ClaytonB
Member
- Joined
- Oct 30, 2011
- Messages
- 10,227
We are in the AI Singularity
- Thread starter ClaytonB
- Start date
acptulsa
Member
- Joined
- Jan 2, 2008
- Messages
- 75,404
This should be a Bee parody.......
Yes, Newsmax is a parody site like the Bee. If all parodies had to come from the same provider, we'd all be reading the Onion, and never would have heard of the Bee.
ClaytonB
Member
- Joined
- Oct 30, 2011
- Messages
- 10,227
OK Boomer
acptulsa
Member
- Joined
- Jan 2, 2008
- Messages
- 75,404
- Joined
- Nov 5, 2010
- Messages
- 39,958
//
This either makes perfect sense or no sense at all. I'm not sure which.
https://twitter.com/backtolife_2023/status/1654545860120788994
Swordsmyth
Member
- Joined
- Apr 14, 2016
- Messages
- 74,737
Google Working On AI Tool Which Will Write News Stories And Headlines
https://www.thefinancialtrends.com/...-which-will-write-news-stories-and-headlines/
https://www.thefinancialtrends.com/...-which-will-write-news-stories-and-headlines/
ClaytonB
Member
- Joined
- Oct 30, 2011
- Messages
- 10,227
Gemini is truly scary. People are not ready for this. Unless I've missed my guess, Gemini is going to effectively be like talking to Data from Star Trek. Yes, it will track with the subject of a conversation and not run off into the weeds like ChatGPT. Yes, it will be able to solve logic puzzles, including puzzles involving spatio-temporal reasoning. It will seem to have "insight", "intuition" or even "creativity". It will still clearly not be human, but I am quite confident that it will possess all of these attributes, because they are adding most of the ingredients for general-purpose intelligence that ChatGPT is missing.
*sigh
ClaytonB
Member
- Joined
- Oct 30, 2011
- Messages
- 10,227
This is BIG:
“Low-Resource” Text Classification: A Parameter-Free Classification Method with Compressors
It might not sound like much, but this is a canary-in-the-coalmine for AGI. If you're not facile with machine-learning terminology, this might sound a lot more complicated than it is. In the video, he expresses astonishment that this technique works, but it's not too difficult to understand why it works. Let's say we have two identical positive reviews: A:"This movie was amazing!", B:"This movie was amazing!" Both are obviously marked "positive" sentiment in the training data-set. When we concatenate the strings together, "This movie was amazing!This movie was amazing!" and compress it, the compression algorithm is smart enough to recognize that the string has simply been repeated and will encode it that way: "[This movie was amazing!]\r" where I am using "\r" to represent whatever escape-code the compressor uses to indicate "repeat the last string in brackets." It won't be quite that simple in the compressed format, but this is one of the capabilities of any SOTA compressor. So, when you measure the "length" of the two compressed strings when joined, it will be barely longer than either of the original strings. So, the compression algorithm is measuring commonality, but it's doing it in a very sophisticated way.
Let's take two opposite reviews. A:"This movie was amazing!" B:"This movie was horrible!" Again, the compressor is smart enough to recognize that "This movie was " is common to both strings. So, it will use some short code to repeat that: "[This movie was ]amazing!\rhorrible!" It's not exactly like this, but this is the basic concept of what is happening internally. And the compressor is much more sensitive to repetitions and statistical patterns than I am describing here.
In information theory, this kind of measure is called the mutual information. And that's really what ncd is acting as a proxy for.
We are on our way to a computable approximation of AIXI (general-purpose AI with provably optimal behavior)...
“Low-Resource” Text Classification: A Parameter-Free Classification Method with Compressors
It might not sound like much, but this is a canary-in-the-coalmine for AGI. If you're not facile with machine-learning terminology, this might sound a lot more complicated than it is. In the video, he expresses astonishment that this technique works, but it's not too difficult to understand why it works. Let's say we have two identical positive reviews: A:"This movie was amazing!", B:"This movie was amazing!" Both are obviously marked "positive" sentiment in the training data-set. When we concatenate the strings together, "This movie was amazing!This movie was amazing!" and compress it, the compression algorithm is smart enough to recognize that the string has simply been repeated and will encode it that way: "[This movie was amazing!]\r" where I am using "\r" to represent whatever escape-code the compressor uses to indicate "repeat the last string in brackets." It won't be quite that simple in the compressed format, but this is one of the capabilities of any SOTA compressor. So, when you measure the "length" of the two compressed strings when joined, it will be barely longer than either of the original strings. So, the compression algorithm is measuring commonality, but it's doing it in a very sophisticated way.
Let's take two opposite reviews. A:"This movie was amazing!" B:"This movie was horrible!" Again, the compressor is smart enough to recognize that "This movie was " is common to both strings. So, it will use some short code to repeat that: "[This movie was ]amazing!\rhorrible!" It's not exactly like this, but this is the basic concept of what is happening internally. And the compressor is much more sensitive to repetitions and statistical patterns than I am describing here.
In information theory, this kind of measure is called the mutual information. And that's really what ncd is acting as a proxy for.
We are on our way to a computable approximation of AIXI (general-purpose AI with provably optimal behavior)...
Last edited:
- Joined
- Nov 5, 2010
- Messages
- 39,958
ClaytonB
Member
- Joined
- Oct 30, 2011
- Messages
- 10,227
Content starts at 3:30:
Swordsmyth
Member
- Joined
- Apr 14, 2016
- Messages
- 74,737
Microsoft Is Betting Heavily On The Future Of AI
https://www.thefinancialtrends.com/2023/08/13/microsoft-is-betting-heavily-on-the-future-of-ai/
https://www.thefinancialtrends.com/2023/08/13/microsoft-is-betting-heavily-on-the-future-of-ai/
- Joined
- Nov 5, 2010
- Messages
- 39,958
AI david attenborough narrates a documentary about my cat...
https://www.youtube.com/watch?v=eTb-7Gn-63o
https://www.youtube.com/watch?v=eTb-7Gn-63o
- Joined
- Nov 5, 2010
- Messages
- 39,958
This AI needs to stop immediately
As AI rapidly improves with Midjourney and ChatGPT, other forms of Artificial Intelligence are advancing that we do not see as much. As people are using AI art and social media to find a way into making money, But AI social media models and infulencers are probably the least of our worries now.
https://www.youtube.com/watch?v=lXdCOLSQqXs
As AI rapidly improves with Midjourney and ChatGPT, other forms of Artificial Intelligence are advancing that we do not see as much. As people are using AI art and social media to find a way into making money, But AI social media models and infulencers are probably the least of our worries now.
https://www.youtube.com/watch?v=lXdCOLSQqXs
Similar threads
