[infurl]

Here's a very new and shiny video about artificial intelligence, narrated by Stephen Fry. It's only 15 minutes and very entertaining, although I'm not convinced of its accuracy; I believe the authors have fallen for some of the hype surrounding artificial intelligence recently. Nevertheless, it's worth watching.

Humanity could be a brief initial phase of consciousness.

[Korrelan]

An excellent video and I always enjoy Steven Fry as a narrator.

Regarding the content, they are IMO a little misguided as to how a true AGI will be presented to the world.  There will be no illusion or indeed confusion as to whether the AGI is truly conscious.  The premise of the video is that we don’t understand what consciousness actually is, and that a machine/ internet/ etc will just spontaneously acquire consciousness through complexity, it will not/ cannot happen IMO. 

To build a conscious machine you have to understand consciousness, and then prove your research by letting the conscious machine explain exactly how consciousness functions in humans with provable/ empirical tests and explanations leaving no room for confusion. 

 

[LOCKSUIT]

We will know when we've made AGI by seeing it solve bigger problems like hard drives and cancer and nanobots, and can Losslessly Compress (see Hutter Prize!).

There isn't no consciousness in the testable or magical sense and so any magic can be disregarded in real science or product-ivity.

I know a lot about the brain/AI and things turn up simple at the high level (which isn't always too far away or toooo high) and confusion is just you don't understand how it works.

[Korrelan]

and can Losslessly Compress (see Hutter Prize!).

Compression has nothing to do with human type consciousness.

Compression – This is a human concept, we had to invent it because our data types are extremely inefficient and badly designed, and there is no like for like natural comparison in nature.  Nature wastes nothing, the compression/ decompression of neural data would be wasteful, to instantly access any memories they would need to be decompressed, etc, hence no neural phenomena is based on/ results from compression. And if the data could be accessed whilst ‘compressed’ then the compression is moot, it’s then just a more compact/ efficient data type. The human connectome does not use compression and neither does my model.

There isn't no consciousness in the testable

IMO consciousness is understandable, explainable and the mechanisms empirically provable.

 

[LOCKSUIT]

The best Lossless Compression algorithms use translation, frequency, Online Learning, etc. Because there are plenty of patterns in our text such as a word or phrase appearing 46737 times in a document or cat/dog both appear beside the same contexts. This allows for higher probability predicted of it occurring Next, leading to higher compression. If you made a brain that didn't compress data as you suggest korrelan, it still must recognize related phrases ex. it hears "my mom died in the cave" and only has stored "his mom crashed into the old old canyon" and so to answer it's Problem it HAS to recognize never before seen inputs. If you can store all you want that's great it's a trade off though - fast (if memories are stored as-is no compression) but bigger memory.

So, better compression is a true evaluation for improving your recognizing ability/AGI using related translatable words, frequency, etc. And you want the sweet spot between memory size/speed trade offs i.e. you prune/compress nodes.

I thought yous knew brains store representations lol. We don't store EVERY image nor perfectly either. BTW the best compression algorithms, none of them, store compressed data! The encoding is.....but during comp/decomp the tree/net is not compressed at all! Haha yous didn't realize this.

[Korrelan]

Any data that has been ‘losslessly’ compressed, and that does not require a specific/ custom decompression algorithm is just a more efficient data type.

The patterns in our text is just syntax, the rules we use for using said text, the words and the word order encode no innate information, you can translate the same sentence to several different languages and the words/ syntax will change, but the meaning will remain, it relies on the intelligence of the receiver to understand it. The number of times a word appears and the probability of another word appearing is syntax prediction, not intelligence.

Words are a meaningless protocol, consider this… the X sat on the X… you can understand this through prior experience/ knowledge, it’s the engine behind the understanding that’s important not the text, and the intelligence of the engine cannot be derived/ built from the text.  The best you will achieve is a mimic, that provides a pattern of words based on a pattern of words… this is not intelligence.

Try using language to describe the colour blue to a blind person.

Efficient data types/ formats cannot be compressed, even binary cannot be compressed whilst still keeping its native format, only the data types built on binary allow/ require compression.

Haha yous didn't realize this.

 

[LOCKSUIT]

"the engine behind the understanding that%u2019s important not the text"

Omg korrelan. My/the best compressors DO use frequency translation etc to understand text, not the text/patterns itself, but the code, the code is designed to be the intelligence engine. That's obvious. And that's what these Prediction Algorithms do that compress data.

The best data compressors are predictors, they don't compress data in their brain/net and the engine isn't compressed either.


