Hi Jan technology has advanced a lot , chatgpt is like a single celled amoeba, there's tech out there that has no problem what you feed it. You just need to feed it core principles and it can ingest all the garbage out here and produce gold. More advanced systems don't care what it ingests , it will produce the truth from the mess👍
Lets imagine for a moment you are mentally or physically disabled or are old and need assisted living, under these conditions it makes sense to externalize intelligence and processes. Take Japan or dessert Arab countries with the money or the tech or both for example , this could be a great place to setup prototypes for automated farming. Self running sustainable farms could ensure a year round food supply. Maintenance and building robots could work with mining and processing robots to build and maintain our infrastructure. These prototype cities could meet the basic needs of food, shelter/clothing. We will be one step closer to eliminating poverty. In doing so we also create other problems for the enabled populace , health issues, obesity etc so we may need the enabled populace to workout to get food and shelter. Money fosters world peace and social interaction, as you get to talk and deal with people who you might have ignored were there not the exchange of money its like a bull ring with it you get a hook to train any beast. Our imagination is the limiting factor
It can't perceive what it can't steal from us. It would take a whole freakin' server farm to match my fart. they need to go full quantum to get anywhere
near what a devoted organic can do. F AI !!
If it was up to me , I would outlaw this tech - death as a penalty. Because organics can't be trusted with it .... at our present level.
See for yourself: https://chat.openai.com/
v3.5 can be fed back that it responded badly when it does.
v4 is more advanced and can access plugins and updated Web info.
There are even more advanced versions. It is possible to teach them on curated, more accurate documents as well.
BTW, I don't know when the human feedback received for incorrect answers is incorporated and made available back to us in v3.5 - I haven't checked. It might be something that is available in the next minor upgrade, e.g. v3.6
So I could feed back to it that the theory of a flat earth is the only correct earth theory and that the globe theory is a sham?
Jan
At the risk of being political: we're way past the point where we could eradicate poverty.
We just choose not to and instead continue to gather personal wealth to the extreme.
All the ChatGPTs of the world will not change that.
Jan
You could, but you'd be wrong to think your feedback isn't further verified for correctness before it is actually used for enhancement.
Even this can be automated... using ChatGPT 4...
You miss the point. Who will decide what is right and what is wrong? You find something that's 'right' according to you, and I'll get you someone who proves the opposite.
Jan
Interesting. When I made a chessprogram with AI, in LISP, linguistic models were being used in AI research to detect the political bias of a text.
In that sense, all texts can be seen as political.
If you don't harness one yourself, you're at the mercy of somebody else.
If you don't have a plan for this... someone else has a plan for you.
Humans have always adapted to new technologies, from the stone age till today.
Academics, scientists and big tech involved in the creation of AI are really the guys who know to some extend what AI is capable of and ask for regulation, for they know that this technology outsmarts them and they seem to loose their grip on it. Never before has something being created that has the potential to dominate us on all fronts. In all our existence we never had a technologie smarter than combined humanity, or at the very least combines all human knowledge.
I wonder if someone build an amplifier with AI schematics 🙂
Academics, scientists and big tech involved in the creation of AI are really the guys who know to some extend what AI is capable of and ask for regulation, for they know that this technology outsmarts them and they seem to loose their grip on it. Never before has something being created that has the potential to dominate us on all fronts. In all our existence we never had a technologie smarter than combined humanity, or at the very least combines all human knowledge.
I wonder if someone build an amplifier with AI schematics 🙂
Agree. We had a civil service that used AI to get an answer for an application - the rules were difficult and there was quite some 'grandfathering' (a public service term) so humans did take a lot of time.
But then - there was a law suit, and nobody could explain the outcome. And that AI program was within a fixed domain, only fed with past cases in the field.
Like someone said above, the big problem is first to train ChatGPT in our domain. The one who does that will NOT trust the outcome, I'm sure. But many will admire it. And others 🤑 the bucks that the new garbage generates.
But then - there was a law suit, and nobody could explain the outcome. And that AI program was within a fixed domain, only fed with past cases in the field.
Like someone said above, the big problem is first to train ChatGPT in our domain. The one who does that will NOT trust the outcome, I'm sure. But many will admire it. And others 🤑 the bucks that the new garbage generates.
AI finds another use:
https://consumer.ftc.gov/consumer-a...use-ai-enhance-their-family-emergency-schemes
https://consumer.ftc.gov/consumer-a...use-ai-enhance-their-family-emergency-schemes
I did a few tests for power circuits. It can revise versioned file formats, but the contents so far aren't usable. Take into account these were early tests with no fine-tuning or encoding of example files and descriptions, so prompting might have been lacking in the first place.
I'd say it is a promising avenue.
The way forward will be acting on curated information from a LLM base.