hi all. about 20 years ago, i remeber number of little home chatbots programed by various hobbyists. they were always primitive, rarely with some learning feature (based on dictionary expansion only) and usually just trying to be funny and make jokes.
there were quite short time fun with them before they turned out repetitive.

today, we experience blast of deep learning neural network applications. for example i like to play with from-words-to-picture generators like dall-e (tried 3 or 4) and i was really impressed by the case of how google fired an employee because he believed that their tested artificial intelligence had become conscious!

although I don't believe it really happened, since that, i desire to have such artificial "creature" installed on my home computer to chat with!
with program which is capable to persuade someone that it is "alive" it must be really fun!

so - question:

do you somebody know about downloadable chatbot like that (not intended for support or marketing but just to communicate with human and learn from it and from internet) ??? (maybe, say if it is for win or linux or both)

i think, some independent projects in beta would exist but my english is rather shitty to search google effectively. if they are not, i believe they will be soon and will make company to lonely ppl for example.

I'd say such a project is stupid and not to get it.

Why? Because a chat box using a neural network is just a way to brute force output by giving it hundreds of thousands or millions or billions of inputs and tests and randomizing parts until it gets the desired result.

However the desired result from a chat is... well, silly and has no strict output. So you'll get semi-intelligible strings with no real knowledge what came before; Or it's far more likely it will either have a database of strings and options to choose from in a tree and throwing it a wrench of an input will make it give odd outputs. It could also be based on likelyhood of the next word following after the last 3 words, and choosing that as the output scanned from most internet sites stories and whatnot which may make it seem very realistic but has no 'intelligence' of it's own.

Though if you enjoy the fantasy, you might like Tron. Go enjoy some old movies :)

Go out, try speaking to people.

Expand your own neural network.

blah blah. you really don't have to react by hook or crook having just shit to say. sit back and relax. thanks for opinion, but i search for the best possible home computer chatbot, ideally with a neural network touch (if possible with home pc performance).

im not interested in tron or other movies

I don't think this is a project worth perusing.
It's a seriously exhausted field, having been used for entertainment purposes as far back as <checks> 1966.

If you're investigating chatbots via neural networks as a historical curio, knock yourself out.

But Tron is about computer programs that are sentient and smarter than any chat box :P

People take this way way too seriously. The Loebner Prize was fun, and so are chatbots (in fact, a genuinely efficient AI conversationnist -soulless but indistinguishable from humans- would be boring). The giggles come from the quirks, tricks, limitations and surreal/random results.

For what it's worth, Mitsuku (kuki ai) has been the regular winner of these last years' competitions. I'd enjoy toying with it a bit, but (in addition to being presented as some blade-runner-esque young woman, which is already embarrassingly dumb) the online version requires us to make an account and log in, and nope. Still, if you're not bothered by that sort of thing, you can have a go : https://www.kuki.ai

so what.. still, it's not an installable program anyway. that means it's not related to the subject of the question

i dont understand how do you mean that "if you're investigating chatbots via neural networks as a historical curio"

not "via" but "with" or "using".. and not that "historic curio" 20 years old chatbots for fun, but nowadays ones using neural network.. something like google's LaMDA..

i dont know if you noticed the news record.. so copy paste:

When Google engineer Blake Lemoine claimed an AI chat system that the company's been developing was sentient back in June, he knew he might lose his job. On July 22, after placing him on paid leave, the tech giant fired Lemoine for violating employment and data security policies.

i would love to have and test something like that.. thats all :-)

I find it odd he was fired just for making a statement. There's more below the surface to this story. Maybe something big tech DO NOT want us plebes to have as common knowledge. The whole field of artifical intelligence/life/sentience/consciousness, neuro-this and neuro-that and all "we're nothing but biochemical automatonry" philosophy in general.

Well...... anyone else find it. Sinister?

BTW there is an official label such advocates of the above use for "not brights" like me. "Neuromaniac". Honestly, google "neuromania vs neurophilosophy" see for yerselves.

If you want a laugh, try "neuromarketing" as well! ;)

PS/ Didn't Richard Dawkins (in his rabid obsession to find such an unholy grail to put any "delusions" of spirituality to the sword once and for all) put up a monetary prize for anyone that created an AI that fooled Dawkins into believing it was real human? Quite a substantial prize as well if I remember right. Nor do I know if it was ever claimed.

No one should be fired for such a statement. They went too far with that.

It's a matter of non-disclosure agreements, that's stuff people get fired about. He didn't express an opinion, he leaked stuff he contractually wasn't supposed to.

Without asking a question questions myself or doing tests of gibberish or other things, it's hard to say if it's sentient or not.

Some chat bots go out and are given minimal information, and pick up how to reply by copy-pasting replies they get from real users, making them seem a bit lifelike. Until you start asking the same thing over and over again. One way to identify a bot is to ask them their age over and over, you'll start getting ages at first normal 18 21, 30, etc, then you'll start getting 0 a million, a billion, and other nonsense answers.

As my reply earlier and from Tim Pool, likely it is using the most likely word following another word or set of words (not too unlike some compression algorithms)

Makes me wonder how they would act if you started intentionally misspelling words and see how it acted.

without having possibility to chat with that LaMDA we cant say anything. but the guy started making shit like hiring lawyer to defedn the chatbot from turning off.

but sense of my question is, if somebody knows about downloadable open project of similar chatty AI.. :)

Have to see how things progress. But for now i'll assume it wasn't sentient.



Not sure on the chatty part, but i remember a demo one when i was a kid from a BBS or a CD. Was okay overall...

Then there's one in 'life and death 2' or something, though it seemed to look for key parts in your message and act more like a shrink of 'how did that make you feel? Uhuh, go on... Tell me more...' etc, very vague and useless replies.