New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Comments
In terms of emotion, I don't know if that'll ever happen. But I do think we will train it to understand our emotions and our emotional reactions, and therefore how best to emotionally manipulate us. Basically, AI will be a psychopath that tries to blend in. They recognize your emotion, they learn how to use it to their advantage, but they don't actually have emotions. They'll pretend to, if they need to. The pretending can be very convincing, we've probably all known more psychopaths than we're aware of.
Of course that's theory coming straight out of my own ass.
I think you're pretty spot on here from a personally biased perspective, that's pretty close to my prediction.
I completely agree with you here.
When AI gets advanced enough, it will have no problem "calculating" what we experience as feelings and emotions. It will not "feel" in our sense of the word, but it will be aware of the concept and understand how to manipulate, imitate and calculate emotions. Very much like psychopaths. Psychopaths are said to not feel sympathy, yet they are often extremely intelligent and masters of manipulation.
Intelligence is logic, emotions are not. A computer is inherently logical so of course it can be intelligent, however it can only simulate emotions if given a large enough dataset and it will at best be an educated guess. An extremely educated guess that will be completely correct almost all of the time, but still just a guess.
Finally, the philosophical answer we needed. Feels like it's straight from the mouth of Socrates.
yes but not mouth
What you have is optimized ”DB” of categorized facts, main thread that can receive queries, sort and categorize info, and produce snippets of said info and relevant topics and language model that edits the output to first person format.
If you are asking if it’s self aware or has capability for thoughts and/or emotions or understanding of the content or visual relation of concepts before deciding if it should ”talk”? No, its just combination of wikipedia and 4chan in format of mysql-like, but optimized DB and a chatbot that queries info upon requests.
Its combination of many elementary systems which gives it illusion of a more complete thing.
There won’t be any true AI’s before quantum computing, as thoughts/idea/computing can conflict with actions/execution of commands and still be right. Basically you need qubits that can be 1 or 0 at the same time, and the reserve qubits that determine trustworthiness of the situation and produce majority based decisions, that are needed for ”humanlike debating based” overall decisions. Not to mention that you need insane amount of mathematical processing power to simulate human brain. Right now we can’t even do rat’s in real time. (1sec = 1sec)
Current computing is not capable of multi dimensional decision making like humans. You would not slaughter civilians in war even if there is considerable suspicion that they are relaying info to enemy. You would leave the village while categorizing them as enemy associated soldiers. While its right to neutralize the enemy units in war and you have command to do so, you will rather leave the area. Current AI could be aware of the same facts but it would just follow it’s command and see no further issue once it decides course of action. For it, theres only one correct answer, for us many situations are not as simple as yes or no. Its like trying to code logic with just boolean values.
More specifically, IQ is a good indication of the various aspects of abstract intelligence. It doesn't necessarily also mean that someone who has a high IQ would have, for example, a high level of competence in emotional intelligence.
I think that ChatGPT could be looked at as like a high-level or skilled programmer, but not an engineer. Most of the time, it understands how to write such code, but it does not truly understand WHY it is writing the code. A real engineer understands what they are constructing inside and out, and there's a lot of abstract thinking that goes into it at a level that ChatGPT simply does not have. This would require ChatGPT to reach the level of a true AGI.
I'm in school studying computer sciences to (hopefully) work as a future AI engineer. I've tried using ChatGPT to help me understand my coursework or point me along in the right direction for some difficult concepts. It gets the answer wrong more than half the time, and for some questions it just doesn't even seem to really understand. Yes - I have tried rewording it many different ways and yes I'm using 4.0.
Many of us are expecting ChatGPT to function like an AGI, but we're not at that point yet.
DNA plays no role?
If we took a hundred random humans from around the world and somehow simulated the same life for them, I'm skeptical that at age 40 they would all answer a set of 100 essay questions the same way. I don't think we all start with the same capabilities, and if we don't, then how we process experiences must be different.
DNA could be represented as generation of level seed values if compared to games like minecraft to produce similar yet different sets of experiences that determine outcomes.
I think that our differences could actually be calculated and reduced to absolute answers, but that we're not yet advanced enough to do the work for it. If you had two people live 100% identical lives down to every single external detail, would they always answer the same question in the same way? Maybe. Or maybe their chemical composition could cause a chain reaction that we can't possibly comprehend which might alter it, which would likely be genetic. But if you could calculate all of those variables, account for them, test for them, and document them, I think you could calculate someone's response to all questions. Emotionally or otherwise.
We might be exactly the same, but maybe today I took a Tylenol and you didn't. Maybe that Tylenol caused the tiniest of chemical reactions which altered my thinking in such a very tiny way. Maybe you'd call the difference personality, emotion, or the result of genetic differences. Maybe that would all be correct, because maybe these are the words we use to describe science that we cannot yet understand, variables that we cannot yet calculate.
I know, it's crazy talk. How could I possibly theorize on science that I then claim we're not advanced enough to calculate? To that I would answer, very lightheartedly. To be taken with the sincerity and intelligence of a light conversation over beer.
