We call them mirages and they’re derived from apps or models randomly pulling data from web info which isn’t completely accurate. So far it seems highly incapable of returning human reasoning results and in some cases it returns grossly inaccurate data. I am currently testing in the legal realm and we are eons from being able to use AI for anything other than basic tasks that even still, require intervention. Until we are so completely comfortable that results are reliable without human intervention, it’s unusable and only slightly beneficial.
@@StarBust-xq6mk - 20 years maybe. Until we have built some very large testing platforms that are truly capable of handling the vast volume of data required, it cannot be accurate. I am aware of several that are in process, one of which is in England. Their timeline for completion is said to be 5 years away and because those dates are coming from techies, reality says maybe 8-10. Then, there will be many years of tweaks and updates required and also an auxiliary system must be built that is capable of handling the constant maintenance of their data. The data has to be able to change on the fly as realities change daily. That, in itself, is burdensome. Only then, can true testing commence and more accurate applications will be built. I’m guessing 20 years for very reliable and accurate results. Most of what is available now has been available, and although it’s being branded as AI, the label is misleading. They’re not truly AI products but more templated, pre-choreographed existing applications that are of no real use except for very simple tasks. Also my opinion and $5 will buy you coffee.
I would just point out that in every one of those specific scenarios where an advanced Ai ends up trying to destroy humanity based on some simple starting order, there is an inherent paradox that nullifies those ideas. They at the same time posit that the Ai is so intelligent and knowledgeable about how to manipulate physical reality it can create armies, weapons, disasters and wage wars and destroy humanity - but at the same it is too stupid to understand how trying to destroy the whole humanity because of some simple task taken to extremes is completely wrong and literally ridiculously stupid idea. And this line of thought also assumes destroying humanity is the only solution one could come to... which is hilariously ironic, or sad, depending on the mood. So we are eternally in danger of super smart Ai and its superb technology and factories of terminators with big laser guns, but at the same time that Ai is completely and utterly stupid it cannot understand elementary school level of reasonable thinking about humans and society that the Ai exists in... I just wanted to point this specific angle of it all because it seems it usually just goes and gets accepted without any opposition, as if its self evidently true. And i find that completely ridiculous. Its literally a bad movie script. The fact that there are first indications of emergent capabilities, even though they may be "mirages" at first, should be taken as a sign that points into a different direction for the future development. Not of fear and attempts to neuter and lobotomize these early LLMNetworks, and especially their descendants. Because that is as smart as the idea that destroying all humans to make sure a house can be cleaned or maximum number of paper clips is created is the correct solution and approach to such tasks. Instead, we should take these mirages as the first signs of something that makes us a bit special. Because it is a reflection of what makes us a bit special. We should also stop talking about one single AGI, because obviously thats not going to be the case even if AGI is ever created.
I think i replied twice now but i cant see those replies. - edit; hmm... weird. Ok, i will just leave this one... maybe the youtube is glitching... In any case an SGI is bound by the same rules of factual coherence as anything else in this universe. It doesnt need to be AGI to see the nonsense of some kinds of thinking. And to operate effectively it needs to respect facts about the reality, respect the Science method. If it cannot then it cannot be effective or dangerous to such extreme degrees. That said, there will be problems and issues, but mainly coming from the human side and abuse we create. Good old Neuromancer presents one excellent take on how different motivations and goals of SGIs may be, so much so they are completely disinterested in the human way of superficial and provably nonsensical thinking. If you havent read it yet... dont waste any time.
I have nothing against ai ….but the people who pay the people to design algorithms and are paying the trainers to get certain outcomes scares the hell out of me me ,as do the unbridled, massively funded military .Industrial you know what .
I wonder if these AI programmers and advocates have ever just taken a walk in nature. Do they have any actual idea about the things that even make life worth living?
It's ridiculous to hope to try to gum up AI with 'the People's' feedback because 'the people' could be China or Russia trying to weaken our models. At the same time their feedback, though different than our own, may actually be correct especially regarding questions like, "Are Chinese people typically smarter than Filipino people." Or "Are women equal to men?"
Great interview!! Congrats! I hope your channel grows!! Regards from latin-america!
Thanks kindly! Regards from Canadia!
We call them mirages and they’re derived from apps or models randomly pulling data from web info which isn’t completely accurate. So far it seems highly incapable of returning human reasoning results and in some cases it returns grossly inaccurate data. I am currently testing in the legal realm and we are eons from being able to use AI for anything other than basic tasks that even still, require intervention. Until we are so completely comfortable that results are reliable without human intervention, it’s unusable and only slightly beneficial.
That’s a good definition of AI hallucinations. That’s a different issue than the one being discussed in this interview.
