[hamdani yusuf]

A tad bit Worrying...

Nobody knows how it Works.
Opens up the possibility of it Miscalculating.

If 99 times out of 100, it says 2+2=4, & gets it wrong once, is Totally Fine.

Can only Hope autonomous ai systems won't be deployed for Nukes.

99.999% accuracy with an Error margin of just 00.001% will be
Game Over!

Humans have even higher error rates. Nonetheless, we survive, so far. Even with nukes available for decades.
What's needed is some reliable system for check and balance, and acceptable common goals.

[Zer0]

A tad bit Worrying...

Nobody knows how it Works.
Opens up the possibility of it Miscalculating.

If 99 times out of 100, it says 2+2=4, & gets it wrong once, is Totally Fine.

Can only Hope autonomous ai systems won't be deployed for Nukes.

99.999% accuracy with an Error margin of just 00.001% will be
Game Over!

Humans have even higher error rates. Nonetheless, we survive, so far. Even with nukes available for decades.
What's needed is some reliable system for check and balance, and acceptable common goals.

https://en.m.wikipedia.org/wiki/Stanislav_Petrov

[hamdani yusuf]

I tried to use Google Bard to write the opening chapter of a novel series that I've been thinking on for a few years. First, I told it the background of the story, the preferred writing style, and the main plot. It sounds supportive and made useful suggestions. But when it was asked to write the first chapter, it blows up and started to hallucinate. It doesn't seem to understand when I tried to correct its mistakes. It kept making the same mistakes.

[alancalverd]

It kept making the same mistakes.

And doubtless reinforcing them because it believes what it has already written!

[hamdani yusuf]

It kept making the same mistakes.

And doubtless reinforcing them because it believes what it has already written!

I'm not sure if Bard has a rigid boundary between training mode and deployment mode. I've read some articles describing AI models which can still learning from new data even when they're running in deployment mode.

[hamdani yusuf]

BTW, here's a nice explanation of transformer model, which is the basic of Chat GPT.

Transformer Neural Networks, ChatGPT's foundation, Clearly Explained!!!

[hamdani yusuf]

ChatGPT - Semantics: Transformers & NLP 2

How do transformers like ChatGPT learn and represent words?

Transformers are a type of neural network architecture that are used in natural language processing tasks like language translation, language modelling, and text classification. They are effective at converting words into numerical values, which is necessary for AI to understand language. There are three key concepts to consider when encoding words numerically: semantics (meaning), position (relative and absolute), and relationships and attention (grammar). Transformers excel at capturing relationships and attention, or the way words relate to and pay attention to each other in a sentence. They do this using an attention mechanism, which allows the model to selectively focus on certain parts of the input while processing it. In the next video, we will look at the attention mechanism in more detail and how it works.

We can encode word semantics using a neural network to predict a target word based on a series of surrounding words in a corpus of text. The network is trained using backpropagation, adjusting the weights and biases of the input and hidden layers until the updates become negligible and the network is said to be "trained". The weights connecting the input neurons to the hidden layer will then contain an encoding of the word, with similar words having similar encodings. This allows for more efficient processing and a better understanding of the meaning and context of words in the language model.

'Attention is all you need' paper - https://arxiv.org/pdf/1706.03762.pdf

=========================================================================
Transformers are a type of artificial intelligence (AI) used for natural language processing (NLP) tasks, such as translation and summarization. They were introduced in 2017 by Google researchers, who sought to address the limitations of recurrent neural networks (RNNs), which had traditionally been used for NLP tasks. RNNs had difficulty parallelizing, and tended to suffer from the vanishing/exploding gradient problem, making it difficult to train them with long input sequences.

Transformers address these limitations by using self-attention, a mechanism which allows the model to selectively choose which parts of the input to pay attention to. This makes the model much easier to parallelize and eliminates the vanishing/exploding gradient problem.

Self-attention works by weighting the importance of different parts of the input, allowing the AI to focus on the most relevant information and better handle input sequences of varying lengths. This is accomplished through three matrices: Query (Q), Key (K) and Value (V). The Query matrix can be interpreted as the word for which attention is being calculated, while the Key matrix can be interpreted as the word to which attention is paid. The eigenvalues and eigenvectors of these matrices tend to be similar, and the product of these two matrices gives the attention score.

[Petrochemicals]

Apparently the hoo-ha about AI is speculative about the future or what they think is in the pipe line, at present they have Narrow AI which is basically a computer programme.

https://codebots.com/artificial-intelligence/the-3-types-of-ai-is-the-third-even-possible

[hamdani yusuf]

Apparently the hoo-ha about AI is speculative about the future or what they think is in the pipe line, at present they have Narrow AI which is basically a computer programme.

https://codebots.com/artificial-intelligence/the-3-types-of-ai-is-the-third-even-possible

It's an old article, which was written before ChatGPT and its competitors. It's highly likely that the author has changed his opinion by now.

24 October 2017 ? 11 minutes

Written by Eban Escott

At some point, future AI models will be able to rewrite its own source code. Some of them will have access to physical world through I/O interfaces like cameras and robotics.
They won't be a monolithic entity. They will form an ecosystem of AI models instead. They will compete and cooperate with one another trying to achieve some terminal goals. IMO, sooner or later those goals will inevitably converge to get closer to the universal terminal goal.