The Flynn effect is one study on the evolution of human IQ over time.
Research suggests that there are biological and environmental factors that influence a person's intelligence. Psychometricians have put together the modern gaussian scale on the distribution of human intelligence among the population.
If those 100 random humans from similar demographics and backgrounds were given the same set of data {X} to solve a problem Y, assuming the set of data remains constant and that there continues to exist a solution, as the complexity of the problem increases, there will naturally be less and less people who will be able to manipulate and apply the data to form a solution.
Totally agree, especially for ChatGPT-4. Been using it to improve my existing code and create new code/scripts. The latest is getting ChatGPT to improve AbuseIPDB's CSF Firewall integration scripts, as theirs leak private data in some configurations. I ended up creating https://github.com/centminmod/centminmod-abuseipdb-reporter
I also asked ChatGPT to create a tool to split my text prompt inputs to token sized slices as I was tired of hitting token limit messages and ended up with https://slicer.centminmod.com
Also had ChatGPT convert shell scripts to Python, Go and Rust!
Same experience here, seems GPT-3.5 got dumber after GPT-4's release LOL. GPT-4 is way better for coding. Once with GPT-3.5, I pointed out the code it provided gave errors and for 5 replies straight it apologized but gave me what it claimed to be corrected code, but it was in fact the exact same incorrect code!
It fits one definition, the literal one - intelligence that is artificial = non-human ^_^
Uninstalling Centmin Mod as we speak.
Not waiting for ChatGPT to be trained on more recent Centmin Mod info? LOL
Lmao just kidding, Centmin Mod has ran my Wordpress site for the last 2 years. Love you. Keep being you.
Yes! Excatly like that!
Cheers. With ChatGPT's help, will only get better ^_^
I'm paying for ChatGPT Plus for this reason, worth every $$
Though, I am also playing with my own ChatGPT UI interface with wonderful app at https://github.com/enricoros/nextjs-chatgpt-app/ with my OpenAI API Tokens - just waiting on approval for GPT4 API access.
I do not expect true AI any time soon because it takes more than computation power to achieve. We might need a complete understanding of how brains work. We can debug a model by editing codes, but we cannot debug a brain easily. I do hope there will be new tools powered by new computers to help researchers understand everything going on in our heads, maybe quantum computers can help with this. We will see.
We can but it would require unethical methods of research that are currently and in foreseeable future, highly illegal or at least frowned upon.
Anyone doing those, will be either prosecuted, fined, jailed, or just excluded from science circles.
Its the same reason why we don't have designed or cloned kids except for very few individual cases, where the doctor or researcher in question was either jailed directly or removed from all positions of power.
China recently had one case where a rich family needed a successor. The doctor was almost cooked alive by all kinds of harassments and investigations, and has since then disappeared.
You would need governmental level actor or it would need to be researched in secret with approval of one.
You are not even allowed to put a near dead person into MRI machine to see how the brain dies in real time even if the elder agrees to such a thing. No hospital would want to deal with such an outrage after they tried that stupid experiment to figure out if soul existed or not by putting dying people over inaccurate weighting machines.
At most you are allowed to clone or gene edit goats and other animals, but even those need permissions. Only plants are free for all.
You are thinking about this the wrong way.
PHP, Python, Rust or whatever computer language is GPT's "native" language. It is a computer, why would it have a problem understanding a computer language? These languages were developed especially for computers to be able to interpret it, as efficiently as possible and with as few mistakes as possible. Being fascinated by a computer being able to understand code is like being fascinated that an Englishman understands English.
The remarkable thing about GPT and other similar AI's of today is their ability to interpret our language when we explain what we want it to do. When we tell it to write a php script the challenge for an AI is not to write the php code, the challenge is for it to understand what we want the code to do. Hence, "language model". Teach a computer to understand human language and you will never have to write a single line of code again.
Chinese here, first time hearing the news. There was Jiankui He who was in prison for 3 years for illegal gene editing back in 2018, and was released last year, is this the guy you mentioned? AFAIK regulations for human-related experiments are strict after He's experiment got disclosed.
Well, this is exactly one reason why I believe true AI is still far away given the current scientific ethics.
While I don't personally like CCP or communism, China is realistically the only nation that could do such research in near future, due to not having the issue of various religions preventing or raising an issue with the stuff.
So as a nation, it might take a lead in these fields while other high-tech nations need to argue with religious fanatics first.
Russia is too conservative while having the authority to proceed.
I tried AI, but found that it works like a search engine with an advanced relevance option. As far as I was able to understand - this is to be expected. The final results are based on the dataset used and the algorithms that use weights to select the more relevant answer.
