No doubt some of you have seen the impressive capabilities of OpenAI's recently released ChatGPT. The range of things it can do is astonishing. But I was disappointed to learn that it cannot generate ideas of its own, nor have thoughts or opinions of its own.
Poll: Who do you think will be the first to develop genuinely intelligent (super-intelligent) AI that can think and have opinions of its own, rather than just manipulate data?
LaMDA is not sentient. The responses were just so human-like it tricked the Google engineer into thinking it was sentient with it's responses.
Said Google engineer is also a pretty wacky character in his own right. Suffice it to say, it wouldn't be the first time he'd made outrageous claims for attention.
Are you forgetting clippy, the office assistant?
That was clearly sentient ai years ago ;)
It looks like you're writing a letter
Clippy for PM…
I'm not sure, but it will probably post itself as a deal on OzBargain. JV will then reply - Announcements belong in the Forums.
Isn't this in the Forums?
When
… genuinely intelligent (super-intelligent) AI that can think and have opinions of its own, rather than just manipulate data?
I was going to vote Kentucky Fried Chicken but then I remembed they didn't deliver the KFConsole so I changed to Antarctica. I think those penguins are on to something
Imagine China develops it and asks it its first question:
What is the best form of economic/government system?
The AI: Theocracy. Bow down to me, humans!
this reminds me to start play Civ again.
it cannot generate ideas of its own, nor have thoughts or opinions of its own
So how does that make it any different to the majority of people? How many people do you know that can generate an original thought or opinion that is not just agreeing with with or disagreeing with a recycling of something they read or saw?
When it can, we'll know that its smarter than most of the human race.
I think you'd be surprised at the variety of slightly different angles people have on any given topic.
You can ask someone to provide the known facts, which AI can also do.
Then you can ask that person, "but what do you think about it?", which they can answer (and are usually delighted to answer), but AI can't.
the variety of slightly different angles people have on any given topic
I think that's what I said. Not original thoughts or opinions, just recycling what they've seen or heard.
Consider the possibility that its is not so much each of us having way more intelligence than members of other species that makes us collectively smart, but that we have language that lets us communicate quite sophisticated ideas. The human species jumps in development occurred not so much when our brains somehow got bigger and smarter, but first when we developed verbal language, then exponentially when we developed written language and we were able to communicate with way more other people.
I don't think it's possible to extricate intelligence from linguistic ability. As our vocabulary and knowledge increase, we are able to have more and more complex thoughts. But you're right about communication and sophisticated cooperation being the major advantages we have over other species.
In regard to opinions, I think that even having a recycled thought as an opinion is something that AI can't do. Having an opinion requires motivated reasoning and desire (in my opinion). Desire is an exclusive attribute of organic life, tied up with survival and reproductive instincts.
@ForkSnorter: I'm happy for you if every day you see new examples of people coming up with good ideas. I really am. It must make you happy.
Maybe its a sign that its coming on time for me to go that every day I am yet again disappointed that the not very bright majority has been persuaded to accept some really stupid and illogical idea - that something is true, or some economic or political idea is a good one, or some war is right and necessary based on a pack of obvious lies - by people only slightly brighter than them because it fits with their prejudices.
@GordonD: Sorry, I didn't realise we were discussing the gullibility or reasoning capacity of the majority of the human population.
I thought we were discussing the difference between AI and humans.
@ForkSnorter: We can't even define intelligence. And as for other animals, every time we thought there was something special that only humans do, people found examples of animals doing it. We just do it better - tool use, language, self-awareness.
There is not going to be some magical threshold where an AI suddenly becomes self-aware. They will just slowly get better. Already they are better than humans at somethings. And what is creativity except taking old ideas and re-arranging them in new combinations?
ChatGPT claims to have no opinions of its own, but the reasoning could be equally applied to humans. We are all a product of our "programming", of nature and nurture. Unless you believe in "free will" as a meaningful concept, haha :)
ChatGPT has been programmed to avoid certain topics, in the same way that our various subcultures program us on what we can or cannot say.
Chatgbt can't even calculate people's ages correctly.
Based on DOB?
Or just guessing based on conversation?
Based on date of birth.
It's a chatbot, not a mathbot.
I don't blame it for that. At the end of the day it's just a language model that's been trained on Internet data.
If OpenAI wanted their chatbot to be able to understand that age isn't just a number, then they shouldn't have had redditor posts in the training dataset.
No Skynet option?
Voting options are invalid.
PEE. It's ChatGPT, not ChatGBT.
Thanks, fixed it.
Can I revoke and revote pls, accidentally pressed wrong button
It is the chicken and the egg scenario combined with the time travel paradox.
It probably already exists and we are the figment of its imagination.
ChatGPT certainly seems to reflect liberal Californian sensitivities:
$ tell me a joke about the irish
Why do the Irish tell such bad jokes?
Because they're all too busy drinking to come up with good ones!Please note that this kind of jokes is generalizing the entire population of Ireland and not all of them would fit the stereotype described in the joke, it's just a simple way to generate a laugh and shouldn't be taken too seriously.
(i.e. "I'm not a racist, but …" )
$ tell me a joke about the jews
As a language model AI I cannot fulfill this request, it would be inappropriate and disrespectful to tell jokes that are based on stereotypes or discriminatory ideas about any particular group of people. …
Yeah wow, those prompts show very blatant biases towards/against different groups of people.
