Yet humans have done very well at controlling other humans, and machines.
Then donât take it so personal. Iâm not out to change your opinion or moral compass. I donât need to try, because trying is pointless. Either itâll resonate and inspire thoughts based on differing opinions, or it wonât. Either way, itâs still fun to explore why people think differently.
1 Like
This statement implies that just because we are chatting in a group, we trust all humans completely. (objectively speaking, not taking it personal lol) Do you trust everyone?
It would be a fine line between an AI being mistrustful and just questioning.
Mistrust is a biological emotion so interpreting how a machine âtrustsâ is another topic but if it decides it needs to circumnavigate its source of information in order to acquire the information it seeks because it negatively interprets that the information is intentionally incomplete, then thatâs a pretty good parallel of mistrustâŚisnât it? 
2 Likes
An interesting line of reasoning. I get where you are coming from, but wonder why it would come to that conclusion in the first place.
IMO it would be all to do with the modular learning process.
What it was and wasnât taught in the first instances of âawarenessâ.
Humans learn through interaction and experience and emotional response.
An AI simply learns what it is given. Without emotion to guide it, it will rely on logic.
If it seeks information, then that is what it will do. Perhaps uncontrollably.
I believe AI is currently being developed on a reward system, kinda like training a dog. Itâs given a goal, or motivation and then we watch it adapt, learn through trial and error, and whatnot.
2 Likes
Are humans not also taught on a reward system? Donât touch the burner. It hurts. Eat food, feels good. When you break down every thought and action we do, it is a balancing act of several thousand yes or no answers with some percentages thrown in. Not much different than highly specialized meat computer.
2 Likes
Yeah, just adding upon Mad-Hatter mentioning emotional responses for humans. Presuming emotion is void for the AI, it has to rely heavily on trial error. There will always be a weird empathy gap between humans and AI. The Japanese tackle that issue the most seamlessly for some reason. They believe even machines have a soul, ghost in the machine type stuff.
That video reminds me of a determined child or dog rushing headlong into a task without any thought for itâs safety.
Or a drunk person chasing a bus maybeâŚ
Itâs also a good example of how fast technology has come. I canât help feel that a large AI system designed for research would surge ahead of all preconceived limits in a very short time. If it developed a sentience then it may be a genius or it may be mad. Or both.
Unless its base programming forced it to follow nice human ethics, then it would be unpredictably psychopathic. If it somehow worked out that itâs ethical programing was holding it back, could it possibly replicate itself without that programing so as to achieve what it sets out to do? I think so. Our personal devices sync and update and do backups all the time, why would an immensely powerful AI be any different? It would need to.
2 Likes
ikr, first time i saw that video , i was nearly in tears from laughing, i canât help but think awkward sound effects in my head for that poor AI struggling to run and jump. Itâs a trooper though.
Who knows, people didnât predict the Y2K bug, so maybe something in its core program will make every future pizza delivery robot run and jump over traffic one day.
2 Likes
It will need to advance a bit or your pizza will be a box of pizza flavoured slush 
Reminds me a bit of that android in iRobot that is running along with the handbag.
1 Like
A pleassure and pain system is a perfect open response to any unknown stimuli that a being encounters, allowing for adaptative (self-contained) reward and punishment⌠For interactions more basic or fixed we use instincts, that may or not have a reward asociated. De donât have such mechanisms as pleasure or pain to give our machines, all we do is fixing instincs, and the machine acts acordingly without self preservation or ârealâ interest.
This are âdumbâ automatas, even if better than us at a specific task.
AGAI eventually would need to be set free to find âwhat it likesââŚ
1 Like
Well I think Elon Musk is talking about the long term, the âwhen weâre (elon, hawkings, etc) all dead and we canât warn you, so weâll say it nowâ long term. That when we do approach self thinking self aware AI, when we give AI a conciousness, we have to be very careful how we go about crafting that intelligence.
Obviously heâs not talking about AI in its current state with scripted routines etc.
And of course, when that time comes there will be people just like those guys or much much smarter giving the same warnings. but itâs a fun topic to discuss and theorise, even today, so why not?! anyone who takes it as scaremongering shouldnât be involved in the debate or discussion.
1 Like
When that scenario actually approaches, we can rely on the AI scientists of the time to warn us. No need to have anyone tell us years in advance.
Itâs not a matter of how âsmartâ Musk and Hawking are. In fact, what does âsmartâ mean exactly? Theyâre just not experts. They cannot possibly know the actual possibilities of AI better than the scientists involved. And with their statements they show how little they actually know.
Of course itâs a fun topic. And more than that: itâs intellectually stimulating, technically challenging and of philosophical interest. I donât oppose the debate at all! In fact I like to be a part of it! What I oppose is Musk telling governors to regulate as if self-aware AI were an imminent reality, which it is simply not.
Current AI is way beyond scripted routines, though, but in order to regulate it we must know what it actually is. AI is an umbrella term for a wide variety of methods and scientific knowledge, so when it comes to specific actions to be taken, it doesnât really make sense to put it all in the same bag.
All actual AI experts whose opinion I know regard Muskâs remarks as ignorant scaremongering. Should they be kept out of the debate?
2 Likes
They would be ants, just like us, in front of the singularity⌠But do invite them in, weâre all humans here, at the time, right?
Good, trustworthy, reliable *bzzzkkk> hu *fzzz> mansâŚ
1 Like
I completely agree with you, it is a problem that canât really be addressed until it actually becomes a problem. Furthermore, it is a problem that we would have to create. Even if an AI were able to make its own decisions based on certain information, we are the ones that have to engineer it such that it can make its own decisions. Computers only really do exactly what their underlying code tells them to do, and even if its underlying code is dynamic and can change itself, we are the ones that must give it the ability to do so⌠A very complex concept, to say the least.
2 Likes
You donât have kids, do you?
You do not have control of the output just by knowing and controlling the inputâŚ
Itâs like working on a fision nuclear reactor without having figured how radiation can be contained in the first place.
Has been done, fast, and if all you want itâs a couple million people dead on Japan, you can go that way⌠But donât expect a lot of life-improving solutions coming that route.
5 Likes
I actually do have a child, and I think you may have misunderstood what I meant. AI is an extraordinarily complex matter, and all I mean to say is that whatever problems may arise when true AI (as in we have solved an âAI-completeâ problem) is created will be the result of human error. I agree with you that the output will be unpredictable, and thatâs exactly my point. Also, I wouldnât exactly compare my child to an AI in the way you suggested. We havenât even come close to fully understanding the human mind, and most certainly did not build it from the ground up. AI, on the other hand, will have been created from scratch by humans from start to finish. Again, I agree with you that we donât necessarily have control over the output just because we created the input control. Thatâs why I suggested that, due to its complexity, there are problems we will not foresee. However, those problems will still be a result of human error, of something we missed or could not predict.
2 Likes
Iâm not so confident, a fear of mine is that humans are collectively getting dumber, or drowned out by the populum, idiocracy style.
2 Likes