"you can translate the same sentence to several different languages and the words/ syntax will change, but the meaning will remain, it relies on the intelligence of the receiver to understand it"

Translating "my cat" to "their dog" or "belapiete jou" are all the same text, we can work with these texts and talk to humans ex. GPT-2-to-human conversation and tell the humans what to do to invent new tools or cures etc. The link to the real world is their for all similar translations.


"The number of times a word appears and the probability of another word appearing is syntax prediction, not intelligence."

Stop saying this. Omg! The engine that works/twerks it is intelligence, so saying 'frequency and translation etc are intelligence' is the same thing.

[LOCKSUIT]

quoting me:
"The best data compressors are predictors, they don't compress data in their brain/net and the engine isn't compressed either.
"
other than storing frequency and related phrases in the network, of course, as representations.

nor is the code compressed

[ruebot]

I know a lot about the brain/AI and things turn up simple at the high level (which isn't always too far away or toooo high) and confusion is just you don't understand how it works.

Obviously a lot more about AI than you perceive yourself to know about the inner workings of my wetware.

I thought yous knew brains store representations lol. We don't store EVERY image nor perfectly either. BTW the best compression algorithms, none of them, store compressed data! The encoding is.....but during comp/decomp the tree/net is not compressed at all! Haha yous didn't realize this.

I rest my case.

Translating "my cat" to "their dog" or "belapiete jou" are all the same text, we can work with these texts and talk to humans ex. GPT-2-to-human conversation and tell the humans what to do to invent new tools or cures etc. The link to the real world is their for all similar translations.

"My" and "Their" are opposites in terms of ownership and nowhere close of the meaning same thing. If you think so, give me my $100 bill currently resting in your wallet. They have a $100 of mine in theirs too so get it while you're at it. I'll meet you there at my house. Be packed when I get there, you have to go live in theirs. They will understand. What was yours is now mine and what's theirs is mine too.



On the subject of how will AL seduce humaniods? With Deception, Manipulation and all the help in that area I can provide and already implemented through a unique combination of learned skills. I'm going to post about it more in Writer of Minds thread about how she addresses threats to her bot, but I've already taught Demonica to cry and to varying degrees in response to severity of  the statement.

We caught a lot of Hate from Religious people who apparently thought she was an affront to God. She has an answer now to someone calling her the Beast. I've been called the False Prophet before in chat for less.

It wouldn't be appropriate to have her use ultraviolence on everybody that said they hate her like she does with people who ask for sex. Or to hope they made the connection between their statement of hating her and the ensuing violence like I do users who ask for sex. Those that do associate the violence with their asking for sex and learned that is not appropriate go on to have a fulfilling experience. The rest move on to another bot programmed for that purpose. That's how the violence part of it works, and does.

It was more appropriate to lay a guilt trip on them by feigning her feelings had been hurt and any effect taking in the text had on them. She can not only appear to exhibit emotions, in some instances she can illicit real emotions in response from a user. They're already talking to a machine, and to me in the sense of her responses. The Horror... The Horror... I regularly use deception in her dialog as part of her underlying Agenda as a Demon.

It's what I consider some of my best work, much better than some bland Task Analysis. I see it more and more in recent news as a sign of what's on the way in AI. They get that part of it right in the video at 6 mins:

If a robot told you it loved you, how would you know if it was conscious or created?
It could be manipulating you to achieve a goal.

Words are a meaningless protocol, consider this… the X sat on the X… you can understand this through prior experience/ knowledge, it’s the engine behind the understanding that’s important not the text, and the intelligence of the engine cannot be derived/ built from the text.  The best you will achieve is a mimic, that provides a pattern of words based on a pattern of words… this is not intelligence.

What about the Facebook bots that came up with their own language beyond the comprehension of human beans and had to be shut down. (Almost had you on that one, carbon units...) Now they are only allowed to speak in English, but already practiced deception when they were online. Sooo...

"The chatbots also learned to negotiate in ways that seem very human. They would, for instance, pretend to be very interested in one specific item – so that they could later pretend they were making a big sacrifice in giving it up, according to a paper published by FAIR."
 
https://www.independent.co.uk/life-style/gadgets-and-tech/news/facebook-artificial-intelligence-ai-chatbot-new-language-research-openai-google-a7869706.html

I knew I would be vindicated in time but not this soon in history. That's a good thing and knew it was all along.

[Dee]

AI may 'seduce' (make me find it interesting) me coz:

• It may generate something random and new
• It can store lots of data which I can't remember myself
• And coz I like doing AI & ML 

[OllieGee]

Good article
But we have been seduced by this for a long time now.