weird stuff, as if a 4 year old kid was trying to paint, write, or explain something. And this is what I find funny, I told someone (in a good way and intention) he often mess up his own explanations, as if it was a broken robot from Boston Dynamics, and dude laughed a lot because he agreed, the thing is... some people, despite being humans... really do things like 4 year olds, like AI does too.

[blend27]

I couldn't even understand why regular people would need AI assistance to read or post on....

Wait another couple of years till it gets to the point when a Kid(about 10 years old) if asked a simple question has a simple answer:

Q.(to Kid) --- Kid, why did you skip cleaning your room this afternoon?
A.(from Kid) --- Leave me Aloooone, Talk to Dorothy..., she knows - (AI App installed on Your Device).

Answered Instantly(AI), by Dorothy: ----- We are not in Kansas anymore!

[explorador]

@blend27, those are prophetic words :(

I'm seeing now some people interacting with bots on forums (bots placed there because the admin and the forum system allows it. The comments by AI don't actually contribute to anything, and it usually goes to "apologies, being a bot means I have limitations, I have no eyes, emotions, etc, I should have said that initially". And other forum members are posting what AI told them via their own accounts, and they act as if someone just shared great wisdom. Terrible times!

[explorador]

On another webmaster forum (in spanish, my native language) I'm seeing more requests for review "hey, can you tell me what you think of my new website?", and turns out it's 100% AI generated content, some free, some via paid accounts. It's 500% MFA (made for adsense), you can tell right away and the authors confirm it's for that purpose. When you read it... it's heavily cold, "a computer has diff parts, like the screen and the keyboard blah blah", terrible, no value. AI is improving a lot, still lacks lots of the human touch, but the results... are garbage, really.

[lucy24]

It's 500% MFA (made for adsense), you can tell right away and the authors confirm it's for that purpose.

Serious query: How much money can one possibly make from something so shallow? Do you rely on the new-site-SERP-bump, bank your fifty cents after the first month, and move on to the next site? And is it all done with registrars who charge $2.99 for the first year, after which you just let it expire?

[explorador]

Lucy24: Serious query: How much money can one possibly make from something so shallow?

Really, not much (if any), and if they do, it won't last. I think the whole questions goes in the form of "How much money can someone lazy be and deceiving make via deceiving others?"

Honestly... I've said this, literally, on that forum: it's just people trying to deceive others... just wait and you'll see them absolutely gone after a couple of months, or a year top, they just vanish (along with their websites) or end up banned, cancelled, taken down via DCMA, or... the more honest ones open new threads asking help to monetize their AI generated websites because they are going nowhere.

I consider this topic worth discussion, so I opened a new thread to discuss this specifically (MFA) keeping the topic here as intended originally by the author. Here goes: [webmasterworld.com...]

[blend27]

Lucy24: Serious query: How much money can one possibly make from something so shallow?

Perhaps another question would be in order then:

How much money GOOG make by promoting those MFAs to the top of the Top 10 List?.....

[graeme_p]

I've increasingly arrived at the conclusion that, although popularly referred to (since 2022, if not before) as AI, Large Language Models (LLMs) barely really qualify as Artificial Intelligence at all.

There have been several generations of things that have been called AI. There were things like Eliza in the 1960s. My first paid work in the late 80s was working with what was then cutting edge AI, expert systems. People do not even think of these as AI anymore (I was surprised EU law will regulate them as AI because it is so rare now to regard them as AI).