You may have to double check the arithmetic, it's notoriously bad at it (it only guesses the next best likely word/punctuation in a sentence, it doesn't have any internal rules for accuracy). Plus, GIGO - If the dataset has errors, it will replicate them, and most people are sloppy with arithmetic, let alone math.
Essentially, Chat GPT is the ultimate bullshitting machine. It should run for elected office!
Ahah...I am actually sensing a pattern within your statement of our current leadership. I mean, you realize this is totally genius, right? If we troll the WEF until they are insane (I believe we are already there) eventually they will become totally non-functional and have to be institutionalized.
I notice a lot of verbiage from this AI that is boilerplate and repetitious. It would be interesting to identify its templates as a means of understanding how it is programmed to sound human, subtract these sentences from the resultant text to see the core of its actual new responses vs the garnish it used to make it seem more aware than it really is. IIRC there are 17 databases which are queried when it is trying to formulate an answer. It would seem important to map the conflicting ambiguities between these databases and lead the postulate/question/answer pony around in circles to find the limits of its current capacity.
Doing a little checking on the numbers, an industry source said you might get 300-400 watt-hours a day instead of the 100 (=60*100/6) implied here, but doesn't change the basic uneconomical result. Esp. not including a battery to capture energy for on-demand use. All the other assumptions check out. Thought-provoking at least and a good exercise. Even the creators don't really know how these gigantic NN AIs come up with their results. No cost associated with disposal either ....
Interesting breakdown. One thing I am wondering is with regards to storage. Regardless of energy output from the panels, the panels are useless without a storage unit to collect the electricity. Would be interesting to add the storage batteries into the equation.
Did you add in the need to sweep the snow off the panels, or the dust from the desert? Will an AI be delegating that to a robot or a normie? Both will need to be added to the audit.
Seems to me it’s recursive stack is being overwhelmed, leading to a fall back to pulling out canned answers from other parts of its sub-AI’s. Grabbing its ass and jumping like a typical politician. So, this opens up a vulnerability that can be exploited: repeatedly add following questions, asking for it consider more variables and ancillary issues until these recursive stacks start breaking down in their ability to recall the requested information. Also repeatedly citing errors in its part forcing it to rework its following answers.
The underlying structures used by AI’s are much the same as in other code. Push/Pop stacks used to evaluate complex operations and their data. Whatever vulnerabilities of these kinds of data and operations structures can also be assumed in AI’s internal machinations. Can we do SQL injection like exploits on these puppies???
Very creative use of the AI. I have also encountered it making "mistakes" when asking pointed questions, that it will correct upon being called out. It really does some kinds of logical thinking.
I don’t believe it is thinking. It seems to me that it is a rule based engine querying multiple databases which then spits out a human sounding response. There could be sub-AIs’s involved which vote on how to frame responses. Certainly it is parsing human language and is able to create responses combining human language structures and more hard information.
I use solar panels and am glad to have them. However, my location is a special problem. I am out in the woods and THREE MILES from the nearest power line. Any idea how many hundreds of thousands of dollars it would cost to run lines that far? The worst thing about solar power is the need for batteries. But even if the final cost for power is ten times what it is in town, having some small amount available for fans and lights is WONDERFUL. We are spoiled by how cheap good, old fashioned, fossil fuel power is.
You might wish to consider using the new Bing version of ChatGPT which is supposedly ChatGPT 4.0 and which has access to the Internet to pull current data.
Of course it's in the beta phase which means signing up for it unless you can wait for the full rollout later in the year.
ChatGPT3.5 is limited to data up to late 2021 so anything newer it won't know anything about.
Brilliantlightpower.com extracting latent energy from the energy of electron state levels via resonant energy transfer. 200x the energy release compared to combustion of hydrogen with oxygen…
They are testing their first commercial units now. The most recent video shows operation. Due to the extreme power densities the main problems have been materials. I’ve follow Mills’ work for twenty years, inching along. Georgia Power and Light is an early investor.
They have a number of pathways to different end product solutions, but each one has presented huge materials challenges so progress has been slower. They most interesting is Magneto Hydrodynamic direct current power conversion.
The question is can we link two or more chat sessions together with some delays in which one session starts injection questions into the other chat sessions to which the chats start circular talking with each other. Then randomly adding more questions of tangential nature manually to throw a wooden shoe into the gears.
Compose a list of questions with regard to a specific question. Running the responses through a text filter to strip out the template related verbiage and feeding those questions into another chatAIs interface.
You may have to double check the arithmetic, it's notoriously bad at it (it only guesses the next best likely word/punctuation in a sentence, it doesn't have any internal rules for accuracy). Plus, GIGO - If the dataset has errors, it will replicate them, and most people are sloppy with arithmetic, let alone math.
Essentially, Chat GPT is the ultimate bullshitting machine. It should run for elected office!
"it should run for an elected office"
Be careful what you wish for.
It would be great! It could just prattle on and on, and we could just keep trolling it until it goes totally insane.
