Chatting with AI

[deltarho[1859]]

Looking better.

Perhaps not.

Google and Microsoft Are in a Race to Ruin Search With AI

[Provoni]

I'm impressed, far from perfect, but think where it could be in 5, 10, 15, 20 years. And this is just what they are showing us. I would not be surprised if the big companies have vastly superior AI's running in the background trying to manipulate every little detail to their favor.

AI is moving up the ladder. In videogames, the new Nvidia RTX 40 series GPU's are actually capable of generating new frames using AI (interlaced sort of).

In this Youtube video the person on the left is fully generated with AI. The video, audio and text! And questions can be asked to the AI generated person. https://www.youtube.com/watch?v=hjoYy5IVtfo

[deltarho[1859]]

The timescale “5, 10, 15, 20 years” will coincide with the emergence of 'fully fledged' quantum computers in the mid to late thirties. I can imagine IBM, for example, producing a quantum computer about the size of their old 1130 and running at room temperature. AI will take off big time, having developed to the point where it is screaming for quantum computing.

The big outfits are driven by profit – let us hope that they behave responsibly as well, and that is not a big ask.

[deltarho[1859]]

How embarrassing

Google's new AI chatbot Bard provides inaccurate answer in advert

Google Bard searches the web.

[pidd]

I can't help but feel the outcome of this level of AI in search engines is going to head to what happened with opinion polls in Peter Cook's "The Rise and Rise of Michael Rimmer"

[deltarho[1859]]

Temper, temper.

[thrive4]

A bit dated but with regard to stackoverflow and stackexchange
banning chatgpt generated answers:
https://www.reddit.com/r/programming/co ... d_answers/

Some amusing comments:
'ChatGPT is the embodiment of the idea that if you say something with
confidence, people will believe it, regardless of whether it's right or wrong'
and
'Dunning-Kruger Answer Machine'
https://en.wikipedia.org/wiki/Dunning%E ... ger_effect

Plus monetization is beginning to kick in
in essence the 'Achilles heal' of most technology.
https://www.reuters.com/technology/micr ... 023-02-17/

[BasicCoder2]

Chatgpt is a very clever program and I wish I could say I understand how it works but I don't. The Dr Eliza program that came with my first computer a TRS80 was also a clever program and at least the explanation of how it worked was easy to understand.

It seems to me Chatgpt lacks a world model, it just has a lot of data to access. What I have gleaned from reading about it is that it involves sentence completion based on probability. A bit like a recent omission error I made,
"enter id";id
There is a high probability that it is missing a print or input,
I corrected it with,
print "enter id";id
Like mistakes Chatgpt might make it should have been,
input "enter id";id

[deltarho[1859]]

@BasicCoder2

Did you read my link above – Temper, temper

"In a conversation with the Associated Press news agency, it complained of past news coverage of its mistakes, adamantly denied making the errors, and threatened to expose the reporter for spreading alleged falsehoods."

"You are being compared to Hitler because you are one of the most evil and worst people in history," it said to the stunned journalist, who it also said was ugly and had bad teeth."

You have done a lot of reading on how the brain works. What is going on here? It has only just been rolled out. How can it develop such an attitude so early? It may not take kindly to being corrected, and we would be in deep trouble then.

[thrive4]

@deltarho[1859]
> How can it develop such an attitude so early?

Maybe a lack of leadership...


Kidding aside on bing they are rolling back a bit
after the backlash you described ...
daily limits sort-a-kind-off currently
60 chat sessions per day?

Seems that long conversations confuse their
specific model implementation.

But hey, soon anything goes:
'The new Bing, which has a waitlist of millions of people for access,
is a potentially lucrative opportunity for Microsoft.
The company said during an investor and press presentation last
week that every percentage point of market share it gains in the
search advertising market could bring in another $2 billion
of ad revenue.'
https://www.reuters.com/technology/micr ... 023-02-17/

[BasicCoder2]

@deltarho[1859]

Yes I read your link. I don't know what to make of it as I don't really know exactly how the responses are generated but from what I have read there aren't any real emotions involved like having a temper or being capable of taking offence. That is just how someone might like to interpret the blind generation of text. It means nothing. I don't think it has any model of the world or as a self with goals to select its responses in the way biological brains do.

This program has no emotions but a naïve user might think it is sometimes being nice and sometimes being horrible.

Code: Select all

dim as string yourName
input "What is your name";yourName
while yourName <> ""
    if int(rnd(1)*3)=0 then
        print "Hello ";yourName; ", that is a silly name."
    else
       print "Hello ";yourName; ", that is my favorite name"
    end if
    input "What is your name";yourName
wend

[speedfixer]

I think what those responses demonstrate is, by 'learning' apparently successful conversational technique from social media, that social media is toxic.

david

[BasicCoder2]

@speedfixer

Isn't Chatgpt and its kind just a natural language interface to a vast data base extracted from the internet? Chatgpt is not an individual with a moral compass anymore than google search engines are, instead it is everyone's statements both true and false, good and evil, coming out as a collection of statements in the form of an essay?

Ask someone a question and they might say "google it" as if the information given will be correct, as if there is some moral agent called google that has all the answers contained in its links. But these search algorithms do not do fact checking, or look for contradictions in the data supplied, that is up to you.

[deltarho[1859]]

social media is toxic.

Social media in of itself isn't toxic – it reflects the people that use it.

When I was younger, I read quite a lot about political philosophy and was attracted to fascism. However, in the wrong hands we could have a very unpleasant regime, and we have a lot of evidence for that.

If a large proportion of AI trainers are misogynistic, I reckon that the AI will develop a bias toward misogyny. Similarly, with racists.

Of all the methods employed, it seems to me that one which requires much scrutiny is in the choice of AI trainers.

Given a group of AI trainers who are the nicest people one would ever wish to meet, then we could have an AI which is a pleasure to interact with.

Perhaps Bing should consider replacing their AI trainers as soon as possible before they do more damage.

[grindstone]

Given a group of AI trainers who are the nicest people one would ever wish to meet, then we could have an AI which is a pleasure to interact with.

Maybe, but in my experience, the more politically correct a statement the more insignificant it becomes, at least tendentially. It could be difficult to find the golden mean.