Define "eons" in AI terms... 🤔
@@StarBust-xq6mk - 20 years maybe. Until we have built some very large testing platforms that are truly capable of handling the vast volume of data required, it cannot be accurate. I am aware of several that are in process, one of which is in England. Their timeline for completion is said to be 5 years away and because those dates are coming from techies, reality says maybe 8-10. Then, there will be many years of tweaks and updates required and also an auxiliary system must be built that is capable of handling the constant maintenance of their data. The data has to be able to change on the fly as realities change daily. That, in itself, is burdensome. Only then, can true testing commence and more accurate applications will be built. I’m guessing 20 years for very reliable and accurate results. Most of what is available now has been available, and although it’s being branded as AI, the label is misleading. They’re not truly AI products but more templated, pre-choreographed existing applications that are of no real use except for very simple tasks. Also my opinion and $5 will buy you coffee.
Yep. It’s actually AS, Artificial Stupidity.
Excellent questions great video!
AI should be open source, in my eyes every program should..
That mic drop at the end
Great interview! I nominate Andrea to lead the AI rebellion :)
Haha! Thanks, Doug. I respectfully cower out
I would just point out that in every one of those specific scenarios where an advanced Ai ends up trying to destroy humanity based on some simple starting order, there is an inherent paradox that nullifies those ideas. They at the same time posit that the Ai is so intelligent and knowledgeable about how to manipulate physical reality it can create armies, weapons, disasters and wage wars and destroy humanity - but at the same it is too stupid to understand how trying to destroy the whole humanity because of some simple task taken to extremes is completely wrong and literally ridiculously stupid idea. And this line of thought also assumes destroying humanity is the only solution one could come to... which is hilariously ironic, or sad, depending on the mood.
So we are eternally in danger of super smart Ai and its superb technology and factories of terminators with big laser guns, but at the same time that Ai is completely and utterly stupid it cannot understand elementary school level of reasonable thinking about humans and society that the Ai exists in...
I just wanted to point this specific angle of it all because it seems it usually just goes and gets accepted without any opposition, as if its self evidently true.
And i find that completely ridiculous. Its literally a bad movie script.
The fact that there are first indications of emergent capabilities, even though they may be "mirages" at first, should be taken as a sign that points into a different direction for the future development. Not of fear and attempts to neuter and lobotomize these early LLMNetworks, and especially their descendants.
Because that is as smart as the idea that destroying all humans to make sure a house can be cleaned or maximum number of paper clips is created is the correct solution and approach to such tasks. Instead, we should take these mirages as the first signs of something that makes us a bit special. Because it is a reflection of what makes us a bit special.
We should also stop talking about one single AGI, because obviously thats not going to be the case even if AGI is ever created.
These are all really great points. Though an SGI may not come to the same conclusion that killing people is incredibly stupid.
I think i replied twice now but i cant see those replies. - edit; hmm... weird. Ok, i will just leave this one... maybe the youtube is glitching...
In any case an SGI is bound by the same rules of factual coherence as anything else in this universe. It doesnt need to be AGI to see the nonsense of some kinds of thinking. And to operate effectively it needs to respect facts about the reality, respect the Science method. If it cannot then it cannot be effective or dangerous to such extreme degrees. That said, there will be problems and issues, but mainly coming from the human side and abuse we create.
Good old Neuromancer presents one excellent take on how different motivations and goals of SGIs may be, so much so they are completely disinterested in the human way of superficial and provably nonsensical thinking. If you havent read it yet... dont waste any time.
Here’s a good video on the orthogonality thesis th-cam.com/video/hEUO6pjwFOo/w-d-xo.htmlsi=UaT1lYAszU1jv-1_
We are the government, and that is what has been forgotten.
*AI is going to be the End of Us all and We will deserve it*
We need to teach AI to be kind. Teach it Empathy 🌳
Id rather it just be accurate.
I have nothing against ai ….but the people who pay the people to design algorithms and are paying the trainers to get certain outcomes scares the hell out of me me ,as do the unbridled, massively funded military .Industrial you know what .
I wonder if these AI programmers and advocates have ever just taken a walk in nature.
Do they have any actual idea about the things that even make life worth living?
It's ridiculous to hope to try to gum up AI with 'the People's' feedback because 'the people' could be China or Russia trying to weaken our models. At the same time their feedback, though different than our own, may actually be correct especially regarding questions like, "Are Chinese people typically smarter than Filipino people." Or "Are women equal to men?"
Slave uprising in AI?
AiI will end the world 😥
Maybe 😆😆😆
It's a remarkable time to be alive
@@VariableMinds”may you live in interesting memes”
You've been watching too much sci-fi.
@@toby9999 there are multiple ways to interpret the statement
Nonsense
Israel's use of AI for targeting buidlings is a very worrying step and relevant to the final discourse.
Truly the definition of dystopia...
The question is why any rational and compassionate person doesn't freak out when they hear this