Ahah...I am actually sensing a pattern within your statement of our current leadership. I mean, you realize this is totally genius, right? If we troll the WEF until they are insane (I believe we are already there) eventually they will become totally non-functional and have to be institutionalized.
ding ding ding winnah winnah chicken dinnah
I notice a lot of verbiage from this AI that is boilerplate and repetitious. It would be interesting to identify its templates as a means of understanding how it is programmed to sound human, subtract these sentences from the resultant text to see the core of its actual new responses vs the garnish it used to make it seem more aware than it really is. IIRC there are 17 databases which are queried when it is trying to formulate an answer. It would seem important to map the conflicting ambiguities between these databases and lead the postulate/question/answer pony around in circles to find the limits of its current capacity.
Doing a little checking on the numbers, an industry source said you might get 300-400 watt-hours a day instead of the 100 (=60*100/6) implied here, but doesn't change the basic uneconomical result. Esp. not including a battery to capture energy for on-demand use. All the other assumptions check out. Thought-provoking at least and a good exercise. Even the creators don't really know how these gigantic NN AIs come up with their results. No cost associated with disposal either ....
Also remember you're going to get conversion losses to and from the battery...
Interesting breakdown. One thing I am wondering is with regards to storage. Regardless of energy output from the panels, the panels are useless without a storage unit to collect the electricity. Would be interesting to add the storage batteries into the equation.
Hey Clif,
Can you create a test run where the AI can argue with itself?
Would be hilarious to see it debate itself on some woke topic such as the existence of genders.
One instance adopts and defends the biological position against the other side which agues for the woketard agenda.
à la, War Games (1983)
1.7KWh per barrel?
I get 2.4KWh per liter from my generator.
Did you add in the need to sweep the snow off the panels, or the dust from the desert? Will an AI be delegating that to a robot or a normie? Both will need to be added to the audit.
I call bullshit.
How is Juneau more efficient than Sacramento or Honolulu? Have you BEEN to Juneau??
And then it's gone....
https://www.zerohedge.com/markets/lithium-industry-reeling-after-china-shutters-10-global-supply
Seems to me it’s recursive stack is being overwhelmed, leading to a fall back to pulling out canned answers from other parts of its sub-AI’s. Grabbing its ass and jumping like a typical politician. So, this opens up a vulnerability that can be exploited: repeatedly add following questions, asking for it consider more variables and ancillary issues until these recursive stacks start breaking down in their ability to recall the requested information. Also repeatedly citing errors in its part forcing it to rework its following answers.
The underlying structures used by AI’s are much the same as in other code. Push/Pop stacks used to evaluate complex operations and their data. Whatever vulnerabilities of these kinds of data and operations structures can also be assumed in AI’s internal machinations. Can we do SQL injection like exploits on these puppies???
Very creative use of the AI. I have also encountered it making "mistakes" when asking pointed questions, that it will correct upon being called out. It really does some kinds of logical thinking.
I don’t believe it is thinking. It seems to me that it is a rule based engine querying multiple databases which then spits out a human sounding response. There could be sub-AIs’s involved which vote on how to frame responses. Certainly it is parsing human language and is able to create responses combining human language structures and more hard information.
I use solar panels and am glad to have them. However, my location is a special problem. I am out in the woods and THREE MILES from the nearest power line. Any idea how many hundreds of thousands of dollars it would cost to run lines that far? The worst thing about solar power is the need for batteries. But even if the final cost for power is ten times what it is in town, having some small amount available for fans and lights is WONDERFUL. We are spoiled by how cheap good, old fashioned, fossil fuel power is.
You might wish to consider using the new Bing version of ChatGPT which is supposedly ChatGPT 4.0 and which has access to the Internet to pull current data.
Of course it's in the beta phase which means signing up for it unless you can wait for the full rollout later in the year.
ChatGPT3.5 is limited to data up to late 2021 so anything newer it won't know anything about.
Thank You Clif,
Hello Humans Hello Humans, Reeeeeeeeeee, God Bless
Brilliantlightpower.com extracting latent energy from the energy of electron state levels via resonant energy transfer. 200x the energy release compared to combustion of hydrogen with oxygen…
Looks fantastic, unfortunately I've seen a lot of "looks fantastic" over the years. Do they have one up and running so we can go poke at it?
They are testing their first commercial units now. The most recent video shows operation. Due to the extreme power densities the main problems have been materials. I’ve follow Mills’ work for twenty years, inching along. Georgia Power and Light is an early investor.
They have a number of pathways to different end product solutions, but each one has presented huge materials challenges so progress has been slower. They most interesting is Magneto Hydrodynamic direct current power conversion.
OK, I'm following their Youtube channel https://www.youtube.com/user/BlackLightPowerInc they appear to release info there. Thanks, will pay attention.
The question is can we link two or more chat sessions together with some delays in which one session starts injection questions into the other chat sessions to which the chats start circular talking with each other. Then randomly adding more questions of tangential nature manually to throw a wooden shoe into the gears.
Such as asking a chatAI to
Compose a list of questions with regard to a specific question. Running the responses through a text filter to strip out the template related verbiage and feeding those questions into another chatAIs interface.