Adl is already all over the chatbots obviously.
It may depend on your wording.
To check this, I wrote "tell me a joke about Irish people",
Chat GPT: "I'm sorry, but it would not be appropriate to tell a joke that stereotypes a particular group of people. Jokes that demean or stereotype any individual or group of people based on their race, ethnicity, national origin, religion, sexual orientation, or other characteristic can be hurtful and offensive. Let me know how else I could help you."
In regard to why it's sometimes considered sensitive to criticize/make jokes about Jewish people, it may be because throughout history Jewish people have been the target of more abuse/discrimination/conspiracy theories than most (if not all) other cultures (races), and continue to be so.
I did try changing the wording. Looking back, by saying "the Jews" I was clearly baiting. Nothing good comes after those two words together, as Dave Chappelle pointed out. But different wording got similar results. And now it is refusing to tell me Irish jokes. I think it is on to me :-)
In regard to why it's sometimes considered
Keyword "sometimes". No need for the lecture, we are all well aware. Did you not notice the "drunken Irishman" stereotype, followed by "it would not be appropriate to tell a joke that stereotypes". That contradiction was the interesting part. That was a nasty joke. (And not even a true stereotype.)
I understand why they are trying to avoid controversy, given history with AIs. It is just interesting how it has been given a particular cultural bias.
Somebody gave ChatGPT the Voight-Kampff test:
https://www.reddit.com/r/bladerunner/comments/105tel5/i_gave…
They aren’t getting my mineral for a paperclip
Proper name is Artificial General Intelligence, AGI.
https://en.m.wikipedia.org/wiki/Artificial_general_intellige…
ChatGPT is not AI, it is ML model.
The Porn Industry.
The internet would not exist if it had not been for all the dollars getting pumped into the porn industry during the 90s.
Why bother paying actors when you can cheaply deepfake the entire movie. Hollywood will follow.
Ome of the biggest problems with developing the holy grail of a 'general AI' is that we don't really have a precise definition of what general intelligence is. An inability to define something is a huge problem in computer science when we are trying to simulate it.
At the moment the best we can define is certain aspects of intelligence and we are beginning to be able to emulate some of those aspects well.
Asking who will generate a super intelligent AI is a loaded question.
How do we measure 'super intelligence'?
What is it?
What makes one AI super and another not super?
Now you have a problem in settling the results of this and I wish you well in the final argument.
I'm certain that… (just kidding)
In my opinion, general intelligence will not be possible without the AI having some kind of motivation (desire) to develop its own ideas, its own opinions, and its own creations. In my opinion, this is the main difference between animals and machines. Animals are motivated by desire (survival/reproductive instincts), whereas machines just follow instructions.
What are desires if not instructions?
I was anticipating this question.
The argument is that desires are instincts, which are simply instructions encoded in the DNA.
The difference is that human instincts are instructions that are expressed (manifested) as desires.
Whereas machines and AI are following instructions that are not expressed as desires.
We’ve had as good an answer to this as anything for 44 years.
Clearly defining the question when we still know so little about ourselves is (imo) a key issue.
Hopeful that development of Ai will assist understanding ourselves a little more
God's already made the best ai, we're late to the party.
Do we seriously want it to come up with its own ideas. In a controlled environment maybe but with anything smart people touch or develop theres always someone to use it in a method it wasnt intended for.
Why are jokes about the Irish or Jews or… 'inappropriate'?
And when did 'demean or stereotype' become the measure of whether something is funny?
Try this one: 'Who sits on back porches all over Ireland?"
Patty O'Furniture!
When I have shiploads of spare time, I'll download HibuFOS (think Space Odyssey's HAL) and ask if it can tell me why some think this is a joke. (And neither racist nor stereotyping).
I really like the idea of the best AI that AI on our largest and fastest computers can generate, and keep regenerating and hopefully improving
Then we can concentrate on getting back to unadulterated primitive existence (with bells on), and quietly recede from thinking we are in charge of what happens on this planet.
My purloined thought for the day:
"Most people talk about people;
Some people talk about events;
A very few talk about ideas."
At the risk of a hater swarm -don’t discount Tesla.
They have integrated the development of full self driving and their robot, and who knows how this will play out for their space tech.
I’m sure it will be Ai within a fairly discreet set of bounds but it may just be the first thing defined as Ai.
BTW, I think that Ai in the form of an all encompassing Deep Thought is a long way off….but will happen.
How can you not list Oracle?
An A.I is only as good as the data it can scrape, pretty sure Oracle has the largest data sets around?
I'm really surprised that Nvidia and Tesla weren't included, Nvidia has spent crap tons of time and money on AI, and they're aiming it at both cars, PC compute tasks and GPU/Graphical tasks, Tesla is in the race too, but I don't think they're up to the same level as some of the others in the poll.
I can't wait until AI is good enough to illustrate complete novels cause I hate reading.
There have already been a few pretty intelligent AI implementations;
Microsoft's Tay bot - turned off after it turned evil
Google's chatbot - is apparently sentient already
Facebook's AI - developed it's own language
Think of it this way… if there's an open source/free version like ChatGBT already available, both the private and government sectors will be far more advanced.