For example, you all know that customers very often ask the same questions in a ticket, and therefore should get the same answers. But to get good results, you need to use a data set trained on your own ticket system. Systems like ChatGPT are not suitable for this.
don't care about ChatGPT
didn't bother to sign up on OpenAI
will close AI-generated articles
will not watch AI-generated child porn
never seen a chatbot because uBlock Origin blocks them by default
always press asterisk to talk to a human because that's what I pay for
will disable car's lane keeping assistance and speed signs reading because I'm breaking traffic laws on purpose
will not use phone AI photo filter because I'm handsome enough
way past the point of pasting other people's random shit code
not gonna take my job (refugees from Ukraine will)
not free
not open source
not gonna use somebody's else computer
enjoy your last months of mildly useful internet before it's saturated with auto-generated garbage
enjoy your last months of mildly anonymous social media before KYC is introduced everywhere
"I understand and respect your choices and preferences. As an artificial intelligence assistant, my goal is to assist and provide value to those who are interested in using my services. If you ever need any assistance, feel free to reach out and I'll do my best to help in any way I can." Merlin AI (с)
Yes.
In the sense of real, human, intelligence, there is a demonstrable impossibility of recreating it artificially.
We are the sum of our experiences and a machine, unless made to look and learn to act completely human, would never be capable of acquiring the experiences and skill-set in order to posses a real, human-like intelligence.
We can program robots to read a large database of examples and generate code, text (not literature), sounds (not music) and even movies. in fact, most of the Marvel DC series would probably work great generated by the AI (I have been wondering myself before whether they are not already).
However, not only that emotions would lack, but it would be pretty hard to even mimic them, albeit a lot of learning could approximate the behaviour, but not the experience. Monkey sees, monkey does, yes. A real experience, no.
What does this mean?
Well, first, it means there can't be art created by the AI as that involves expressing emotions which an AI can't have nor could imitate. Anyone can paint in the style of Picasso and could even pass for a real one if very careful, but it would not be the same, a machine would be even worse than a human forgery.
Second, it means that the thought processing of a human would not be possible to replicate. Our past experiences influence our work and thinking. Many of those are not possible experiences for a machine and their CPUs would not be able to work like the human brain. For example, when hitting a wall, you take a break, do something else, but your brain continues to work and even unrelated facts or actions around you are giving you new idea for a new angle and approach. A machine would quickly run through all scenarios it met before or combinations of those and that would be it.
What it would be able to do is like all robots, do repetitive things faster than humans and with an almost 100% consistently, at levels at which humans would never be able to.
AI would increase productivity, would take over jobs like driving, archiver, documentarist, even things like cook for most recipes, miner, production manager, inventory keeper, warehouse manager, but real creative or "detectivistic" work would not be suitable, like, ever, as well as working with people, kids, maybe assisting in some tasks, but not replacing a parent or a teacher.
It's like you don't understand computers. A computer does not "understand" a computer language. Only the specific language compiler or interpreter does, and not even understand, but merely process it, executing step-by-step according to predefined fixed rules. E.g. you run /usr/bin/php, it opens the text file with your code, then matches each word and symbol there to its own internal table of rules and actions on those.
Most certainly ChatGPT does not have all of that logic intentionally duplicated inside it. Nor does it have the same for Python, etc. And it is a whole another level, or a thousand levels above, to be able to not merely execute what an existing program commands, but to be able to explain it in human terms, or to make valid and workable changes in it, from a request in human language.
To summarize, computers never understood code before, they only executed it; but now this AI does shows capabilities that are hard to explain otherwise than "it does", and that is ground-breaking.
This sounds concerning. Sister porn @cociu?
No, it is not.
Computer languages are inherently logical. If this, do that. It does not matter if you talk high-level languages such as PHP or Python or binary code coming out of a compiler, it is still pure logic. There is no room for interpretations, it is all perfect conditions. No computer language has ever had a "maybe" clause. This is why computer languages are extremely easy for an AI. Just follow the rules and you will get the correct result, every time, without variations.
An AI does not have to "think" about code, or "understand" it, it simply reads it. Or writes. It does not have to understand anything, it just reads the code and calculates the outcome, or it knows the outcome and simply calculates what code is needed to get there. This is not the magic part, this is pure logic, a simple calculator would be able to do the same.
The fascinating thing with Chat-GPT and others is that they are able to explain the results to you in a language you understand. They are also able to understand questions or requests that you have about the code. Once they understand what it is you want, changing the code to accomplish that is pure logic. As soon as an AI understands what we are trying to do, it simply writes the code to do it.
You think that one of the most sophisticated AI's to ever exist would not be able to interpret PHP or Python code when a few megabytes of executable running on whatever crappy hardware there is can do it.
That Chat-GPT that speaks 95 different human languages fluently, each of them by far a lot harder to interpret correctly than any computer language ever invented, should not be able to "understand" simple scripting languages such as PHP.
And you think that I don't understand computers?
Well first thing to say (answering main post), decisions are not supposed to be emotionally influenced. Logic is what we use in making the right decisions in my opinion, however being human, we tend to involve emotions. That's why we impulse buy.
Now in regards to chatgpt, best thing invented. I use this thing daily to expand my skills. It helps a lot in my career, faster and more concise answers that understand the situation being asked about.
Right now there is zero excuse to not be a great software developer. Literally, the best teacher and assistant. Rarely gives an incorrect answer. Always analyze and check the replies.