[Petrochemicals]

Apparently the hoo-ha about AI is speculative about the future or what they think is in the pipe line, at present they have Narrow AI which is basically a computer programme.

https://codebots.com/artificial-intelligence/the-3-types-of-ai-is-the-third-even-possible

It's an old article, which was written before ChatGPT and its competitors. It's highly likely that the author has changed his opinion by now.

What opinion? Chat gpt is the same level "AI" as spam filters.

[hamdani yusuf]

What opinion? Chat gpt is the same level "AI" as spam filters.

If the third type is possible. It's right there in the title.
The accuracy of the AI models are measurable, and can be compare with each other, including their own older version and varians, as well as average humans and experts. Most researches show accelerating progress. AI arms race is happening in front of us, but some of us are willfully ignorant.
They remind me to analysts who thought that it would take centuries before computers can conquer the game of Go, even after deep blue beat Kasparov.

[Petrochemicals]

If the third type is possible. It's right there in the title.

There is only one form of "AI" at present the others are theoretical. At present it is a targeted computer programme. The title is

What are the 3 types of AI? A guide to narrow, general, and super artificial intelligence

[hamdani yusuf]

If the third type is possible. It's right there in the title.

There is only one form of "AI" at present the others are theoretical. At present it is a targeted computer programme. The title is

What are the 3 types of AI? A guide to narrow, general, and super artificial intelligence

The same argument can be applied against human. Currently, there's no human general intelligence. No human individual has average expertise in every field of intelligence. It's even impossible for anyone to beat every expertexpert in every field.
We can argue that humans have inherent physical limitations of compute capacity, which is caused by the size of their brains. Input and output bandwidth put other limitations. AI can increase their intelligence without those limitations.

Apparently the hoo-ha about AI is speculative about the future or what they think is in the pipe line, at present they have Narrow AI which is basically a computer programme.

https://codebots.com/artificial-intelligence/the-3-types-of-ai-is-the-third-even-possible

[hamdani yusuf]

Basically, in every frontier of technology, there are two main groups of people. The first are those who think that it's impossible. The other are those who disagree, and are motivated to prove that their opponents are wrong.

[Petrochemicals]

If the third type is possible. It's right there in the title.

There is only one form of "AI" at present the others are theoretical. At present it is a targeted computer programme. The title is

What are the 3 types of AI? A guide to narrow, general, and super artificial intelligence

The same argument can be applied against human. Currently, there's no human general intelligence.

Yes there is, well at least I am, I can use the Internet, then paint a picture, then think about dinner. General.  I can learn, not something AI is capable of, unless you have a general AI you are not telling the world about.

[Petrochemicals]

Basically, in every frontier of technology, there are two main groups of people. The first are those who think that it's impossible. The other are those who disagree, and are motivated to prove that their opponents are wrong.

No one is saying it is impossible, just like nuclear fusion in a sustained reaction, or the atom bomb in 1940, perpetuated flight in the days of da vinci, it just isn't currently present. Chat gpt has specific programming to search the internet and rehash information, if I ask it to create a picture it will not be able to as it is programmed to rehash text, not images. Also I cannot ask it to" try again differently", it requires specific input, otherwise the programme goes out of control, not general AI. AI bots have been known to stage wars

https://www.theguardian.com/technology/2017/feb/23/wikipedia-bot-editing-war-study

[hamdani yusuf]

Yes there is, well at least I am, I can use the Internet, then paint a picture, then think about dinner. General.  I can learn, not something AI is capable of, unless you have a general AI you are not telling the world about.

What makes you think that AI cannot use internet? Or do any of those things you mentioned?
Can you detect cancer from radiograph? Or do brain surgery?
Don't you think that search engine algorithms are AI?
How do you define learning?

[Petrochemicals]

Yes there is, well at least I am, I can use the Internet, then paint a picture, then think about dinner. General.  I can learn, not something AI is capable of, unless you have a general AI you are not telling the world about.

What makes you think that AI cannot use internet? Or do any of those things you mentioned?
Can you detect cancer from radiograph? Or do brain surgery?
Don't you think that search engine algorithms are AI?
How do you define learning?

I can do anything I wish given enough time, anti gravity, nuclear fusion etc. I can learn, I realise I need to understand something better and try to ascertain the knowledge to do so. This is what general AI will understand also. Chat gpt refuses to paint no matter how much you ask it, perhaps its goung through adolescence.

[hamdani yusuf]

I can do anything I wish given enough time, anti gravity, nuclear fusion etc. I can learn, I realise I need to understand something better and try to ascertain the knowledge to do so. This is what general AI will understand also. Chat gpt refuses to paint no matter how much you ask it, perhaps its going through adolescence.

Can you fly to outer space?
Newer AI models started to become multi modal. At least OpenAI, Google and Tesla had already announced their development. They will include embodiment and agency.

[hamdani yusuf]

How do you define learning?

Better yet, you can watch Neural Networks Learning in this video.

Timestamps
(0:00) Functions Describe the World
(3:15) Neural Architecture
(5:35) Higher Dimensions
(11:55) Taylor Series
(15:20) Fourier Series
(21:25) The Real World
(24:32) An Open Challenge